you should make this open source so people can use and tweak this to make it more and more accurate and interesting if they want to would be very cool. great video btw
would love to see if someone could add more variables to this. I.e finishing position, gap to teammate and the cars performance relative to the rest of the field. I think that including some of these would make the system a lot more accurate in comparing drivers from different teams
@@hbrown689 I'd add finishing position, gap to teammate, car performance relative to both the first (or second, if their car was the fastest) and last cars. I would also weight their age and number of races against their teammates, since we know that for a rookie, especially, there's an adjustment period, and they won't receive the same treatment as the more experienced driver in the team, and we also know that, especially before the 1990s, age was a huge factor in performance, as they did not treat themselves as athletes the way they do nowadays. It would help give things like Schumacher's, Alonso's and Hamilton's rookie performances a bit more of nuance, as beating your teammate as a rookie is an unbelievable achievement, as well as Hamilton's and Alonso's current performance despite their age, compared to their much younger teammates. And as much as it could skew things a bit, I'd remove DNFs. Whereas nowadays most of the DNFs are driver's faults, historically, and particularly for the drivers before the 2010s, mechanical failures were so much more frequent that they could have DNFs for many races in a row, by no fault of his own (see Kimi Raikkonen 2005, for example). I guess in the end, driver's mistakes would kind of cancel each other out, for the most part. I would honestly be more than happy to be a part of a group willing to undertake this (it would be too much work for a single person). In the end, it would be a better measurement, but it would still let out a few subjective things, unfortunately, like the differences between how teammates were treated in certain eras, like, for example, in the 1990s and 2000s, for example, teams generally had very clear "2nd drivers" who would get development parts later,
This guy should make the rating system to compare against all the ninteen other drivers in the grid, than only your teamate, when then we would get honest elo rating on the board. Fangio was back in the day atleast dominating his teamate so hard, that he got so high elo rating, while Alonso drove for so long, that he in time to got to that high elo rating.
@@jout738 Agreed, just because it would be cool what would show up. However, we don't get the full picture. For example, a driver has a mechanical failure or race ending crash.
Thats why Max is the best this sport has ever seen, Senna with “only” 3 WDC’s proves that this sport is not about statistics, but 18 year old Max did that 30 laps Spain 2015 against Kimi in a Ferrari, in a car Max drove for the first time. We all know who the best this sport has ever seen is. Max is a cleaner and more consistent Senna
@@yerrie1908Driver complained about Max for years. He has since matured into a great champion but be real. They called that boy Crashstappen for a reason😂
as a programmer, I know it really pisses me off when someone point out something i overlooked after I am finished. Sadly, I’m here to ask you a question. Did you compare the teammates in races where one of them had a mechanical failure? Because that could mean the car is taken into account…
There's so many things that definitely weren't taken into account, "equal machinery" only goes so far, most of the times the team will prioritize a certain driver or car and put better parts, or better set up the car, there's also strategy and pit stop priorities to be taken into account. Drivers with a long career and that are known to enforce the "number one driver" in their contracts (like Alonso) will likely get a better elo not necessarily being the better driver
@@AdamastorPequeno01 you re absolutely right, but it would get way too complicated for a simple elo system, i think with this one it already took an eternity
In the early '80s, I was about five years old, my father was driving. Suddenly, someone made a bad maneuver in front of us and my father honked in protest. When the cars were parallel, the driver of the other car made a gesture of apology and my father, who had been mumbling some insult, instantly became silent and then…laughed. The old man in the other car was Fangio.
so back in the 80s, fangio locked up, your dad moved up the inside, gave him the peace sign, and fangio nodded in respect your dad is the goat of F1 confirmed, put him in a head to head with alonso and its gg
so back in the 80s, fangio locked up, your dad moved up the inside, gave him the peace sign, and fangio nodded in respect your dad is the goat of F1 confirmed, put him in a head to head with alonso and its gg
Please elaborate how this is a particular british thing, as I do not really associate chess with being british. As it is one of the few thing the britsh did not steal from India
@@QuixoticCowboy oh are we gatekeeping? I’ll make it painfully clear then; it’s British because likely only Brits would make this creative correlation and apply the system to it as both chess and F1 are obscenely popular in the UK. It’s about sports, not history. I’m not even British so cool your jets. It was a joke clearly not meant for you.
I already struggle to remember the Caro-Kann, Italian, Sicilian, Jobava London, and Ruy Lopez openings. Now you're telling me I have to remember the Verstappen, Alonso, Pérez, Hamilton, Räikkönen, Schumi, Button, Sebi, and how many others...
I don't watch Formula-1 and clicked on this video just to find out how would you apply the Elo system into such a sport. The video was so amazingly entertaining so I instantly forgot I was going to turn it off after the 30 seconds. The edit and the way of telling the information are astonishing. I would subscribe to your channel just because you deserve it. I really hope you'll have enough enthusiasm to make every video as cool as this one just because this one is perfect and you are, honestly, the most underrated content creator I've seen at 2023. Just keep doing this art, mate, you are the GOAT
Always knew Alonso was the GOAT sadly needed few more points to be 5wdc and better cars in his career to show how truly great he was/is. Alonso is one of those where stats don't show how good he is.
8 pts away from being a 5- time world champ, and if he made a bit better carreer choices maybe even more, basically if he had a bit more of luck hed probably have the most amount pf WDCs imo at least 7 or 8
@@nitr097sticking with ferrari until 17 and 18 would have DEFINITELY given him 25-ish more wins over the extra 4 years (the 16 car had no wins under vettel, so it isnt more) and probably would have gotten a WDC in 2018, it would have been close in 2017 but just maybe
@@nitr097 IF the team changes were perfect, alonso could be a 15 time champ And the team changer would be 2009 brawn which he was offered, 2010-2013 rebull which he also was offered, 2013 and 2014 2021 merc Of course these changes are impossible to do and to know with period information BUT with future info like now you could do it
This video perfectly shows what it's like to be a Software developer with the quadrillion edge cases you never thought about before you started the project.
Let's make a "K" in equation the difference between finishing positions. If teammates are 1-2 the K is 1 so it's little difference, but if they are finishing 3 and 12 K=9 (mayby we can always multiple that number by 5 or 6 or something like that to get higher difference in points)
@@insertcoolname96car performance is taken into account because you’re only comparing teammates here, if your teammate ended 2 and you ended 10 is because theres a clear skill gap, so I think this sistem is quite good
@@dtar380 Yes, but imagine your teammate ends up 1st and you second, 30 seconds behind him. Not because you are good, but because your car is so much better than the rest. I can see your point, but I think it could depend on the performance of your car.
I just posted a similar response before seeing yours. A weighted position doesn't say as much as time differential. As an example gap/elapsed time. If your team mate DNF's then the scale is 1.
The problem with this whole approach is that is not the ELO of f1 as a whole, but rather who won more over their teammates. The best way to try to represent this on an ELO would be like iracing rating. You enter with a fix amount of ELO, and depending on the finish position and everybody’s else’s elo will determine how much you gain or loose. For ex. A low elo driver can win a lot of points by finishing 7th but a high ELO driver can loose even by finishing 6th because rivals have low elo
Yeah I agree, this system is horrible, it shouldn't just be comparison with teammates (although it matters well). But even if we just take teammate comparison, the weighting is garbage.. Verstappen is going the greatest season of all time, beating Perez 18-2, and those 2 wins of perez take off about as much points off Verstapen as 6 Verstappen wins. For context, in both cases, Verstappen lost to Perez because of problems in qualifying.
Yes, my thoughts also. The system can be updated by taking into account the relative power level of each car and see how well the drivers compare to the power level of their car. If Ferrari has the best car by a mile, than getting into the podium should give fewer ELO points. Finishing in the points with a tractor, is some thing that should boost your ELO points. Also you need to take into account if mechanical failure/crashes was self-inflected or not. The formula would be a lot more complicated, yet, it seems like a nice subject for a thesis on data analysis.
This episode is so clever and such a unique cross topic narrative. You’re tweaking the YT algorithm to potentially capture nerds in F1, racing, chess, data science, and coding. Kudos!
2012 says why alonso is the goat, close to win a wdc with only 3 wins and a car that was consistently the 3rd fastest, the driver who got the only 2 poles of ferrari that year in wet conditions
What about an ELO system for the teams, which gets joined to the driver's ELO to calculate the gains/losses, but that remain separate so drivers can change teams and carry only their personal points. That way you'd be able to somewhat gauge the effect of having a superior car. Going even further maybe it would be interesting to reset team points whenever there's a big regulation change like in 2014, 2022, but use only the teams ELO at the end of the season to calculate the driver ELOs
I would love to see this as a public thing that updates every race/month/season. It will be so interesting to see the changes in elo after each race/season
this might be a horrible take but mabye shumacher just wasent allat, dont get me worng he was a beast but lets not forget two (almost three) of his titles where gained by crashing into his championship rival
Schumacher never had a high rated teammate under this system. Which is one flaw in this. Since it only looks at your teammate, beating a much higher rated rival in a different car (even if that car is equally good, or even better than yours) doesn't mean anything.
@@kukuc96this system also favours beating skilled drivers who are past their prime (exactly what alonso has done for the last decade). If you find a skilled driver, they have a high ELO, if they are past their prime you can beat them by a big distance and gain more points. No wonder alonso went to the top, it’s a unique situation for him.
As an engineer who doesnt code (or have the time to do the hard work you put into this), what I would love to see is an ELO weighted by how much (in seconds) a driver beat his team mate in each race - throwing out races with DNF's, where they make still have an ELO adjustment, but with a factor of 1.0 vs the time differential. Obviously it would have to be adjusted for the length of the race, so gap/elapsed time or something of the sort. This would give emphasis to the driver's abilities, even if they're in the fastest car. Such as Hamilton vs Bottas and Hamilton vs Rosbergwith with Merc, or Verstappen vs Perez with RB. MV is not surprising to anyone who doesn't worship Hamilton. He really is a phenom. Despite some boring racing, its wonderful watching him develop and truly master his craft. Laser precision and as repeatable as a robot. This will probably get me downvotes. But I'm not surprised that Hamilton wasnt anywhere close. All but Mclaren, his WDC's were in by fsr the fastest car, for most of those years every team in the top 10 was running a Merc powerplant. He excelled at being consistent, but had a fast and stable car to do it. The stand outs on talent have always been those who coupd consistently manage a pointy car, MV, Schumi, Prost, & Senna. Anyone on the grid can be fast on the fastest *stable* car. This also explains the massive gap MV has had to team mates. Alex Albon does a great job explaining it in his interview. Fantastic work and a fun watch. I'll sub.
It's a great start for a rank, there're loads more information needed to develop it, time interval between teammates (P1 and P2 is one thing P1 +20 seconds lead to P2 is another) increase elo for time diference; dnfs for driver error (how would one do it i don't know, so let's just ignore it); as you said qualifying; amount of times undefeated should have an increment to your score This is a amazing work!!
the issue with time interval is that in the 50s 20 seconds would be actually pretty good, so you'd have to adapt to every era somehow, factor in when cars are broken but not enough to DNF (faulty brakes, no power steering, gearbox...), how many laps the car has been broken, maybe add team strategies to win points for a championship, also team strategies blunders (hello ferrari)... there is so many variables at play that it's virtually impossible to solve.
@9t8268 That does add more impact from the cars and the relative balance of power of the teams though. It makes it harder for teammates who have far and away the best (or worst) cars to gain or lose points compared to drivers in midfield teams where there is a much greater chance of having several positions between the teammates. Without getting into car performance, I am not sure there is a whole lot you could do to improve this.
"Mr V's tests a new ranking system. He makes 2 teammates fight each other, 99000 teammates, 35 exact same persons, -1 drivers against each other, drivers in multiple different teams, drivers with the same position. A customer asks how many wins does a driver has, the excel promptly catches on fire."
For those who don't understand that joke, it's based on a stupid joke many devs love to tell : "a programmer enter a bar, and buys 1 drink, 50000000 drinks, -1 drink, apple drink, Null drink, ... A customer enters the bar ans ask for the bathroom, everything catches on fire."
I’d be interested to see the list with the 50s and 60s (and perhaps the 70s) removed. It’s was pretty much a different sport back then. Love the hard work on this!
I'm with you, but I'd take it in a different direction. Not necessarily "removed" but maybe split into 2 or 3 big eras? Something like the very early years, then a cut around the late 60s when aero started to become a thing and maybe another cut around the 90s/late 90s where safety became an even bigger concern, simultaneously giving us a pretty good "modern" era. Other people would probably find better split-points and you'd have the issue of how to deal with drivers that participated across multiple of these splits, but it would slightly alleviate the issue of comparing drivers between vastly different phases of the sport.
@@IguitarVreakIi think breaking it up by tech eras makes a huge amount of sense since the cars are so different between each one and require different driving styles. Early, aero and modern would be a good split.
@@kingofmemes5017like you said, each decade of car needs a different driving style. Adaptability is what makes a driver great. It may just be the biggest factor anyways. That means splitting era cars really only tells you who drives those style of cars best. Not the most accurate in finding out who is the actual best driver of all time. Just raising a point against yours. Honestly, I myself have no solution to this problem. I don't think there is any good solution.
6 wins x 6 points would make it 1350 indeed but that's without taking into account the other factors of finishing lower than some other drivers in Singapore for example.
It will require at least 10 consecutive victories before Verstappen gets to Fangio. And since Fangio could only start driving F1 when he was nearly 40...if that guy was 10 years younger we would have never needed a discussion on who'd hold the most WDC...that would have been Fangio till the end of time. Verstappen might however be the first in 73 years to match Fangio...the rest doesn't even come close.
@@JayQ2kAs Verstappen rises he slowly gets less points off Perez as well since the gap between their ELO goes up so it'll take a fair bit longer than 10 I think.
Very interesting work, well done. Obviously looking at teammate head-2-head's is limited to just finishing positions, and subject to who your teammate is (who's to say how the ratings of Fernando and Max would be impacted if Fernando joined RBR for a bit as an example). Also, those 'all time great' performances, or margin of dominance over your teammate isn't taken into account, nor are external factors. Still very fun exercise!
The ELO system was made with competitive pairings in mind. Verstappen gains a lot of points on Perez because Perez is not necessarily there to win the race, but to help Max win the race - they're teammates after all, they're not always competing against each other. Piastri as you pointed out is close to last because in many races McLaren call was to give Norris the lead even in races where Piastri was beating his teammate. I know this may sound a little nitpicky 'cause how else would you be able to normalize the competition across so many factors in another way, but I guess that's exactly my point - for ELO to work on F1 it would have to be a completely differente weighted average system
Hold on hold on, giving people ratings based on how they are compared to their teammate doesn’t quite work because what if they have a shitty teammate who’s only driving cause their dad owns the team
had to scroll way too long for this comment. all other smaller problems are obsolete when the basis of the formuma is that shaky… no offense to the guy tho obviously
This would be fine if most top drivers on the ELO ranking have matched up against a large variety of teams and teammates even if one of them was a shitty nepo driver. But if let's say Max was paired up with Lance Stroll from 2016 to 2024; indeed, that would make the ELO ranking unrealistic, but three years of Checo ain't too far.
Moreno was highly competent but also pretty much trounced by Piquet in the final years of Nelson's career... I guess the issue here is that it rates later wins against incompetent team mates too highly?
@@intheclarts You actually want the incompetent teammate when they are a rookie in this system. Let's say there is a driver has a final real rating of 800, but they still start with 1000 as rookies, so you can take more points off them when they are rookies. The other option is beating a high rated teammate when they are not that good anymore, but they still have a high rating from their past success. Although they are not in the top, I imagine Leclerc gained a good chunk of points from Vettel that way, and probably Nico Rosberg also got a bunch of points from Schumi.
@@kukuc96 Don't know, Piquet had plenty incompetent team mates, Brabham were famous for it. Not sure where he rates on the system, but that he would be below Moreno? Not sure how this computes. Beat every teammate up to Mansell, that was fairly level, then beat every teammate until Schumacher.
What I learned from this is that spec series race categories are ripe for ELO comparisons. With a spec series you can absolutely compare the entire field against each other for each race. Besides that one F3 season where a certain driver seemed to have an upgrade on their suspension compared to everyone else.
Nice video! Regarding the point about factoring in finishing positions to make the model more accurate, that's basically what F1metrics did. They also factored in other things such as age/experience and accounting for mechanical DNF's and Customer/Works cars. You should check it out (it hasn't been updated since 2019 unfortunately)
i love this as being more of a coding edge case demonstration rather than being fully focused on the F1 aspect of it i remember back in uni most of my projects were like "oh yeah, this works, im a genius!", then getting the professor's input file and seeing 90% of my code exploding because i didnt take into account barely any of the edge cases "nah, there wont be empty groups" "nah there wont be repeated inputs" "nah there wont be unnacceptable characters" oh boy, did i learn that year
Not surprised to see Button so high up, especially if we ignore qualifying and only look at the race results. He has been teammates with the two greatest drivers of his era and did extremely well against both of them.
I would gladly accept this. I respected and Feared Alonso because I Supported Michael Schumacher , Kimi Raikonnen & Vettel , I was against him. However, I can't deny his genius and the dude suffered so much coz of Machinery but still Races like a Beast. He's one of the 3 sportsmen that turned me into a fan even though i was rooting Against them in the begining Fernando Alonso, Roger Federer & Lionel Messi are those 3. Alonso is Great in F1 and his Other Racing achievements as well, He's one of the Greatest Racers Overall. There's a Study Published in Britain 2016, Showing he's 3rd Greatest Driver , #1 of Drivers still driving in 2016 and Your ELO Which shows Alonso is The GOAT. WORTHY IMO. I'd definitely say he's the greatest Racer who Debuted in 21st century.
Not a chance. Hamilton, Vettel, Verstappen, Russell all significantly better than Alonso. Alonso just always had very poor teammates in a car that suited his crappy driving style which massively overinflated his stats. He didn’t have a teammate the caliber of Leclerc, Russell, Rosberg, etc and he lost his first season against Button and only Won the next season because Button was in retirement mode. Alonso has lost to Tarso Marquez, Trulli, rookie Hamilton, Button, Ocon.
@@paperplane-db8qf if you understood anything of the code, you'd know that what you're pointing out is already accounted for. Losing for a low elo rookie is bad for his stats, yet he managed to keep it.
this system favours beating skilled drivers who are past their prime (exactly what alonso has done for the last decade). If you find a skilled driver, they have a high ELO, if they are past their prime you can beat them by a big distance and gain more points. No wonder alonso went to the top, it’s a unique situation for him.
in OTB chess, you start with no elo, after X matches, you get set an elo based on your results (for example you score 50% against an average opponent of 1423, your starting elo will now be 1423, you do not need to build up from 1000) this should solve the 'Piastri' problem in your video. To be more complete also the R factor changes from high to low when you have played more matches, meaning that your elo changes in the beginning are more volatile, and when you have more experience it is more stable. This also helps to get you to your accurate level sooner. Great content by the wat
Alonso is the GOAT, definitly, remeber the quote? "I knew he would hit the brakes because he has a wife and two kids at home." ~ 2005 Alonso about racing Schumacher xD
Great and interesting video, good job :) In my opinion, a big problem with this ELO system is that is doesn't count the DNFs correctly. If a driver is leading a race but the car fails and therefore he gets beaten by a teammate, it doesn't seem fair to lose points (yes, I'm talking about Kimi :D )
@ it’s fantastic, the mic in particular really makes it for me. What I love about UA-cam is that it can feel like a conversation and you’ve honestly nailed it. You remind me a bit of Eddie Burbach which is a great thing!
This is actually a lesson in ETL for data science. LOL. Formula 1 is geeks paradise indeed, from the car engineers to the fans, we all just geeks. Real talk, his details show he looked at the data, saw all the weird exceptions and made a plan for them. respect, this man looked at the data! I can guarantee you as of April 2024, there is no AI that would've build this code, this is pretty cool. The man understood the key question he was trying to answer, and all his decisions around ETL were driven by that question. Well scoped, well executed
Hope you get to see this comment - this is the first video of yours that the Algorithm gods have sent me, and it was fantastic. Clearly an immense amount of effort put in here, but the script and editing were great too; funny, informative - just all-around good. Easy subscribe. Nice sign-off too, hope to see this channel explode from 15K subs at the time of writing! Now time to watch the entire back catalogue....
That's why I am a huge advocate for demolishing the constructor championship and making the constructors collaborate on one car that all f1 drivers use to see who is the best driver...
I'm sure others have said this at this point, but, I think you should take qualifying into account. There could be reasons why a driver fell down into "p19" as you said, and that could be mechanical failure and driver error by... another driver. So driver A gets pole position, and Driver A takes off and gets to corner one, but Driver B decides to have a shufti up the inside, but goes to deep on the brakes and buries his nose-cone into Driver A's sidepods, then by no lack of skill on the part of Driver A, he suffers in the points of this ELO because he finished behind his teammate, who only qualified 5th, but managed to get past the carnage unscathed. So in short, qualifying, forced and unforced error, and mechanical failure should be taken into account. I realize this makes it a whole lot more work. Still, I have wanted to come up with some mathematical solution for years, so I salute your effort. I really, really do. EDIT: I have to add another bit to this, to even make it MOAR complicated. And this applies to both Hamilton and Schumacher. If a driver is leading a race, and because his teammate is leading in WC points, one driver is told to by the team to slow down and let the points leader pass, how is that accounted for? I could pile on but what about pitstop errors? This is why qualifying should figure into the equation.
really hope this catches on, I'd love to have MLB-level statistics become a regular part of F1. I want a the current championship standings shown next to ELO on a broadcast one day!
No one can deny Alonso is talent wise, at least top 5 in the sport´s history. But this system as thorough as it seems, goes out the window the moment you see Button´s 3rd
Alonso in general is one of the motor sports goat having achieved and participated in other disciplines ,that can't be said about other. Yeah the elo fails when button is 3.
This is the "Peak" ELO, and it is kind of pointless, I would bet the "peak" ELO of Button was in 2015 after he just beaten Alonso, which probably at the time had a very high ELO...
this system favours beating skilled drivers who are past their prime (exactly what alonso has done for the last decade). If you find a skilled driver, they have a high ELO, if they are past their prime you can beat them by a big distance and gain more points. No wonder alonso went to the top, it’s a unique situation for him.
To make an ELO engine for F1 you need, as he said, an equation to calculate everything. You can do that by making a car performance equation. Basically take the constructors championship after every season and add the difference in stats of speed, grip etc. And use that to give points depending on who beats who in what car.
The problem there is that it's nearly impossible to normalise that from car performance. I think it's very easy to argue that Verstappen and Alonso aren't far away from Hamilton, but because of his long dominant period during the last decade, if you don't normalise for car performance, they all look like complete trash by comparison. If F1 was a spec series, it would be easy as anything to make that work, but since it's not, that's just pretty much impossible to do
Calling this a "not a perfect" system is... understated. In F1 terms it is simply broken. The best model that i've come across, and mind you it still has it's flaws, is the one made by F1Metrics and it surely deserves a read. Anyway, this was a great watch and the amount of time and effort that went into it really shows. Before the youtube days i made something like this myself and for myself (it was way more broken) and i know and really appreciate what you went through!
Getting shafted by poor team strategy, finishing the race but your car was damaged by another driver, finishing but behind teammate due to reliability, botched pitstop, safety cars and red flags can ruin a drivers potential gain of ELO but has nothing to do with the skill of said driver. Ironically Fernando has suffered alot of these yet comes out on top, which means he should probably be every higher. And lastly, teamorders are a thing aswell.
They should somewhat cancel each other out over time, although it is certainly possible that one driver is particularly unlucky. Moreover, the car's reliability and driver's skill do go hand in hand to some extent, at least in when you take into account things like tyre management, overall management of the car's systems' temperatures, reflexes, awareness, driving style etc..Of course there are instances where the car breaks down through no driver's fault at all, and indeed an unfortunate yellow or red flag can ruin an otherwise perfectly good strategy. It's just really tough to factor it all in when, in some instances, a case can be made that a driver's compitence was of influence, especially as it is unclear just how big that influence was. As for team orders, I'd argue the better driver gets relatively few of those.
@@CallACAB93 Alonso vs Ocon at Alpine is a prime example. Ocon at the end of the year bragged he beat Alonso in points, yet it seems more like Alpine beat Alonso with the factors mentioned above and Ocon was just there.
Nah, cause there would have been times he would have also profited from misfortunes of other drivers. Probably easier and safer to keep these out of the algorithm, as it all probably cancels each other out anyway.
Yes, this rating system is about as nerdy as it gets. Some seasons there is a clearly superior car and in others they are pretty equal. I think that this should be taken into account, in seasons where there are four or five different race winners. Rubinho was the ultimate bad luck guy but remember Kimi retiring so many races when he was over a pit window in the lead.
I watched this just because as an age of empires player I always thought elo system to be fascinating and clever. You are very entertaining and fun to watch, thanks for this!
Just like many other similar videos: Keep fudging the numbers till Fangio is dethroned. I do like how everybody seems to start any number crunching system with him as the final answer at the first run through.
There is no way to dethrone someone that could only start F1 when already near 40 and then only be either WDC or 2nd in the standings. If he were like 30 we would never even need a discussion on this ever again. Verstappen is someone however who can end up matching Fangio toe for toe. Took some 73 years though.
Juan Manuel Fangio, el maestro, will always be the greatest driver of all time, no matter which analysis or statistics is done (objectively and accurately). People argue against his stats (like the highest winning and pole percentage), being biased just because of the lower amount of races; but they forget drivers were killed in accidents often in that time, there was a lot more of competition between cars and constructors, and there were a huge amount of retirements and dnfs, and despite that he's the only driver in history to win the world championship with 4 different engines and constructors: Alfa Romeo, Mercedes, Ferrari and Maserati. 5 total championships. I think this record will never be broken. As one of his fellow competitors said: "The fastest car of the season is the one that Fangio drives" There is a very good Netflix documentary about Fangio "A Life of Speed": I strongly recommend you to watch it!
You could also include RacePosition vs TeamFinalConstructorsPosition in that way you would be taking into account how the drivers performed with the machinery they where given. Max winning races with the best car of the season would give less points that for example Alonso winning races with the fourth car in the grid in 2010. (Just to clarify, I'm not looking to misprize Max)
As a stat geek, I absolutely love this video. I think the best way to tweek it "a bit" would be to not only let teammates run against each other, but to compare the pace of the cars in correlation to the other cars, use that result as a base value for the comparison of the expected outcome and let that influence the value of the gain. I admit, right now I have no clue how the code would look like, but I think that this would solve the piastri problem for example. It would also incoorporate the team intern rivalry indirectly, because for example a bad performing Logan Sargent would lower the expected pace of the Williams in comparison to Albons results. It would still not be flawless, but I think a step in the directionof more advanced stat, as it would also take the "greatness" of the car into consideration.
Brilliant, interesting content (I love the conclusion that all of this was just a 'eff around and see what happens, even it turns out just as proof that the method is flawed', because... sometimes you need that kind of experiment. Entertaining, if nothing else!) and absolutely fantastic comedic timing. I think I spent about 20 minutes just replaying the 'maximum number of teammates' section yesterday and crying from how much I was laughing. Have spent the day going through your repetoire of videos and looking forward to seeing more from you in the future!
Another example of the imperfection of this system, is when Hamilton and Alonso were tied at 109 points for 2nd and 3rd place in the 2009 F1 championship. Since Hamilton was a rookie that year (and thus had a low ELO), Alonso (who had a high ELO from winning the 2005 and 2006 championships) lost a lot of points. If it accounted for the actual performance of the drivers, and for the sheer incredible level of Hamilton as a rookie, it would make Alonso lose fewer points.
This was a great video with excellent information mixed with downright brilliant comedic beats. Thanks for the great deal of effort you put into these vids. You have passed the likes of Josh Revell as my favorite F1 youtuber.
I understand basing the ELO on results in-between teammates for a more level playing field, but keep in mind teammates are also not always equal. Either directly, through team orders, or more in-directly through priority among the team and manufacturer. There's several examples of a team's #1 driver getting upgrades sooner, or a #2 driver being handed a T-car or a B-spec car.
Yeah, been there. Building an ELO system for F1 simply doesn't work due to the lack of varied matchups and comparable data, and too many other factors flowing into the results that are outside of a driver's performance, like, depending on system: - strategy errors by teams and bad luck due to things like safety car timing - sabotage within teams (like Santander's blocking of Kimi's setup calls and car development input in 2014, or Red Bull's blocking of Gasly's setup calls in 2019, or what happened at Andrea Moda) - car characteristics massively disadvantaging one driver over another (see Ricciardo vs Norris at McLaren) and much more. An ELO system can't account for factors like these, and ultimately as a result it is pretty hard to build a driver battle based ranking system as a result. Due to these issues, I built a different ranking system earlier this year (with multiple iterations improving on the previous version's issues) that is instead based on driver performance relative to the car they drive. It looks at where a driver qualified relative to where the car should've been based on where it stacked up throughout the respective season, also looks at where it finished relative to where it should've, and calculates a score for each race accordingly. It also accounts for number of total DNFs in the race and calculates the likelihood by which the driver might've gained positions from those. It calculates this score across a driver's 5 year career span to identify a peak and be able to rank them and disregard things such as rookie seasons, or final years in the sport. After throwing version 2.1 of said system at over 50 drivers, the top 5 scores ended up being: 1. Juan Manuel Fangio (1953-1957): 1831.81 2. Max Verstappen (2018-2022): 1828.38 3. Fernando Alonso (2010-2014): 1779.93 4. Kimi Räikkönnen (2003-2007): 1725.57 5. Jackie Stewart (1969-1973): 1666.93 P6 to P10 are held by Alberto Ascari, Michael Schumacher (2 different timespans), John Surtees, and Jim Clark Version 2.1 wasn't perfect, and "Spreadsheet Hell 3.0" improved on it, but thus far has only been used in limited tests and remains unfinished. Unlike version 2, which compensated for DNFs by predicting a finishing position based on performance the rest of the year, version 3 ignores DNFs completely but is still capable of at least counting the quali score of said driver for said race, to see if the driver at least on one lap pace did the car justice or not. Once that system is built and finished and at least 50 drivers have been ranked in it, I'll likely end up discussing it on my channel or on the DRS Train Podcast.
@@FedorAntony Most likely. I can do an updated version of the V2 Max Verstappen spreadsheet once 2023 is over. Remind me in a few weeks when the season is over 😅
@@anirudhsankar5321 Well, the short version is: Max Verstappen overtakes Fangio and gets the top spot with a total score of 1868.13. Which means the top 5 are now: 1. Max Verstappen (2019-2023): 1868.13 best individual season score: 2023 (2045.77) 2. J. Manuel Fangio (1953-1957): 1831.81 best individual season score: 1955 (2028.21) 3. Fernando Alonso (2010-2014): 1779.93 best individual season score: 2012 (2027.16) 4. Kimi Räikkönnen (2003-2007): 1725.57 best individual season score: 2003 (1922.94) 5. Jackie Stewart (1969-1973): 1666.93 best individual season score: 1969 (2037.50) The highest scored season of any driver overall in this data model is Jochen Rindt's 1970 with a score of 2256.02
I read an article where some University did a similar study around 2010 plus or minus 5 years, possibly even more in depth and came to the same conclusion, Alonso is the GOAT.
I think the ELO is more accurate the later the driver races. It's because At the beginning the drivers weren't mixed up, so the ELO rating were to drastic, but now it's more smooth. For example if Alonso would beat Norris (if everyone would start with the same ELO) Norris would have less points, than Piastri, that beat Sargent.
8:49 What an absolute unit Juan Manuel Fangio was. 3 hours all by himself, 96 laps.
and over 89 seconds ahead, and he was older than Alonso at that race.
all that in a suicide-bathtub on bicycle wheels. he was built different. messi of f1 in my book.
@@xander1052 in the 50s they all were a little "older" - it was a sport like golf for elderly rich gentlemen
@@keisuketakahasi4584 fangio was no rich man
@@tomstervelo fangio wasnt but it has the same energy of someone with spare time just deciding to drive for fun
This is it. Alonso is the goat. No more debate. Oh wait..
Tell Tommy and Matt!
@@jeffstaples347 Tommy knows Matt doesn't
Ocon intensifies
I mean, imo, Prost and Alonso are the two GOATs, which is funny because those are both the ones Mr. V put, so maybe those systems do work!
TU TU TU TU…….
you should make this open source so people can use and tweak this to make it more and more accurate and interesting if they want to would be very cool. great video btw
Ngl I have no idea how to do that - I write VB at work but only for internal tools. What would be the best way?
I think you can create a public git repo and Share your code and other related resources there.
would love to see if someone could add more variables to this. I.e finishing position, gap to teammate and the cars performance relative to the rest of the field. I think that including some of these would make the system a lot more accurate in comparing drivers from different teams
I'd like to help you out to make it a public repo@@MrVsGarage, do you have the sheet downloadable somewhere?
@@hbrown689 I'd add finishing position, gap to teammate, car performance relative to both the first (or second, if their car was the fastest) and last cars. I would also weight their age and number of races against their teammates, since we know that for a rookie, especially, there's an adjustment period, and they won't receive the same treatment as the more experienced driver in the team, and we also know that, especially before the 1990s, age was a huge factor in performance, as they did not treat themselves as athletes the way they do nowadays. It would help give things like Schumacher's, Alonso's and Hamilton's rookie performances a bit more of nuance, as beating your teammate as a rookie is an unbelievable achievement, as well as Hamilton's and Alonso's current performance despite their age, compared to their much younger teammates.
And as much as it could skew things a bit, I'd remove DNFs. Whereas nowadays most of the DNFs are driver's faults, historically, and particularly for the drivers before the 2010s, mechanical failures were so much more frequent that they could have DNFs for many races in a row, by no fault of his own (see Kimi Raikkonen 2005, for example). I guess in the end, driver's mistakes would kind of cancel each other out, for the most part.
I would honestly be more than happy to be a part of a group willing to undertake this (it would be too much work for a single person).
In the end, it would be a better measurement, but it would still let out a few subjective things, unfortunately, like the differences between how teammates were treated in certain eras, like, for example, in the 1990s and 2000s, for example, teams generally had very clear "2nd drivers" who would get development parts later,
Brazil 2023 Alonso showed what he's made of. What an unbelievable drive from him to fend off Perez.
This guy should make the rating system to compare against all the ninteen other drivers in the grid, than only your teamate, when then we would get honest elo rating on the board. Fangio was back in the day atleast dominating his teamate so hard, that he got so high elo rating, while Alonso drove for so long, that he in time to got to that high elo rating.
@@jout738 Agreed, just because it would be cool what would show up. However, we don't get the full picture. For example, a driver has a mechanical failure or race ending crash.
Thats why Max is the best this sport has ever seen, Senna with “only” 3 WDC’s proves that this sport is not about statistics, but 18 year old Max did that 30 laps Spain 2015 against Kimi in a Ferrari, in a car Max drove for the first time.
We all know who the best this sport has ever seen is. Max is a cleaner and more consistent Senna
@@jout738ocon was the only driver to beat alonso in team mate battle in 2022, and it was because the car was pure trash.
@@yerrie1908Driver complained about Max for years. He has since matured into a great champion but be real. They called that boy Crashstappen for a reason😂
as a programmer, I know it really pisses me off when someone point out something i overlooked after I am finished. Sadly, I’m here to ask you a question. Did you compare the teammates in races where one of them had a mechanical failure? Because that could mean the car is taken into account…
A year later and no response, you killed him with this
well, we're wating...
There's so many things that definitely weren't taken into account, "equal machinery" only goes so far, most of the times the team will prioritize a certain driver or car and put better parts, or better set up the car, there's also strategy and pit stop priorities to be taken into account. Drivers with a long career and that are known to enforce the "number one driver" in their contracts (like Alonso) will likely get a better elo not necessarily being the better driver
@@AdamastorPequeno01 you re absolutely right, but it would get way too complicated for a simple elo system, i think with this one it already took an eternity
In the early '80s, I was about five years old, my father was driving. Suddenly, someone made a bad maneuver in front of us and my father honked in protest. When the cars were parallel, the driver of the other car made a gesture of apology and my father, who had been mumbling some insult, instantly became silent and then…laughed. The old man in the other car was Fangio.
I hope it's real
@@20ege008 Yes it's real, it become a story my dad told in every reunion for a lot of time.
so back in the 80s, fangio locked up, your dad moved up the inside, gave him the peace sign, and fangio nodded in respect
your dad is the goat of F1 confirmed, put him in a head to head with alonso and its gg
so back in the 80s, fangio locked up, your dad moved up the inside, gave him the peace sign, and fangio nodded in respect
your dad is the goat of F1 confirmed, put him in a head to head with alonso and its gg
Memory is reconstructive. There is like a 90% chance that what you're saying is not true, even if you are actually for real.
Great production quality, excellent editing, top-tier entertainment. Well done Mr. V!
Thank you 🫡
Retrofitting the Chess ELO system to F1 is the most British thing I’ve seen this month. Well done
It's not just for chess but any zero-sum game in general (i.e. you lose = opponent wins, no scenario that's good for both of you).
😂😂😂😂😂
Please elaborate how this is a particular british thing, as I do not really associate chess with being british. As it is one of the few thing the britsh did not steal from India
@@QuixoticCowboy oh are we gatekeeping? I’ll make it painfully clear then; it’s British because likely only Brits would make this creative correlation and apply the system to it as both chess and F1 are obscenely popular in the UK. It’s about sports, not history. I’m not even British so cool your jets. It was a joke clearly not meant for you.
I already struggle to remember the Caro-Kann, Italian, Sicilian, Jobava London, and Ruy Lopez openings. Now you're telling me I have to remember the Verstappen, Alonso, Pérez, Hamilton, Räikkönen, Schumi, Button, Sebi, and how many others...
I don't watch Formula-1 and clicked on this video just to find out how would you apply the Elo system into such a sport. The video was so amazingly entertaining so I instantly forgot I was going to turn it off after the 30 seconds. The edit and the way of telling the information are astonishing. I would subscribe to your channel just because you deserve it. I really hope you'll have enough enthusiasm to make every video as cool as this one just because this one is perfect and you are, honestly, the most underrated content creator I've seen at 2023. Just keep doing this art, mate, you are the GOAT
do you still not watch F1
This is an almost unhinged level of detail and it's a joy to behold!
There's been no hinges around here for a while!
My biggest hobbies are F1, Programming and Chess. This is the perfect video
Always knew Alonso was the GOAT sadly needed few more points to be 5wdc and better cars in his career to show how truly great he was/is. Alonso is one of those where stats don't show how good he is.
8 pts away from being a 5- time world champ, and if he made a bit better carreer choices maybe even more, basically if he had a bit more of luck hed probably have the most amount pf WDCs imo at least 7 or 8
@@nitr097sticking with ferrari until 17 and 18 would have DEFINITELY given him 25-ish more wins over the extra 4 years (the 16 car had no wins under vettel, so it isnt more) and probably would have gotten a WDC in 2018, it would have been close in 2017 but just maybe
Wrong. Kimi should have won 2005. 4 time at best
@@nitr097 IF the team changes were perfect, alonso could be a 15 time champ
And the team changer would be 2009 brawn which he was offered, 2010-2013 rebull which he also was offered, 2013 and 2014 2021 merc
Of course these changes are impossible to do and to know with period information BUT with future info like now you could do it
Bottled 2007 span in fuji and Canada bottled 2010 span in Belgium and overtook of track in silverstone his fault
Might need an update with Verstappen's 2024 season
Agreed
This video perfectly shows what it's like to be a Software developer with the quadrillion edge cases you never thought about before you started the project.
Let's make a "K" in equation the difference between finishing positions. If teammates are 1-2 the K is 1 so it's little difference, but if they are finishing 3 and 12 K=9 (mayby we can always multiple that number by 5 or 6 or something like that to get higher difference in points)
I also thought about this. The problem with it is that this would depend on the performance gap between cars, so maybe it's not the best system.
@@insertcoolname96car performance is taken into account because you’re only comparing teammates here, if your teammate ended 2 and you ended 10 is because theres a clear skill gap, so I think this sistem is quite good
@@dtar380 Yes, but imagine your teammate ends up 1st and you second, 30 seconds behind him. Not because you are good, but because your car is so much better than the rest. I can see your point, but I think it could depend on the performance of your car.
@@insertcoolname96 Making it proportional to finishing time difference could potentially work better then?
I just posted a similar response before seeing yours. A weighted position doesn't say as much as time differential. As an example gap/elapsed time. If your team mate DNF's then the scale is 1.
The problem with this whole approach is that is not the ELO of f1 as a whole, but rather who won more over their teammates. The best way to try to represent this on an ELO would be like iracing rating. You enter with a fix amount of ELO, and depending on the finish position and everybody’s else’s elo will determine how much you gain or loose.
For ex. A low elo driver can win a lot of points by finishing 7th but a high ELO driver can loose even by finishing 6th because rivals have low elo
Yeah I agree, this system is horrible, it shouldn't just be comparison with teammates (although it matters well). But even if we just take teammate comparison, the weighting is garbage.. Verstappen is going the greatest season of all time, beating Perez 18-2, and those 2 wins of perez take off about as much points off Verstapen as 6 Verstappen wins. For context, in both cases, Verstappen lost to Perez because of problems in qualifying.
No surprise that Alonso is getting rated well when the whole system is who has the weakest team mate, and his current team mate is Stroll.
Yes, my thoughts also.
The system can be updated by taking into account the relative power level of each car and see how well the drivers compare to the power level of their car.
If Ferrari has the best car by a mile, than getting into the podium should give fewer ELO points.
Finishing in the points with a tractor, is some thing that should boost your ELO points.
Also you need to take into account if mechanical failure/crashes was self-inflected or not.
The formula would be a lot more complicated, yet, it seems like a nice subject for a thesis on data analysis.
This episode is so clever and such a unique cross topic narrative. You’re tweaking the YT algorithm to potentially capture nerds in F1, racing, chess, data science, and coding. Kudos!
2012 says why alonso is the goat, close to win a wdc with only 3 wins and a car that was consistently the 3rd fastest, the driver who got the only 2 poles of ferrari that year in wet conditions
What about an ELO system for the teams, which gets joined to the driver's ELO to calculate the gains/losses, but that remain separate so drivers can change teams and carry only their personal points. That way you'd be able to somewhat gauge the effect of having a superior car.
Going even further maybe it would be interesting to reset team points whenever there's a big regulation change like in 2014, 2022, but use only the teams ELO at the end of the season to calculate the driver ELOs
I would love to see this as a public thing that updates every race/month/season. It will be so interesting to see the changes in elo after each race/season
Outstanding work!
Thank you so much!!
@@MrVsGarage You are welcome! Amazing channel. Hope you consider doing a member's group soon!
Absolutely brilliant work, congrats from Spain!
Creates a formula that puts Jenson Button above Schumacher... Dares to publish it.
this might be a horrible take but mabye shumacher just wasent allat, dont get me worng he was a beast but lets not forget two (almost three) of his titles where gained by crashing into his championship rival
Schumacher never had a high rated teammate under this system. Which is one flaw in this. Since it only looks at your teammate, beating a much higher rated rival in a different car (even if that car is equally good, or even better than yours) doesn't mean anything.
Also Brambilla, Fittipaldi and Guerney
@@kukuc96this system also favours beating skilled drivers who are past their prime (exactly what alonso has done for the last decade). If you find a skilled driver, they have a high ELO, if they are past their prime you can beat them by a big distance and gain more points. No wonder alonso went to the top, it’s a unique situation for him.
As an engineer who doesnt code (or have the time to do the hard work you put into this), what I would love to see is an ELO weighted by how much (in seconds) a driver beat his team mate in each race - throwing out races with DNF's, where they make still have an ELO adjustment, but with a factor of 1.0 vs the time differential. Obviously it would have to be adjusted for the length of the race, so gap/elapsed time or something of the sort. This would give emphasis to the driver's abilities, even if they're in the fastest car. Such as Hamilton vs Bottas and Hamilton vs Rosbergwith with Merc, or Verstappen vs Perez with RB.
MV is not surprising to anyone who doesn't worship Hamilton. He really is a phenom. Despite some boring racing, its wonderful watching him develop and truly master his craft. Laser precision and as repeatable as a robot. This will probably get me downvotes. But I'm not surprised that Hamilton wasnt anywhere close. All but Mclaren, his WDC's were in by fsr the fastest car, for most of those years every team in the top 10 was running a Merc powerplant. He excelled at being consistent, but had a fast and stable car to do it. The stand outs on talent have always been those who coupd consistently manage a pointy car, MV, Schumi, Prost, & Senna. Anyone on the grid can be fast on the fastest *stable* car. This also explains the massive gap MV has had to team mates. Alex Albon does a great job explaining it in his interview.
Fantastic work and a fun watch. I'll sub.
FERNANDO ALONSO THE GOAT
No, Senna is
but at the end of th day it will be verstappen@@Wild_stream064
@@Wild_stream064 did you see the ELO system?
@@razorr_o yes
@@Wild_stream064 Alonso is the goat! Period!
The GOAT has figured out who's the GOAT. This is all I ever needed
I waited for this, thank you so much . Phenomenal work 👏
Thank you 🫡
It's a great start for a rank, there're loads more information needed to develop it, time interval between teammates (P1 and P2 is one thing P1 +20 seconds lead to P2 is another) increase elo for time diference; dnfs for driver error (how would one do it i don't know, so let's just ignore it); as you said qualifying; amount of times undefeated should have an increment to your score
This is a amazing work!!
the issue with time interval is that in the 50s 20 seconds would be actually pretty good, so you'd have to adapt to every era somehow, factor in when cars are broken but not enough to DNF (faulty brakes, no power steering, gearbox...), how many laps the car has been broken, maybe add team strategies to win points for a championship, also team strategies blunders (hello ferrari)... there is so many variables at play that it's virtually impossible to solve.
@9t8268 That does add more impact from the cars and the relative balance of power of the teams though. It makes it harder for teammates who have far and away the best (or worst) cars to gain or lose points compared to drivers in midfield teams where there is a much greater chance of having several positions between the teammates. Without getting into car performance, I am not sure there is a whole lot you could do to improve this.
@@supernewbie694 Maybe not time interval, but position differences (pro-rated based on number of racers).
"Mr V's tests a new ranking system. He makes 2 teammates fight each other, 99000 teammates, 35 exact same persons, -1 drivers against each other, drivers in multiple different teams, drivers with the same position.
A customer asks how many wins does a driver has, the excel promptly catches on fire."
For those who don't understand that joke, it's based on a stupid joke many devs love to tell : "a programmer enter a bar, and buys 1 drink, 50000000 drinks, -1 drink, apple drink, Null drink, ... A customer enters the bar ans ask for the bathroom, everything catches on fire."
@xsham6224 I love QA testing
I’d be interested to see the list with the 50s and 60s (and perhaps the 70s) removed. It’s was pretty much a different sport back then. Love the hard work on this!
I'm with you, but I'd take it in a different direction.
Not necessarily "removed" but maybe split into 2 or 3 big eras? Something like the very early years, then a cut around the late 60s when aero started to become a thing and maybe another cut around the 90s/late 90s where safety became an even bigger concern, simultaneously giving us a pretty good "modern" era.
Other people would probably find better split-points and you'd have the issue of how to deal with drivers that participated across multiple of these splits, but it would slightly alleviate the issue of comparing drivers between vastly different phases of the sport.
@@IguitarVreakIGreatest driver of the decade would be cool
@@IguitarVreakIi think breaking it up by tech eras makes a huge amount of sense since the cars are so different between each one and require different driving styles. Early, aero and modern would be a good split.
@@kingofmemes5017like you said, each decade of car needs a different driving style. Adaptability is what makes a driver great. It may just be the biggest factor anyways.
That means splitting era cars really only tells you who drives those style of cars best. Not the most accurate in finding out who is the actual best driver of all time.
Just raising a point against yours. Honestly, I myself have no solution to this problem. I don't think there is any good solution.
13:43, so currently Max has overtaken Button's Elo and is about to get fangio then? (since that only ran to belgium)
6 wins x 6 points would make it 1350 indeed but that's without taking into account the other factors of finishing lower than some other drivers in Singapore for example.
@@frankhuurman3955no because it only takes into account the score against teammate. And he won from Perez....
then 1350 it is :)@@toaojjc
It will require at least 10 consecutive victories before Verstappen gets to Fangio. And since Fangio could only start driving F1 when he was nearly 40...if that guy was 10 years younger we would have never needed a discussion on who'd hold the most WDC...that would have been Fangio till the end of time.
Verstappen might however be the first in 73 years to match Fangio...the rest doesn't even come close.
@@JayQ2kAs Verstappen rises he slowly gets less points off Perez as well since the gap between their ELO goes up so it'll take a fair bit longer than 10 I think.
Very interesting work, well done. Obviously looking at teammate head-2-head's is limited to just finishing positions, and subject to who your teammate is (who's to say how the ratings of Fernando and Max would be impacted if Fernando joined RBR for a bit as an example). Also, those 'all time great' performances, or margin of dominance over your teammate isn't taken into account, nor are external factors. Still very fun exercise!
The ELO system was made with competitive pairings in mind. Verstappen gains a lot of points on Perez because Perez is not necessarily there to win the race, but to help Max win the race - they're teammates after all, they're not always competing against each other. Piastri as you pointed out is close to last because in many races McLaren call was to give Norris the lead even in races where Piastri was beating his teammate.
I know this may sound a little nitpicky 'cause how else would you be able to normalize the competition across so many factors in another way, but I guess that's exactly my point - for ELO to work on F1 it would have to be a completely differente weighted average system
Hold on hold on, giving people ratings based on how they are compared to their teammate doesn’t quite work because what if they have a shitty teammate who’s only driving cause their dad owns the team
had to scroll way too long for this comment. all other smaller problems are obsolete when the basis of the formuma is that shaky… no offense to the guy tho obviously
Yeah if there ever were a situation like that I’m sure this theoretical team might skew the data
Why wouldn't it work? Performance speaks for itself
This would be fine if most top drivers on the ELO ranking have matched up against a large variety of teams and teammates even if one of them was a shitty nepo driver. But if let's say Max was paired up with Lance Stroll from 2016 to 2024; indeed, that would make the ELO ranking unrealistic, but three years of Checo ain't too far.
Alonso did achieve the highest in 2014 well before Vitchtits had the chance to disgrace Formula 1
Love this videos so much. The data, the coding, the WTFacts about F1. Simple perfection 1412
Nothing we didn't knew.
We still speaking about Fernando even with the 5th car on the grid. 😂
The most surprising thing about this is Roberto Moreno having the 15th best ELO of all time! Aidan Millward would be proud...
yeah but how about VITTORIO BRAMBILLA ahead of Schumi?
Moreno was highly competent but also pretty much trounced by Piquet in the final years of Nelson's career...
I guess the issue here is that it rates later wins against incompetent team mates too highly?
@@fernandodasilva-producoesa1023must get a non cheating bonus? :)
@@intheclarts You actually want the incompetent teammate when they are a rookie in this system. Let's say there is a driver has a final real rating of 800, but they still start with 1000 as rookies, so you can take more points off them when they are rookies.
The other option is beating a high rated teammate when they are not that good anymore, but they still have a high rating from their past success. Although they are not in the top, I imagine Leclerc gained a good chunk of points from Vettel that way, and probably Nico Rosberg also got a bunch of points from Schumi.
@@kukuc96 Don't know, Piquet had plenty incompetent team mates, Brabham were famous for it. Not sure where he rates on the system, but that he would be below Moreno? Not sure how this computes. Beat every teammate up to Mansell, that was fairly level, then beat every teammate until Schumacher.
What I learned from this is that spec series race categories are ripe for ELO comparisons. With a spec series you can absolutely compare the entire field against each other for each race.
Besides that one F3 season where a certain driver seemed to have an upgrade on their suspension compared to everyone else.
this should include more than just win lose or draw, since P1 and P2 is a lot less of a difference than P1 and P19.
This is amazingly interesting and so entertaining. I was smiling throughout the whole video!
Hey I'm so glad! I'm trying my best to make things interesting and entertaining, both as much as possible!
What a video! I've been gagging for it since you teased it in your last one. Always knew Fernando was the goat 🐐
Nice video! Regarding the point about factoring in finishing positions to make the model more accurate, that's basically what F1metrics did. They also factored in other things such as age/experience and accounting for mechanical DNF's and Customer/Works cars. You should check it out (it hasn't been updated since 2019 unfortunately)
i love this as being more of a coding edge case demonstration rather than being fully focused on the F1 aspect of it
i remember back in uni most of my projects were like "oh yeah, this works, im a genius!", then getting the professor's input file and seeing 90% of my code exploding because i didnt take into account barely any of the edge cases
"nah, there wont be empty groups"
"nah there wont be repeated inputs"
"nah there wont be unnacceptable characters"
oh boy, did i learn that year
Not surprised to see Button so high up, especially if we ignore qualifying and only look at the race results. He has been teammates with the two greatest drivers of his era and did extremely well against both of them.
We don't need an ELO system to know that you are the goat of F1 youtube
I would gladly accept this.
I respected and Feared Alonso because I Supported Michael Schumacher , Kimi Raikonnen & Vettel , I was against him. However, I can't deny his genius and the dude suffered so much coz of Machinery but still Races like a Beast. He's one of the 3 sportsmen that turned me into a fan even though i was rooting Against them in the begining Fernando Alonso, Roger Federer & Lionel Messi are those 3. Alonso is Great in F1 and his Other Racing achievements as well, He's one of the Greatest Racers Overall. There's a Study Published in Britain 2016, Showing he's 3rd Greatest Driver , #1 of Drivers still driving in 2016 and Your ELO Which shows Alonso is The GOAT. WORTHY IMO.
I'd definitely say he's the greatest Racer who Debuted in 21st century.
Not a chance. Hamilton, Vettel, Verstappen, Russell all significantly better than Alonso.
Alonso just always had very poor teammates in a car that suited his crappy driving style which massively overinflated his stats. He didn’t have a teammate the caliber of Leclerc, Russell, Rosberg, etc and he lost his first season against Button and only
Won the next season because Button was in retirement mode.
Alonso has lost to Tarso Marquez, Trulli, rookie Hamilton, Button, Ocon.
"Alonso has lost to rookie Hamilton" you should revise what happened that year
@@paperplane-db8qf if you understood anything of the code, you'd know that what you're pointing out is already accounted for. Losing for a low elo rookie is bad for his stats, yet he managed to keep it.
@@luissun2103That was overturned by his performances against a past prime Raikkonen and Massa.
this system favours beating skilled drivers who are past their prime (exactly what alonso has done for the last decade). If you find a skilled driver, they have a high ELO, if they are past their prime you can beat them by a big distance and gain more points. No wonder alonso went to the top, it’s a unique situation for him.
This guy is built different
in OTB chess, you start with no elo, after X matches, you get set an elo based on your results (for example you score 50% against an average opponent of 1423, your starting elo will now be 1423, you do not need to build up from 1000) this should solve the 'Piastri' problem in your video. To be more complete also the R factor changes from high to low when you have played more matches, meaning that your elo changes in the beginning are more volatile, and when you have more experience it is more stable. This also helps to get you to your accurate level sooner. Great content by the wat
Alonso is the GOAT, definitly, remeber the quote? "I knew he would hit the brakes because he has a wife and two kids at home." ~ 2005 Alonso about racing Schumacher xD
I just want to say, I really like the way you present
Thank you!
Great and interesting video, good job :) In my opinion, a big problem with this ELO system is that is doesn't count the DNFs correctly. If a driver is leading a race but the car fails and therefore he gets beaten by a teammate, it doesn't seem fair to lose points (yes, I'm talking about Kimi :D )
2 races from the end of 2024, I reckon Verstappen probably made that goal of beating Checo enough...
Just wanted to let you know that I love the editing and presenting style of your videos. One of my favourites.
thank you! it takes a little while but I wanted to make something a little different from all the other F1 channels!
@ it’s fantastic, the mic in particular really makes it for me. What I love about UA-cam is that it can feel like a conversation and you’ve honestly nailed it. You remind me a bit of Eddie Burbach which is a great thing!
This is actually a lesson in ETL for data science. LOL. Formula 1 is geeks paradise indeed, from the car engineers to the fans, we all just geeks. Real talk, his details show he looked at the data, saw all the weird exceptions and made a plan for them. respect, this man looked at the data! I can guarantee you as of April 2024, there is no AI that would've build this code, this is pretty cool. The man understood the key question he was trying to answer, and all his decisions around ETL were driven by that question. Well scoped, well executed
Enjoy Alonso being the goat because it’s looking like max might catch him soon
I think the fact that Jenson Button comes out third, proves that you saying "it's not a perfect system", maybe the understatement of the century
Button was a great F1 driver, always underrated, as your ignorant comment shows.
Hope you get to see this comment - this is the first video of yours that the Algorithm gods have sent me, and it was fantastic. Clearly an immense amount of effort put in here, but the script and editing were great too; funny, informative - just all-around good. Easy subscribe. Nice sign-off too, hope to see this channel explode from 15K subs at the time of writing! Now time to watch the entire back catalogue....
I did see it!! Thank you, I do enjoy writing scripts so I try to make them as good as I can, glad it payed off!
The Lance Stroll 1200 rating had me burst out laughing 😂😂😂😂😂😂
“Which driver is most consistent at demolishing their teammate” *Laughs in estaban ocon
That's why I am a huge advocate for demolishing the constructor championship and making the constructors collaborate on one car that all f1 drivers use to see who is the best driver...
Spoiler alert: it's not Pastor Maldonado!
STILL BEAT THE ENTIRE GRID IN A FUCKING WILLIAMS!
I wonder how high Max is rated now…
I'm sure others have said this at this point, but, I think you should take qualifying into account. There could be reasons why a driver fell down into "p19" as you said, and that could be mechanical failure and driver error by... another driver. So driver A gets pole position, and Driver A takes off and gets to corner one, but Driver B decides to have a shufti up the inside, but goes to deep on the brakes and buries his nose-cone into Driver A's sidepods, then by no lack of skill on the part of Driver A, he suffers in the points of this ELO because he finished behind his teammate, who only qualified 5th, but managed to get past the carnage unscathed. So in short, qualifying, forced and unforced error, and mechanical failure should be taken into account. I realize this makes it a whole lot more work. Still, I have wanted to come up with some mathematical solution for years, so I salute your effort. I really, really do. EDIT: I have to add another bit to this, to even make it MOAR complicated. And this applies to both Hamilton and Schumacher. If a driver is leading a race, and because his teammate is leading in WC points, one driver is told to by the team to slow down and let the points leader pass, how is that accounted for? I could pile on but what about pitstop errors? This is why qualifying should figure into the equation.
This is my first video recommended to me by you, and I thoroughly enjoyed it. Subscribed. Loved every minute.
really hope this catches on, I'd love to have MLB-level statistics become a regular part of F1. I want a the current championship standings shown next to ELO on a broadcast one day!
No one can deny Alonso is talent wise, at least top 5 in the sport´s history. But this system as thorough as it seems, goes out the window the moment you see Button´s 3rd
Alonso in general is one of the motor sports goat having achieved and participated in other disciplines ,that can't be said about other. Yeah the elo fails when button is 3.
This is the "Peak" ELO, and it is kind of pointless, I would bet the "peak" ELO of Button was in 2015 after he just beaten Alonso, which probably at the time had a very high ELO...
this system favours beating skilled drivers who are past their prime (exactly what alonso has done for the last decade). If you find a skilled driver, they have a high ELO, if they are past their prime you can beat them by a big distance and gain more points. No wonder alonso went to the top, it’s a unique situation for him.
Will we get an update about this one? I wonder if Verstappen will be the next GOAT.
With how Perez did this year, I doubt it! Checo's ELO could be negative for all I know!
1980: people will do fantastic stuff with technology in 40 years
2023: finally i will know who is the best driver in F1
To make an ELO engine for F1 you need, as he said, an equation to calculate everything.
You can do that by making a car performance equation.
Basically take the constructors championship after every season and add the difference in stats of speed, grip etc. And use that to give points depending on who beats who in what car.
In theory, BIG THEORY. Could you do a sheet showing how one driver did against EVERYOTHER driver in every race of their career?
Yeah, but that factors cars into it HEAVILY. So you scew towards drivers who had long stints in dominant cars like Hamilton, Schumacher or Lauda.
The problem there is that it's nearly impossible to normalise that from car performance. I think it's very easy to argue that Verstappen and Alonso aren't far away from Hamilton, but because of his long dominant period during the last decade, if you don't normalise for car performance, they all look like complete trash by comparison. If F1 was a spec series, it would be easy as anything to make that work, but since it's not, that's just pretty much impossible to do
@@QemeH or drivers that had EXTREMELY fast teammates their whole time racing like prost
This wouldn’t work.
The whole point is that it doesnt do that. Its based on equal machinery so the differences in cars do not have any effect.
Calling this a "not a perfect" system is... understated. In F1 terms it is simply broken. The best model that i've come across, and mind you it still has it's flaws, is the one made by F1Metrics and it surely deserves a read.
Anyway, this was a great watch and the amount of time and effort that went into it really shows. Before the youtube days i made something like this myself and for myself (it was way more broken) and i know and really appreciate what you went through!
Which one by F1metrics?
Getting shafted by poor team strategy, finishing the race but your car was damaged by another driver, finishing but behind teammate due to reliability, botched pitstop, safety cars and red flags can ruin a drivers potential gain of ELO but has nothing to do with the skill of said driver. Ironically Fernando has suffered alot of these yet comes out on top, which means he should probably be every higher. And lastly, teamorders are a thing aswell.
They should somewhat cancel each other out over time, although it is certainly possible that one driver is particularly unlucky. Moreover, the car's reliability and driver's skill do go hand in hand to some extent, at least in when you take into account things like tyre management, overall management of the car's systems' temperatures, reflexes, awareness, driving style etc..Of course there are instances where the car breaks down through no driver's fault at all, and indeed an unfortunate yellow or red flag can ruin an otherwise perfectly good strategy. It's just really tough to factor it all in when, in some instances, a case can be made that a driver's compitence was of influence, especially as it is unclear just how big that influence was. As for team orders, I'd argue the better driver gets relatively few of those.
F1 RNG, just part of the game
@@CallACAB93 Alonso vs Ocon at Alpine is a prime example. Ocon at the end of the year bragged he beat Alonso in points, yet it seems more like Alpine beat Alonso with the factors mentioned above and Ocon was just there.
Nah, cause there would have been times he would have also profited from misfortunes of other drivers. Probably easier and safer to keep these out of the algorithm, as it all probably cancels each other out anyway.
Yes, this rating system is about as nerdy as it gets. Some seasons there is a clearly superior car and in others they are pretty equal. I think that this should be taken into account, in seasons where there are four or five different race winners. Rubinho was the ultimate bad luck guy but remember Kimi retiring so many races when he was over a pit window in the lead.
This video is so good!! Honestly, I hope the algorithm gods help you get the popularity you deserve, as your videos are pretty good!!
Thank you!! It's been going well so far so I can't complain!
I watched this just because as an age of empires player I always thought elo system to be fascinating and clever. You are very entertaining and fun to watch, thanks for this!
All these exceptions make for a great bit of F1 trivia 😂
Just like many other similar videos: Keep fudging the numbers till Fangio is dethroned. I do like how everybody seems to start any number crunching system with him as the final answer at the first run through.
He really had a crazy career even if it was 8 short seasons. Some of the stories of him carrying on at tracks where everyone else passed out it wild
There is no way to dethrone someone that could only start F1 when already near 40 and then only be either WDC or 2nd in the standings. If he were like 30 we would never even need a discussion on this ever again.
Verstappen is someone however who can end up matching Fangio toe for toe. Took some 73 years though.
Juan Manuel Fangio, el maestro, will always be the greatest driver of all time, no matter which analysis or statistics is done (objectively and accurately).
People argue against his stats (like the highest winning and pole percentage), being biased just because of the lower amount of races; but they forget drivers were killed in accidents often in that time, there was a lot more of competition between cars and constructors, and there were a huge amount of retirements and dnfs, and despite that he's the only driver in history to win the world championship with 4 different engines and constructors: Alfa Romeo, Mercedes, Ferrari and Maserati. 5 total championships.
I think this record will never be broken.
As one of his fellow competitors said: "The fastest car of the season is the one that Fangio drives"
There is a very good Netflix documentary about Fangio "A Life of Speed": I strongly recommend you to watch it!
You could also include RacePosition vs TeamFinalConstructorsPosition in that way you would be taking into account how the drivers performed with the machinery they where given.
Max winning races with the best car of the season would give less points that for example Alonso winning races with the fourth car in the grid in 2010.
(Just to clarify, I'm not looking to misprize Max)
As a stat geek, I absolutely love this video. I think the best way to tweek it "a bit" would be to not only let teammates run against each other, but to compare the pace of the cars in correlation to the other cars, use that result as a base value for the comparison of the expected outcome and let that influence the value of the gain. I admit, right now I have no clue how the code would look like, but I think that this would solve the piastri problem for example. It would also incoorporate the team intern rivalry indirectly, because for example a bad performing Logan Sargent would lower the expected pace of the Williams in comparison to Albons results. It would still not be flawless, but I think a step in the directionof more advanced stat, as it would also take the "greatness" of the car into consideration.
Please continue the over the top content!!!
Came into the video knowing Alonso is the goat and ended the video being certain Alonso is the goat.
Can you share us the list? Would be nice if you had it in the description.
this man is like the embodiment of 50% madness + 50% genius. so thankful found this channel
Brilliant, interesting content (I love the conclusion that all of this was just a 'eff around and see what happens, even it turns out just as proof that the method is flawed', because... sometimes you need that kind of experiment. Entertaining, if nothing else!) and absolutely fantastic comedic timing. I think I spent about 20 minutes just replaying the 'maximum number of teammates' section yesterday and crying from how much I was laughing. Have spent the day going through your repetoire of videos and looking forward to seeing more from you in the future!
I respect your kind of stoic self torture 😂
For some reason you guys seem to enjoy when I suffer so I guess I can't stop 🤷♂️😅
2:20 this stupid system is why I quit my computer science class in school
Would be interesting to see every driver's iRating (overall result, not just vs teammate)
It probably wouldn't change it all too much at least at the top end
@@minigungaming5915 It would punish drivers like Alonso Raikonen and Button just for not having a car capable of winning races
Another example of the imperfection of this system, is when Hamilton and Alonso were tied at 109 points for 2nd and 3rd place in the 2009 F1 championship. Since Hamilton was a rookie that year (and thus had a low ELO), Alonso (who had a high ELO from winning the 2005 and 2006 championships) lost a lot of points. If it accounted for the actual performance of the drivers, and for the sheer incredible level of Hamilton as a rookie, it would make Alonso lose fewer points.
This was a great video with excellent information mixed with downright brilliant comedic beats. Thanks for the great deal of effort you put into these vids. You have passed the likes of Josh Revell as my favorite F1 youtuber.
gonna message Josh and tell him the bad news! But for real though thank you, these videos do take a while to make so I'm glad people enjoy them!
What would be Verstappens rating now ( after Las Vegas 2024)?
I understand basing the ELO on results in-between teammates for a more level playing field, but keep in mind teammates are also not always equal. Either directly, through team orders, or more in-directly through priority among the team and manufacturer. There's several examples of a team's #1 driver getting upgrades sooner, or a #2 driver being handed a T-car or a B-spec car.
Yeah, been there. Building an ELO system for F1 simply doesn't work due to the lack of varied matchups and comparable data, and too many other factors flowing into the results that are outside of a driver's performance, like, depending on system:
- strategy errors by teams and bad luck due to things like safety car timing
- sabotage within teams (like Santander's blocking of Kimi's setup calls and car development input in 2014, or Red Bull's blocking of Gasly's setup calls in 2019, or what happened at Andrea Moda)
- car characteristics massively disadvantaging one driver over another (see Ricciardo vs Norris at McLaren)
and much more. An ELO system can't account for factors like these, and ultimately as a result it is pretty hard to build a driver battle based ranking system as a result.
Due to these issues, I built a different ranking system earlier this year (with multiple iterations improving on the previous version's issues) that is instead based on driver performance relative to the car they drive. It looks at where a driver qualified relative to where the car should've been based on where it stacked up throughout the respective season, also looks at where it finished relative to where it should've, and calculates a score for each race accordingly. It also accounts for number of total DNFs in the race and calculates the likelihood by which the driver might've gained positions from those. It calculates this score across a driver's 5 year career span to identify a peak and be able to rank them and disregard things such as rookie seasons, or final years in the sport.
After throwing version 2.1 of said system at over 50 drivers, the top 5 scores ended up being:
1. Juan Manuel Fangio (1953-1957): 1831.81
2. Max Verstappen (2018-2022): 1828.38
3. Fernando Alonso (2010-2014): 1779.93
4. Kimi Räikkönnen (2003-2007): 1725.57
5. Jackie Stewart (1969-1973): 1666.93
P6 to P10 are held by Alberto Ascari, Michael Schumacher (2 different timespans), John Surtees, and Jim Clark
Version 2.1 wasn't perfect, and "Spreadsheet Hell 3.0" improved on it, but thus far has only been used in limited tests and remains unfinished. Unlike version 2, which compensated for DNFs by predicting a finishing position based on performance the rest of the year, version 3 ignores DNFs completely but is still capable of at least counting the quali score of said driver for said race, to see if the driver at least on one lap pace did the car justice or not.
Once that system is built and finished and at least 50 drivers have been ranked in it, I'll likely end up discussing it on my channel or on the DRS Train Podcast.
that's also pretty interesting! can't wait to watch your vid about it once it's done
Nice! I’m afraid to imagine what would happen if you’d count in Verstappens 2023 season, once it’s done. Better than Fangio? 😱
@@FedorAntony Most likely. I can do an updated version of the V2 Max Verstappen spreadsheet once 2023 is over. Remind me in a few weeks when the season is over 😅
@@PsychedeliKompotHi, the season is over. Im very excited to see that spreadsheet!
@@anirudhsankar5321 Well, the short version is:
Max Verstappen overtakes Fangio and gets the top spot with a total score of 1868.13.
Which means the top 5 are now:
1. Max Verstappen (2019-2023): 1868.13
best individual season score: 2023 (2045.77)
2. J. Manuel Fangio (1953-1957): 1831.81
best individual season score: 1955 (2028.21)
3. Fernando Alonso (2010-2014): 1779.93
best individual season score: 2012 (2027.16)
4. Kimi Räikkönnen (2003-2007): 1725.57
best individual season score: 2003 (1922.94)
5. Jackie Stewart (1969-1973): 1666.93
best individual season score: 1969 (2037.50)
The highest scored season of any driver overall in this data model is Jochen Rindt's 1970 with a score of 2256.02
What a time to be watching this. Max must be the ELO speedwagon by now.
I read an article where some University did a similar study around 2010 plus or minus 5 years, possibly even more in depth and came to the same conclusion, Alonso is the GOAT.
I wanna say Nate Silver has done this
Brilliant video. Random question just out of curiosity - is it possible to do this for only qualifying?
Leclerc fan?
I'M THE GOAT
I was eagerly waiting for this, and I was pleasantly surprised, or maybe not actually surprised, to see Alonso at the top.
This is probably my favourite UA-cam video I’ve ever seen. Data, F1 and it’s funny. Perfect 👌🏼
I think the ELO is more accurate the later the driver races. It's because At the beginning the drivers weren't mixed up, so the ELO rating were to drastic, but now it's more smooth.
For example if Alonso would beat Norris (if everyone would start with the same ELO) Norris would have less points, than Piastri, that beat Sargent.