Due to popular demand, here's a Google drive link to my thesis: drive.google.com/file/d/1CkYW0JdMfXXcczFvqjXcfp64D8eMIq1d/view?usp=sharing Huge thank you to the 1,000 people who subscribed to Coachless in its early stage. We changed our default subscription period from 90 to 30 days now, making the Item Analysis feature more accessible. Free version is now also implemented: coachless.gg/items
Just gotta say watching this video and I remember seeing that specifically sorc boots, into mejais soulstealer, into rabadon on vladimir has a 76% winrate. This was the first time I realized this since well 76% winrate is absolutely bonkers and mejais is only ever useful especially as a first item when you are absolutely dumpstering your opponent. Plus if you are playing a late game champ like vlad and are winning that hard in the early game it makes sense there is almost 0 way you can lose besides you having a stroke or all your teammates having one.
I will be waiting for you GG I hope you can find a good Senna otp to learn from as well I'd love to see how one of my favorite champions is itemized i have been too scared to comeback to league since they made the extra nerfs to her damage.
xPetu, you inspired me to dive into the world of math and discover it's wonders. I don't plan on majoring in it but the esoteric and fundamental knowledge I have gained has been partly because of you. When I become a crackpot engineer/scientist, I'll remember you.
This is the kind of analysis I wish we saw in every game dev team. I know your content is primarily LoL focused, but it would be really cool to see if this can be adapted to other MOBAs and become a sort of standard for analyzing MOBAs as a whole. One thing I think would be cool (albeit difficult to implement) is a ML model that takes into account not just the physical vs magical damage, but the specific hero matchups and maybe even team comps as a whole. I imagine that feeding enough of that data into a model would produce some interesting predictions.
The thing I learned from this video is that anytime I'm about to win for certain with shen, I need to recall and buy Rylais so that it has a 100% winrate and thus people start buying it and throwing their games
Hey Petu, I was an assistant coach for a MOBA esports team back in 2021 mostly dealing with data analysis. Although I did not come from a statistics background, I enjoy watching this video. Thanks and good luck on the project!
This is made even more clear in data analytics of a game like TFT where the Delta shows the placement change for a specific item on a specific unit. There you can clearly see exactly how much on average your item decision on each unit will affect the result of your placement.
@@ttv_zombiefetus1710 Typical... I don't even know what you are, to be fair. Do you disagree with what I said? Do you disagree with the use of the word "animal" only? What's your point?
you explain complex subjects in a way that is easy to grasp for someone that has never studied statistics and probability, like a professor! We love you Petu good luck on your and your teams project!
Huh?....The thesis here covers over more advanced statistics topics like Markov chains. It's not a purely easy read but I like where he is going with that research.
I stopped playing but i still enjoy your videos. Because you don't only put your heart and passion in them, but also your intelligence and you make it understandable for anyone. Thank you and good luck on your research!
xpetu you are an inspirational figure for me, firstly I learned how to play through your videos and I also learned about a new application to something I love, next week I will start my college classes in a statistics major and I can proudly say that I found someone to look up to Good luck with your project, we will miss you until your next video
Haven't watched you in years, but I always saw you as a big brained indiviudal. You got me to Diamond in about two months. Thanks man, I also quit this game a long time ago. Love
@@helluvajester the game takes up too much time, toxic player base, stupid chat restrictions, champions can't be balanced for all levels of play, items and champions change, so the game loses a part of it that maked you want to play, the game overall reduces your mental state if you care the least about it. I quit due to these. I made it to Challenger in the turkish server, probably corralates to masters low grandmasters in eu, playing nidalee top with grasp. They removed the item that made it viable, divine sunderer
Favourite league creator full stop. Been missing your more regular uploads, but I love hearing a dope person succeed in life way more so it's plenty offset and makes me happy to hear brother. Keep being real and we'll still be around when you're back
This is the exact vibe we need for the league community. Congratulations on your thesis and graduation, have fun traveling and the best of luck in your product development journey. I'm sure it will be a success.
I'm going to need a full causality analysis on whether the treatment (T = playing Shen) led to your Master's degree, or if the Master's was a confounder making you pick Shen. Please have the paper on my desk by Monday! Great video btw haha
Riot has always seemed to have an issue with data analysis where they know enough to collect and aggregate an enormous amount of data, but not enough to avoid all of the huge pitfalls inherent to statistical analysis. Every so often, Phreak accidentally makes this obvious in his videos. Most recently, at least to my knowledge, when he defended the ping changes by saying that survey responses showed a reduction in reports of negative communication interactions. The first comment under that video was pointing out that if those were the metrics Riot was using, their best course of action would be to remove all ability to communicate, thus bringing the credible reports of negative communication to zero. Statistics is an extremely difficult and counterintuitive field. Being smart and able to run the numbers isn't all that's required to make good use of it, but it doesn't come across as if the people at Riot appreciate that fact.
No lol. That first comment under the video is what's called slippery slope fallacy lmao. All these armchair data scientists in the comments are hilarious.
umm no the best course of action would be to delete the game because therere still some ways to communicate disrespect to others without pings and chat
As much as your videos make me so happy, we all have goals that we strive for and need to achieve. I wish you the best of luck, and hope that in your travels, there is a chance I get to meet you. Thank you, and I can not wait to wait for more tech in December.
as someone who will be graduating next week, hearing ya plans of what you'll be doing is quite inspiring. even if i am not a big data-person myself it is a pleasure sensei and until the holidays come, and as someone whose favorite season is the spooky season, i will wish you a nice fall as well
I love how you took a something that you love and applied your expertise to it, to then share your finding with others. This is a beautiful representation of human ingenuity and our ability to meaningfully connect subjects that would perhaps at first sight seem distant. Love what you do. Keep it up bro ❤ .
It was very entertaining to see, thank you! I hope you find lots of support for Coachless, and bring a qualitative change to understanding of itemization and item balance in league (and maybe all other mobas). Makes me wonder, how wide could the application range extend.
I stopped playing League probably 2 years ago, but I still enjoy watching your content and you almost make me want to get back into the game. Keep up the great work man! I hope your endeavors are successful!
Bro I love numbers and maths as a whole, this like genuinely fascinated me. Its actually so crazy how these statics are quite misleading so I would definitely love to see if you developed something for league that eliminates the selection bias
League is the biggest scam on history. People who play near the server have 1 to 3 seconds in advantage, and the proof is there. All games have many server locations except league!
You are the most positive and interesting LoL youtuber I have ever watched. I love your analytics, comments and thoughts about the game, they are all very useful and interesting to listen to. I find both videos like this and your normal videos very enjoyable. Thank you for the years of entertainment, I hope you have a pleasant journey and am looking forward to your next video when you come back!
This was extremely interesting, and I will definitely take a look at coachless! I'm a maths teacher and theorycrafting nerd myself, so this was a holy grail of nerdgasm. Thank you for your effort and for the link; will definitely read (the parts not above my head of) your thesis! All the best this fall
Great video as per usual, petu. You might be the first person to come up with this insight of an item's average influence on win rate which I think is fantastic. I have a few ideas for you to consider if you haven't already. Firstly, the idea of 'lose less' items on top of 'win more' ones as well, so one that works well when behind. Essentially this idea can be extended to all levels of prior win rates, so items that work well when even, etc. when you take this into account, then essentially the win difference should really be localised to small bins of prior win rates. (This doesn't even begin to consider the prior win rate mechanism - how to estimate this especially when champions have different win graphs over time i.e. scaling vs early game comps) extending this further, items are really dependent on matchups (you highlighted the difference between matchups against MR and armour), although how fed the enemy team magic vs attack dmg carries are also influences which item is optimal. also rest of team should influence it (e.g. for a top laner, jg dmg is especially important, mid to a lesser degree, and bot/supp to even less). team comps also matter, e.g. liandries might have a disproportionately high win rate against tanks, meaning that regardless of whether another item is better when ahead or behind (e.g. mejais if ahead or zhonyas/banshees if behind), liandries is still optimal against more than 3 tanks on enemy team. this idea extends to basically all items against more uniform teams (e.g. frozen heart against auto attackers, zhonyas into assassins, thornmails into healers, serpents into shielders, bork, etc). In fact there might be some self selection already in champ select, e.g. rammus picked more into AAs, gwen picked into tanks, etc, that kind of skew builds from the get go. Items are super complicated, and essentially when broken down into matchups is almost impossible to model. Perhaps when you ignore that and a few other aspects (e.g. different itemisations of scaling vs win now comps, teamfight vs split comps, and stuff mentioned earlier), then perhaps you can end up with some manageable PDEs. Perhaps this taps into the most important part, which is the fact that a human can analyse all these different aspects, to purchase good items so frequently and so quickly, shows the amazingness of our gamer brains. good job on your grad, enjoy your holiday, and best of luck on the chapters ahead.
@@xPetu Hahaha I made an almost identical comment to this one before scrolling down and seeing it, love to see someone else thought the same thing and also thought about it. Does make me wonder, if you took all of this analysis we do and tried to make a high level bot out of it, what level it could get to. If you got a product you were happy with, it would be fascinating if you could get special permission from Riot to let it loose in soloqueue and see where it ends up
I agree with everything you said! You have so many thing to factor in if you want to make a reliable argument it almost seems impossible. I think the endgame of these analyses will be to just hand all the recordings from the latest patch to an AI, and just start giving it questions. Skipping all of the "1 question leads to 3 questions, leads to 9 questions" problem. And the AI just takes EVERYTHING into consideration and says "buy item x, y, z ... "
Happy that your can achieve more form League than most, especially from studying Mathematical Algorithms and applied Statistics Edit: Happy Holiday whenever you decide to upload, and always remember- you can always learn, after you finished your formal studies
I love how when you start looking at data and statistics how easy it is that you remove yourself so much from the original domain, that you forget to look at what the objects really do in the domain.
@@Kepitano Yea I was too tired when writing this😂. Basically I wanted to say that, no matter the field, if you start analysing the statistics of a process, you should always go back and revise what those Steps exactly do in the context of the chain to be able to make informed analytical decisions. It can always happen that certain parts of a process-chain look more impactful than they really are, but they only ARE that impactful because of parts of the process which you did not consider in your firsthand analysis. Ps: - You can swap out process-chain with objects or whatever. - Damn it's hard to formulate stuff in a generalised way without it getting out of hand quickly 😂
Congrats about your finished student life. I love hearing you simplify complex concept within a game. Then I can apply them elsewhere less fun, but more useful =) Thanks!!!!
Congratulations on the graduation! The tech is great! super interesting data. If I was actually playing league, I'd definitely be paying for coachless just to go into the numbers. I guess we finally have the mathematically correct build.
Was waiting for this to come out! Congratulations on finishing the Master's and good luck in your endeavors. I will definitely look at your thesis and try to understand XD, the linear models course I am taking rn should help.
This is an incredibly satisfying video. Thank you, truly. I'm a huge numbers nerd and love to OBSORV and learn. Bring Baus into the fold for data analysis! his knowledge is underrated, especially of matchups and item purchase order IMO.
That is actually such a cool field Though even with the win probability added, I feel like selection bias still plays a part, since you're more likely to buy defensive items when it's *likely* that your win probability is going to drop (thus meaning the average win probability will be lower than an item you buy when it's likely it'll rise anyways). Since the win probability depends on your team and the state of the enemy team as well, there would be an expected change in win probability regardless of specific item choices (mostly leaning towards the team that's winning slowly winning more and more). But then, comparing items specifically in games with similar starting win probabilities would inform whether an item might actually be the better choice when behind or not, and you mentioned selecting for specific conditions which could fix that issue.
What you're basically doing now is referring to the second order terms of the Taylor expansion of whatever function governs the win rate, by looking at the (expected/perceived) rate of increase/decrease of the winrate. Basically (this is speaking from a physicist's method of analysis not a statistician's) by analysing higher and higher orders you approximate the real (currently unknown) function.
xPetu I'm so happy for u man, crazy how time flies.I started paying a lot of Shen when I was in the secondary, and now I'm too doing a engineering career, I hope to be has intelligent like u ingenier xPetu. Greatings and congratulations from Argentina!
Thank you for mentioning that Kim et al. reference! This has given me inspiration for a term project needed for a class I am taking this semester about Neural Networks :)
THANK YOU for making this video! Ever since taking taking a statistics course in college I've felt like a bunch of available game statistics are skewed and misinterpreted. I can't tell you how many times I've been in games where someone build's sub-optimally for the situation simply because the build "has the highest winrate". I hope more people notice this and adopt a modular build mindset.
These things (item winrate data, op-gg, tier lists) only help people in lower ranks, like lower than Diamond. Blindly following guides or popular builds is oftentimes a common practice in low elo, and it is what's making people lose and not improve at League. All you need to do is understand matchups and the basic needs of your champion. For Example Zhonya's is great against Fizz despite it granting 50 armor; on the other hand, Banshee's isn't optimal vs Cassiopeia despite it granting 50 magic resist. You'd use Zhonya's active against Fizz's ult, and the spellshield passive is useless because Cassiopeia could easily pop & ignore it due to her short cooldowns. I'm hoping those people I speak about would be more aware of such observations, but in the end, they are low rank for good reason :)
this is very interesting, I was allways looking for someone or some programm to do the exact calculations about correct itemisation.. I hope you'll succeed with your projekt!
nah mate take ur time to enjoy the world. you will never be this healthy and agile. go on, take ur time. im just happy youre making progress :) ive been watching for like 5 years or more
This is super interesting and i love this concept and idea, ive asked this question about item WR's myself but actually finding a way to answer it and explain it is incredible, great video!
i admire you a lot man, im struggling with college right now (im studying economics) but seeing you playing the game and finishing your degree makes me soo happy and gives me motivation. i hope everything goes perfect in your future, i always love your shen gameplay and your analysis.
If the available data includes timestamps on when the item was purchased during a game, it would be interesting to see how that items win rate changes depending on when it was bought as a first item. Then you could also compare that same item to how the win chance changes if you managed to get your first item before your role opponent did. Then you could go on and see the same things for the second items win chance changes. Furthermore you could analyze more in depth how the item win chance is affected by melee physical, ranged physical, melee magic and ranged magic matchups. (Watched the whole video, but unsure if it was only for physical and magic or if it also included melee and ranged)
It's been such a long time since I've seen such an interesting video about LoL. Used to main shen more than 5 years ago, havent touch the game since then. Always nice to see something so impressive and more importantly from a fellow Shen main.
I'm so impressed by the work you've done on Coachless. You are truly an enigma amongst the plethora of LOL content creators on this platform. Keep it up!!! You've gotten me interested in LOL Statistics, and perhaps statistics as a whole!
Super interesting way of encapsulating this topic. I wonder if you looked into how buying an item after the other compounds their win probability differential and if its a linear or exponential relation, or somehow has diminishing returns or even an inverse relation. Would be fun (although probably incredibly tricky) to expand this study into looking at all 6 slots in the build. Good stuff brother.
Watched this while making dinner and honestly insanely interesting points brought up. Had my attention all video. Very excited to see the future of coachless. Good luck with everything Petu!
Way you live your life is very inspiring to me and I really appreciate videos like this. Congratulations on your graduation and I hope it will all work out!
Don’t know who you are but this is quality content. I’d like to see an addition specifically of items for when a player is behind, or has a below 50% win chance, further refined by whether they were against ad or ap. It would be nice to know the best item to stabilize when you’re losing, not just which one to push your advantage when you’re winning
Just wanted to say that everything in the video is very well done, but you are fundamentally misrepresenting "selection bias". Selection bias does not occur because a human has made a selection, unless you are modelling items based on sampling from a pool of items. Your study would suffer from selection bias if you were only looking at a group of pro players and generalizing to the whole pool of league players. Selection bias occurs when the choosing of the DATA is biased, not the randomness of the data. Rather, the selection you are referring to IS your data, and what you are trying to do is to infer whether or not this choice of the player to purchase item X affect the win rate, the probability of winning, given that X was purchased. What you should be talking about is survivorship bias in relation to certain items with high win rate, as items that are NOT purchased when behind are obviously inflated in winrate because they are ONLY or MOSTLY purchased when ahead. Selection bias has not been introduced to your study because players make a choice, rather your data is non-random, and your inferences must take into account this fact.
Have you seen Shending help is smurfing on the EUW ladder recently? Either shen is in a great spot this season or Shending is a monster on your champion (he is monstergood either way xD)
@@xPetuRiot was really stupid this last patch. They buffed champions that were already good/strong and pushed them into OP just because they want to see them at worlds. Jarvan Jax etc, absolutely no need for buffs lol.
That statics is awesome and for sure opens a lot of possibilities for better analysis. But isn't there a flaw in saying a high deltaW implies that the Item is better in most situations: A high deltaW could also imply that the winrate of having the components of the item is bad. Meaning the build path of the item is so bad that it artificially increases the deltaW. In context that on shen the defensive items have a worse deltaW, could also mean that the components of tank items are generally stronger than the components of offensive items. Which in my intuition would actually make a lot of sense since many tank items are mostly bought for their stats not for their abilities, but bruiser items are often bought because their abilities greatly synergies with the specific champion. Of course more analysis would be needed in order to check that thesis. But I still think that it would probably make more sense to make deltaW = W1-W2 where W1 = Winrate of selling the item W2 = estimated Winrate after buing the legendary item before the one you currently purchased At last I have to say that I didn't check if you already did that. If that's the case sorry.
You could probably also add the dW for the components (when they are used to make that item.) Though if you have a negative dW I don't think it matters, because that means the components are stronger, but you're building toward an item that is worse, so you might as well buy more of the base items. But yeah as soon as it's positive, you might get bias where like some previous item is just as good as the finished item, so you get a dW of 0, but it's also better than another legendary.
This man probably revolutionized the computing of items worthness in League of Legends. I am 100% sure that even big competitive teams didn't actually do these calculations, this means that with enough players testing different builds we should have the OPTIMAL item for each champions for a said patch. Having this information has no price today when you think of the amounts of money involved. If just one team could a systematicaly have the better builds that would probably increase their winrate. Congrats man you probably started a new era for the competitive League of Legends community
These calculations still are not really reliable as a guideline i think (afaik my statistic knowledge is pretty rusty tho) because the Data inherently is flawed/biased.
But how do you get the win probability for the match at the time the item is purchased? I mean you can't take 100 matches that have the exact same state of the match when the item is purchased and then see how many of them ended up in a victory/defeat... Because there will be not one single match that is exactly the same as the other at the time the first item is purchased. Even when you say let not take the "exact same state" but just "very roughly the same state", there are way too many factors that are relevant (tall 10 champs in the match, objectives done, items of all champs, platings, turrets, gold , positioning on the map of all players etc...). So either you get no examples to compare to get a win rate. Or your samples are way to different from each other and therefore not comparable. So there must be another way how you got this win probability. Id really love to know how since i like this topic and your work!
I skimmed the thesis, and the thesis does indeed admit that the variance of win probability is sketchy to analyse cause of the chaotic nature of league matches, but its aided by three factors. The first is the use of statistical average over a large data set, we're talking dozens of thousands of games with a similar enough game state. As such the odd game in which human error or a statistical improbability occurs does not significantly alter the average winrate variance. The second is the bias of league players in their item selection. Because items are chosen based on game state, the conclusions you can take from looking at item winrate variance is reliable. For instance, Shen players tend to buy hearsteel in winning lanes, and winrate increase shows this to be a good item choice. On the other hand, hollow radiance is bought into AP matchups, but the fact winrate decreases shows that its the wrong item to buy in this gamestate. The third is that the thesis uses initial and post-purchase winrate probability to determine winrate variance, so factors that occur after the first item is purchased are irrelevant. Sure, the actual variance is dependent on the gamestate of your particular game, so the average isnt practically reliable. If you buy heartsteel as shen into vayne as your team is down five kills, for instance, you are probably not increasing your winrate probability. But the fact that you are buying an item inappropriate for your gamestate makes you a statistical minority.
I've pretty much recently discovered your content and, aside from that accent that feels so satisfactory to hear, your commentary and demeanor gives me the "this is an individual I can learn from" plus is entertaining to watch. You've become another star in my night sky of fav content creators, but in the unique category of "The Bob Ross of league but he also has some cool teacher vibes". I wish you the very best!
I quit league out of necessity; I can't game that long without a controller anymore due to fatigue and pain. I still am subscribed and like to watch your channel. Your approach to things and the way you articulate your thoughts is really pleasant to listen to. Good luck with your future endeavors.
sorry for the dumb question but is it even possible to play with a controller? that sounds impossible, to me - I suppose you're heavily used to the controller hence why you find it easier? I'm confused, really. I can't imagine playing this game with a controller :(
@@InuKagKikyI play with a steam deck, I’m not great by any means but I got 54% wr on jinx and I’ve gotten to mastery 40 in the past year. I can’t imagine playing with a joystick but with the trackpad I heavily prefer it over a mouse and kb. I enjoy being able to just lay in bed watching Tv while I play league.
Hey man, this is the first video I saw from you. I stopped playing league years ago but I always wanted to code a tool like yours. Your video got me really hooked since I just started statistical machine learning at Uni. I think your tool is really cool and I hope you succeed with your journey!
I watched you almost since day 1. Got really into Shen and hit 500k on him. Then i stopped playing league and so i havent watched you ever since. This video have brought me ton of nostalgia just by hearing your voice talking about statistics and more of the things i dont understand XD. I thank you for all your videos. Hours of content i could enjoy. I hope everything works out well for you in life. Much love bro
Fantastic video! As someone interested in machine learning and statistics I had been thinking why this type of actually useful data was not available since the game is so popular.
Just found this by the algorithm. I don't even play league anymore but I love your style and you seem like a great guy. Very interesting stuff! Makes me want to see this in other games too like TFT !
Man, I really really hope you and your team/project finds success, from hearty I say this. Being able to see your growth and dedication to this game makes me have some more hope in people haha. I will try your product knowing that there will be people behind it that actually care about doing something great. I understand that money is the main reason and that's more than fine, makes sense, but the fact that there is passion for the game included makes it special and I'm thankful for that.
Due to popular demand, here's a Google drive link to my thesis: drive.google.com/file/d/1CkYW0JdMfXXcczFvqjXcfp64D8eMIq1d/view?usp=sharing
Huge thank you to the 1,000 people who subscribed to Coachless in its early stage. We changed our default subscription period from 90 to 30 days now, making the Item Analysis feature more accessible. Free version is now also implemented: coachless.gg/items
Just gotta say watching this video and I remember seeing that specifically sorc boots, into mejais soulstealer, into rabadon on vladimir has a 76% winrate. This was the first time I realized this since well 76% winrate is absolutely bonkers and mejais is only ever useful especially as a first item when you are absolutely dumpstering your opponent. Plus if you are playing a late game champ like vlad and are winning that hard in the early game it makes sense there is almost 0 way you can lose besides you having a stroke or all your teammates having one.
How about for a video title: "How xPetu's Thesis on League of Legends made me stop watching his videos"
I will be waiting for you GG I hope you can find a good Senna otp to learn from as well I'd love to see how one of my favorite champions is itemized i have been too scared to comeback to league since they made the extra nerfs to her damage.
xPetu, you inspired me to dive into the world of math and discover it's wonders. I don't plan on majoring in it but the esoteric and fundamental knowledge I have gained has been partly because of you. When I become a crackpot engineer/scientist, I'll remember you.
Thank you for your analysis.
Bro is so invested in shen he is writing his masters thesis and paper on him
Should've written a Challenger thesis instead.
@@justinjakeashton I see what you did there 😂😂😂
He has 327k subs and almost 200k views in this video soo yeah there is people interested in this info
@@VictorAlfonsoTrades i'am br kkk
Sounds like it will help everyone on any champ and against any champion
This is the kind of analysis I wish we saw in every game dev team. I know your content is primarily LoL focused, but it would be really cool to see if this can be adapted to other MOBAs and become a sort of standard for analyzing MOBAs as a whole.
One thing I think would be cool (albeit difficult to implement) is a ML model that takes into account not just the physical vs magical damage, but the specific hero matchups and maybe even team comps as a whole. I imagine that feeding enough of that data into a model would produce some interesting predictions.
This guy made an entire thesis just to to divert our attention from the fact that he is unironically buying protobelt on shen.
thats what stat tells him
Shen is strong, all items Shen item.
but that protobelt let him catch jinx and close out the game =D
@@QuantemDeconstructor "It's a hunter item," energy.
I mean he also has Riftmaker so clearly playing some kind of AP build.
GL xPetu, hope to hear more about your academic life and work
The thing I learned from this video is that anytime I'm about to win for certain with shen, I need to recall and buy Rylais so that it has a 100% winrate and thus people start buying it and throwing their games
Unless ppl start using his stats!
@@alex.polychronopoulos4487 even more of a reason because it makes more people that care about optimization of items promote this channel
😂😂😂
Well then it would corrected itself, cause people would be losing with it.
That wouldn't affect the results he's looking at because your ΔW most likely wouldn't increase after such an action.
Hey Petu, I was an assistant coach for a MOBA esports team back in 2021 mostly dealing with data analysis. Although I did not come from a statistics background, I enjoy watching this video. Thanks and good luck on the project!
"I have done the analysis" *shuffles papers*. I love it dude.
lol i'd noticed the paper shuffling but it didn't really make sense, now i get it. gold
This is made even more clear in data analytics of a game like TFT where the Delta shows the placement change for a specific item on a specific unit. There you can clearly see exactly how much on average your item decision on each unit will affect the result of your placement.
rito, just give petu a job
Yeah. Phreaks. And consequently, get that animal out of that position.
Typical league player calling someone an animal over a game 😭
@@ttv_zombiefetus1710 Typical... I don't even know what you are, to be fair.
Do you disagree with what I said? Do you disagree with the use of the word "animal" only?
What's your point?
@@ttv_zombiefetus1710 i mean hes not wrong
They won't. xPetu just proved he is an intelligent human being. Riot has no place for such people.
you explain complex subjects in a way that is easy to grasp for someone that has never studied statistics and probability, like a professor! We love you Petu good luck on your and your teams project!
Huh?....The thesis here covers over more advanced statistics topics like Markov chains. It's not a purely easy read but I like where he is going with that research.
I am proud of you Petu, I hope you succeed in this endeavours of yours
GL on your journey my brother! Wish you the best, you deserve it! :)
I stopped playing but i still enjoy your videos. Because you don't only put your heart and passion in them, but also your intelligence and you make it understandable for anyone.
Thank you and good luck on your research!
same! came for league, stayed for the personality
xpetu you are an inspirational figure for me, firstly I learned how to play through your videos and I also learned about a new application to something I love, next week I will start my college classes in a statistics major and I can proudly say that I found someone to look up to
Good luck with your project, we will miss you until your next video
Haven't watched you in years, but I always saw you as a big brained indiviudal. You got me to Diamond in about two months. Thanks man, I also quit this game a long time ago. Love
@@helluvajester is that even a question
@@helluvajester Why wouldn't you quit
Best advice out here rn
I wanna see Xpetu have fun playing Space Marine2
@@helluvajester the game takes up too much time, toxic player base, stupid chat restrictions, champions can't be balanced for all levels of play, items and champions change, so the game loses a part of it that maked you want to play, the game overall reduces your mental state if you care the least about it. I quit due to these. I made it to Challenger in the turkish server, probably corralates to masters low grandmasters in eu, playing nidalee top with grasp. They removed the item that made it viable, divine sunderer
@@efegokselkisioglu8218 You should watch Lourlo play ad nidalee top, he makes it look super fun and pretty strong
Bro, as an engineer, it's umbeliveble nice to hear you bringing biases and data to the conversation! YOU FKING ROCKS, SHEN GOD!
Every professional speaks about biases. Otherwise they aren’t a professional.
crazy if multimillion dollar company riot games hasnt done the same calculations
@@asd-ww4nk None of the games optimizes for balance. All games optimize for fun.
@@MajkaSrajka maybe indie but big companies optimize for balance (they still fail)
Hearing "as an engineer" is the quickest way to discredit yourself tbh.
Favourite league creator full stop. Been missing your more regular uploads, but I love hearing a dope person succeed in life way more so it's plenty offset and makes me happy to hear brother. Keep being real and we'll still be around when you're back
This is the exact vibe we need for the league community. Congratulations on your thesis and graduation, have fun traveling and the best of luck in your product development journey. I'm sure it will be a success.
I'm going to need a full causality analysis on whether the treatment (T = playing Shen) led to your Master's degree, or if the Master's was a confounder making you pick Shen. Please have the paper on my desk by Monday!
Great video btw haha
😂
Riot has always seemed to have an issue with data analysis where they know enough to collect and aggregate an enormous amount of data, but not enough to avoid all of the huge pitfalls inherent to statistical analysis.
Every so often, Phreak accidentally makes this obvious in his videos. Most recently, at least to my knowledge, when he defended the ping changes by saying that survey responses showed a reduction in reports of negative communication interactions. The first comment under that video was pointing out that if those were the metrics Riot was using, their best course of action would be to remove all ability to communicate, thus bringing the credible reports of negative communication to zero.
Statistics is an extremely difficult and counterintuitive field. Being smart and able to run the numbers isn't all that's required to make good use of it, but it doesn't come across as if the people at Riot appreciate that fact.
like statistic teachers say:"raw data is garbagge data"
No lol. That first comment under the video is what's called slippery slope fallacy lmao. All these armchair data scientists in the comments are hilarious.
@@brotendo That's not what a slippery slope argument is.
@@brotendo thats not a slippery slope argument at all lol. Youve had a shocker.
umm no the best course of action would be to delete the game because therere still some ways to communicate disrespect to others without pings and chat
Dude, amazing, congrats on this study and i hope u get where u want to be!! I've been following u since a few years now, love ur content
congratulations on your thesis! this is a beautiful topics, well written, and i’m excited to see the final product!
As much as your videos make me so happy, we all have goals that we strive for and need to achieve. I wish you the best of luck, and hope that in your travels, there is a chance I get to meet you. Thank you, and I can not wait to wait for more tech in December.
as someone who will be graduating next week, hearing ya plans of what you'll be doing is quite inspiring. even if i am not a big data-person myself
it is a pleasure sensei and until the holidays come, and as someone whose favorite season is the spooky season, i will wish you a nice fall as well
WIshing all the success for Coachless and all your endeavors Shensei. You are one of a kind.
I love how you took a something that you love and applied your expertise to it, to then share your finding with others. This is a beautiful representation of human ingenuity and our ability to meaningfully connect subjects that would perhaps at first sight seem distant.
Love what you do. Keep it up bro ❤ .
Same to you Petu! Its so hype to see you follow through on this project!
It was very entertaining to see, thank you! I hope you find lots of support for Coachless, and bring a qualitative change to understanding of itemization and item balance in league (and maybe all other mobas).
Makes me wonder, how wide could the application range extend.
I stopped playing League probably 2 years ago, but I still enjoy watching your content and you almost make me want to get back into the game. Keep up the great work man! I hope your endeavors are successful!
I stopped playing League since the day I was born, but I still enjoy watching this content, yet I still don't want to get into the game.
Good decision! @@dry-cleaning6255
This is such a good analysis! Amazing work on your thesis and the video.
I never thought I'd be interested on a thesis on League of Legends, but I know what I'm gonna be reading for the next few days.
Where is it?
@@joshuahall3622 he linked it in the top comment in a google drive link
Bro I love numbers and maths as a whole, this like genuinely fascinated me. Its actually so crazy how these statics are quite misleading so I would definitely love to see if you developed something for league that eliminates the selection bias
Already miss you, Petu. But also wish you the best in making coachless as great as you can!
Shending you good vibes!
Title seems to be click-bait, he didn't quit league, and instead he just wants to sell you on his coaching platform.
League is the biggest scam on history. People who play near the server have 1 to 3 seconds in advantage, and the proof is there. All games have many server locations except league!
You are the most positive and interesting LoL youtuber I have ever watched. I love your analytics, comments and thoughts about the game, they are all very useful and interesting to listen to. I find both videos like this and your normal videos very enjoyable. Thank you for the years of entertainment, I hope you have a pleasant journey and am looking forward to your next video when you come back!
This was extremely interesting, and I will definitely take a look at coachless! I'm a maths teacher and theorycrafting nerd myself, so this was a holy grail of nerdgasm. Thank you for your effort and for the link; will definitely read (the parts not above my head of) your thesis! All the best this fall
This is actually so interesting data! Great job man, hope your project is a big success!
Great video as per usual, petu. You might be the first person to come up with this insight of an item's average influence on win rate which I think is fantastic. I have a few ideas for you to consider if you haven't already. Firstly, the idea of 'lose less' items on top of 'win more' ones as well, so one that works well when behind. Essentially this idea can be extended to all levels of prior win rates, so items that work well when even, etc. when you take this into account, then essentially the win difference should really be localised to small bins of prior win rates. (This doesn't even begin to consider the prior win rate mechanism - how to estimate this especially when champions have different win graphs over time i.e. scaling vs early game comps)
extending this further, items are really dependent on matchups (you highlighted the difference between matchups against MR and armour), although how fed the enemy team magic vs attack dmg carries are also influences which item is optimal. also rest of team should influence it (e.g. for a top laner, jg dmg is especially important, mid to a lesser degree, and bot/supp to even less).
team comps also matter, e.g. liandries might have a disproportionately high win rate against tanks, meaning that regardless of whether another item is better when ahead or behind (e.g. mejais if ahead or zhonyas/banshees if behind), liandries is still optimal against more than 3 tanks on enemy team. this idea extends to basically all items against more uniform teams (e.g. frozen heart against auto attackers, zhonyas into assassins, thornmails into healers, serpents into shielders, bork, etc). In fact there might be some self selection already in champ select, e.g. rammus picked more into AAs, gwen picked into tanks, etc, that kind of skew builds from the get go.
Items are super complicated, and essentially when broken down into matchups is almost impossible to model. Perhaps when you ignore that and a few other aspects (e.g. different itemisations of scaling vs win now comps, teamfight vs split comps, and stuff mentioned earlier), then perhaps you can end up with some manageable PDEs. Perhaps this taps into the most important part, which is the fact that a human can analyse all these different aspects, to purchase good items so frequently and so quickly, shows the amazingness of our gamer brains.
good job on your grad, enjoy your holiday, and best of luck on the chapters ahead.
yep all matchups related / context-specific stuff is still coming, working on it tirelessly!
@@xPetu Hahaha I made an almost identical comment to this one before scrolling down and seeing it, love to see someone else thought the same thing and also thought about it. Does make me wonder, if you took all of this analysis we do and tried to make a high level bot out of it, what level it could get to. If you got a product you were happy with, it would be fascinating if you could get special permission from Riot to let it loose in soloqueue and see where it ends up
I agree with everything you said! You have so many thing to factor in if you want to make a reliable argument it almost seems impossible. I think the endgame of these analyses will be to just hand all the recordings from the latest patch to an AI, and just start giving it questions. Skipping all of the "1 question leads to 3 questions, leads to 9 questions" problem. And the AI just takes EVERYTHING into consideration and says "buy item x, y, z ... "
You are awesome, love these types of analyses and it's conclusions. Congrats and wish you luck!
Happy that your can achieve more form League than most, especially from studying Mathematical Algorithms and applied Statistics
Edit: Happy Holiday whenever you decide to upload, and always remember- you can always learn, after you finished your formal studies
Always enjoyed your content. Best of luck with your academic endeavors and with coachless o7
Hoping you have success on your endeavor!
A bit sad there will be no more videos for a while, but you gotta do what you gotta do
Man, I love you! U are an awesome person, and I'm sure that you will succeed in whatever project you are doing! praying for u ma man
I love how when you start looking at data and statistics how easy it is that you remove yourself so much from the original domain, that you forget to look at what the objects really do in the domain.
what are you talking about
@@manphan6689domain knowledge is very important in data intepretation
@@manphan6689you look at game numbers too long you forget original game
What the heck? You ought to rewrite that bro
@@Kepitano Yea I was too tired when writing this😂.
Basically I wanted to say that, no matter the field, if you start analysing the statistics of a process, you should always go back and revise what those Steps exactly do in the context of the chain to be able to make informed analytical decisions. It can always happen that certain parts of a process-chain look more impactful than they really are, but they only ARE that impactful because of parts of the process which you did not consider in your firsthand analysis.
Ps:
- You can swap out process-chain with objects or whatever.
- Damn it's hard to formulate stuff in a generalised way without it getting out of hand quickly 😂
Congrats about your finished student life. I love hearing you simplify complex concept within a game. Then I can apply them elsewhere less fun, but more useful =) Thanks!!!!
Yoo i also did my Masters Thesis on League of Legends too! That's so wild that 4 months after my graduation I see this video lol
I have a friend doing it now too XD
whats the title/theme?
They're developing a program that uses computer vision to detect League of Legends twitch streams' highlights
name and title please i want to read some of it
Really impressive content. Thank you for being one of the only content creators to actually elevate our level of knowledge on the game.
Congratulations on the graduation!
The tech is great! super interesting data. If I was actually playing league, I'd definitely be paying for coachless just to go into the numbers.
I guess we finally have the mathematically correct build.
Good luck with your endeavours Petu! And have safe travels.
Was waiting for this to come out! Congratulations on finishing the Master's and good luck in your endeavors. I will definitely look at your thesis and try to understand XD, the linear models course I am taking rn should help.
This is an incredibly satisfying video. Thank you, truly. I'm a huge numbers nerd and love to OBSORV and learn. Bring Baus into the fold for data analysis! his knowledge is underrated, especially of matchups and item purchase order IMO.
That is actually such a cool field
Though even with the win probability added, I feel like selection bias still plays a part, since you're more likely to buy defensive items when it's *likely* that your win probability is going to drop (thus meaning the average win probability will be lower than an item you buy when it's likely it'll rise anyways). Since the win probability depends on your team and the state of the enemy team as well, there would be an expected change in win probability regardless of specific item choices (mostly leaning towards the team that's winning slowly winning more and more).
But then, comparing items specifically in games with similar starting win probabilities would inform whether an item might actually be the better choice when behind or not, and you mentioned selecting for specific conditions which could fix that issue.
Stop overthinking it, buy item accoirding to your GAMEPLAY STYLE and what your team needs to WIN.
Everything else is sperging into nothingness.
What you're basically doing now is referring to the second order terms of the Taylor expansion of whatever function governs the win rate, by looking at the (expected/perceived) rate of increase/decrease of the winrate.
Basically (this is speaking from a physicist's method of analysis not a statistician's) by analysing higher and higher orders you approximate the real (currently unknown) function.
Kind of like how the build sites show Mejai’s as having a really high win rate on mages so you should always buy Mejai’s, right?
xPetu I'm so happy for u man, crazy how time flies.I started paying a lot of Shen when I was in the secondary, and now I'm too doing a engineering career, I hope to be has intelligent like u ingenier xPetu.
Greatings and congratulations from Argentina!
Good luck on your project xPetu
Thank you for mentioning that Kim et al. reference! This has given me inspiration for a term project needed for a class I am taking this semester about Neural Networks :)
THANK YOU for making this video! Ever since taking taking a statistics course in college I've felt like a bunch of available game statistics are skewed and misinterpreted. I can't tell you how many times I've been in games where someone build's sub-optimally for the situation simply because the build "has the highest winrate". I hope more people notice this and adopt a modular build mindset.
These things (item winrate data, op-gg, tier lists) only help people in lower ranks, like lower than Diamond.
Blindly following guides or popular builds is oftentimes a common practice in low elo, and it is what's making people lose and not improve at League. All you need to do is understand matchups and the basic needs of your champion. For Example Zhonya's is great against Fizz despite it granting 50 armor; on the other hand, Banshee's isn't optimal vs Cassiopeia despite it granting 50 magic resist. You'd use Zhonya's active against Fizz's ult, and the spellshield passive is useless because Cassiopeia could easily pop & ignore it due to her short cooldowns. I'm hoping those people I speak about would be more aware of such observations, but in the end, they are low rank for good reason :)
this is very interesting, I was allways looking for someone or some programm to do the exact calculations about correct itemisation.. I hope you'll succeed with your projekt!
nah mate take ur time to enjoy the world. you will never be this healthy and agile. go on, take ur time. im just happy youre making progress :) ive been watching for like 5 years or more
This is super interesting and i love this concept and idea, ive asked this question about item WR's myself but actually finding a way to answer it and explain it is incredible, great video!
5:20 saddam hussein hiding in xpetu's masters paper
XD
I was about to comment that, pattern recognition is truly incredible
We will miss you master, but I hope you enjoy your trip. We love you, take care!
Good luck on your travels brother. Very interesting content
Hey man have a incredible successful journey
i admire you a lot man, im struggling with college right now (im studying economics) but seeing you playing the game and finishing your degree makes me soo happy and gives me motivation. i hope everything goes perfect in your future, i always love your shen gameplay and your analysis.
If the available data includes timestamps on when the item was purchased during a game, it would be interesting to see how that items win rate changes depending on when it was bought as a first item. Then you could also compare that same item to how the win chance changes if you managed to get your first item before your role opponent did.
Then you could go on and see the same things for the second items win chance changes.
Furthermore you could analyze more in depth how the item win chance is affected by melee physical, ranged physical, melee magic and ranged magic matchups. (Watched the whole video, but unsure if it was only for physical and magic or if it also included melee and ranged)
You're a legend, thanks for your work/information and I hope you'll be able to make your dream come to reality!
I could watch these videos for hours!
Good look and I am sure coachless can only improve with you on the team love your videos and can’t wait for the return
Spoiler Blocker!!!
It's been such a long time since I've seen such an interesting video about LoL.
Used to main shen more than 5 years ago, havent touch the game since then. Always nice to see something so impressive and more importantly from a fellow Shen main.
I'm so impressed by the work you've done on Coachless. You are truly an enigma amongst the plethora of LOL content creators on this platform. Keep it up!!!
You've gotten me interested in LOL Statistics, and perhaps statistics as a whole!
Super interesting way of encapsulating this topic. I wonder if you looked into how buying an item after the other compounds their win probability differential and if its a linear or exponential relation, or somehow has diminishing returns or even an inverse relation. Would be fun (although probably incredibly tricky) to expand this study into looking at all 6 slots in the build. Good stuff brother.
Watched this while making dinner and honestly insanely interesting points brought up. Had my attention all video. Very excited to see the future of coachless. Good luck with everything Petu!
Thank mr Swain
Way you live your life is very inspiring to me and I really appreciate videos like this. Congratulations on your graduation and I hope it will all work out!
Don’t know who you are but this is quality content. I’d like to see an addition specifically of items for when a player is behind, or has a below 50% win chance, further refined by whether they were against ad or ap. It would be nice to know the best item to stabilize when you’re losing, not just which one to push your advantage when you’re winning
Nice video. Very fascinating and showcasing a different way of focusing on probability in league. Thanks for your video.
Just wanted to say that everything in the video is very well done, but you are fundamentally misrepresenting "selection bias". Selection bias does not occur because a human has made a selection, unless you are modelling items based on sampling from a pool of items. Your study would suffer from selection bias if you were only looking at a group of pro players and generalizing to the whole pool of league players. Selection bias occurs when the choosing of the DATA is biased, not the randomness of the data.
Rather, the selection you are referring to IS your data, and what you are trying to do is to infer whether or not this choice of the player to purchase item X affect the win rate, the probability of winning, given that X was purchased. What you should be talking about is survivorship bias in relation to certain items with high win rate, as items that are NOT purchased when behind are obviously inflated in winrate because they are ONLY or MOSTLY purchased when ahead. Selection bias has not been introduced to your study because players make a choice, rather your data is non-random, and your inferences must take into account this fact.
that is inspiring brother! good luck with your project man.
Have you seen Shending help is smurfing on the EUW ladder recently? Either shen is in a great spot this season or Shending is a monster on your champion (he is monstergood either way xD)
yeah Shending is cracked (+ Shen buffed for no reason when he's already strong XD)
@@xPetuRiot was really stupid this last patch. They buffed champions that were already good/strong and pushed them into OP just because they want to see them at worlds. Jarvan Jax etc, absolutely no need for buffs lol.
Absolute banger video, thank you for this!
2:45 flash is insane
I started playing Shen this split 3. Thank you for your guides Petu. See you on December.
0:42 what is that unholy ignite range???
exactly! thank you!
That statics is awesome and for sure opens a lot of possibilities for better analysis. But isn't there a flaw in saying a high deltaW implies that the Item is better in most situations:
A high deltaW could also imply that the winrate of having the components of the item is bad. Meaning the build path of the item is so bad that it artificially increases the deltaW.
In context that on shen the defensive items have a worse deltaW, could also mean that the components of tank items are generally stronger than the components of offensive items. Which in my intuition would actually make a lot of sense since many tank items are mostly bought for their stats not for their abilities, but bruiser items are often bought because their abilities greatly synergies with the specific champion.
Of course more analysis would be needed in order to check that thesis. But I still think that it would probably make more sense to make
deltaW = W1-W2 where
W1 = Winrate of selling the item
W2 = estimated Winrate after buing the legendary item before the one you currently purchased
At last I have to say that I didn't check if you already did that. If that's the case sorry.
You could probably also add the dW for the components (when they are used to make that item.)
Though if you have a negative dW I don't think it matters, because that means the components are stronger, but you're building toward an item that is worse, so you might as well buy more of the base items.
But yeah as soon as it's positive, you might get bias where like some previous item is just as good as the finished item, so you get a dW of 0, but it's also better than another legendary.
Congrats on graduating! It's been wild watching you evolve over the years, I hope you have an amazing time traveling.
This man probably revolutionized the computing of items worthness in League of Legends. I am 100% sure that even big competitive teams didn't actually do these calculations, this means that with enough players testing different builds we should have the OPTIMAL item for each champions for a said patch. Having this information has no price today when you think of the amounts of money involved. If just one team could a systematicaly have the better builds that would probably increase their winrate. Congrats man you probably started a new era for the competitive League of Legends community
These calculations still are not really reliable as a guideline i think (afaik my statistic knowledge is pretty rusty tho) because the Data inherently is flawed/biased.
Congrats for your work,
and thank you for bringing a more scientific approach to video games,
I wish you the best for your project!
But how do you get the win probability for the match at the time the item is purchased? I mean you can't take 100 matches that have the exact same state of the match when the item is purchased and then see how many of them ended up in a victory/defeat... Because there will be not one single match that is exactly the same as the other at the time the first item is purchased. Even when you say let not take the "exact same state" but just "very roughly the same state", there are way too many factors that are relevant (tall 10 champs in the match, objectives done, items of all champs, platings, turrets, gold , positioning on the map of all players etc...). So either you get no examples to compare to get a win rate. Or your samples are way to different from each other and therefore not comparable.
So there must be another way how you got this win probability. Id really love to know how since i like this topic and your work!
yeah, i was kind of confused by this part too...
I skimmed the thesis, and the thesis does indeed admit that the variance of win probability is sketchy to analyse cause of the chaotic nature of league matches, but its aided by three factors. The first is the use of statistical average over a large data set, we're talking dozens of thousands of games with a similar enough game state. As such the odd game in which human error or a statistical improbability occurs does not significantly alter the average winrate variance. The second is the bias of league players in their item selection. Because items are chosen based on game state, the conclusions you can take from looking at item winrate variance is reliable. For instance, Shen players tend to buy hearsteel in winning lanes, and winrate increase shows this to be a good item choice. On the other hand, hollow radiance is bought into AP matchups, but the fact winrate decreases shows that its the wrong item to buy in this gamestate. The third is that the thesis uses initial and post-purchase winrate probability to determine winrate variance, so factors that occur after the first item is purchased are irrelevant.
Sure, the actual variance is dependent on the gamestate of your particular game, so the average isnt practically reliable. If you buy heartsteel as shen into vayne as your team is down five kills, for instance, you are probably not increasing your winrate probability. But the fact that you are buying an item inappropriate for your gamestate makes you a statistical minority.
DNN trained on millions of games, described in my thesis.
I've pretty much recently discovered your content and, aside from that accent that feels so satisfactory to hear, your commentary and demeanor gives me the "this is an individual I can learn from" plus is entertaining to watch.
You've become another star in my night sky of fav content creators, but in the unique category of "The Bob Ross of league but he also has some cool teacher vibes".
I wish you the very best!
This is incredibly interesting man! I understand writing papers like these take a lot of time but i would love to see more LoL statistics
I quit league out of necessity; I can't game that long without a controller anymore due to fatigue and pain.
I still am subscribed and like to watch your channel. Your approach to things and the way you articulate your thoughts is really pleasant to listen to.
Good luck with your future endeavors.
sorry for the dumb question but is it even possible to play with a controller? that sounds impossible, to me - I suppose you're heavily used to the controller hence why you find it easier? I'm confused, really. I can't imagine playing this game with a controller :(
@@InuKagKikyprobs just wrist issues like carpal tunnel or arthritis making it painfull to use a keyboard/mouse
@@InuKagKiky They don't play league with a controller, that's exactly why they quit it out of necessity
@@piroshiki-san oh my, thanks for that, makes sense now
@@InuKagKikyI play with a steam deck, I’m not great by any means but I got 54% wr on jinx and I’ve gotten to mastery 40 in the past year.
I can’t imagine playing with a joystick but with the trackpad I heavily prefer it over a mouse and kb.
I enjoy being able to just lay in bed watching Tv while I play league.
Hey man, this is the first video I saw from you. I stopped playing league years ago but I always wanted to code a tool like yours. Your video got me really hooked since I just started statistical machine learning at Uni. I think your tool is really cool and I hope you succeed with your journey!
Hope you enjoy ML! Highly recommended: 3b1b deep learning / linear algebra playlists on UA-cam
Thanks man! I'll definitely check them out!
hello petu
I watched you almost since day 1. Got really into Shen and hit 500k on him. Then i stopped playing league and so i havent watched you ever since. This video have brought me ton of nostalgia just by hearing your voice talking about statistics and more of the things i dont understand XD. I thank you for all your videos. Hours of content i could enjoy. I hope everything works out well for you in life. Much love bro
Fantastic video! As someone interested in machine learning and statistics I had been thinking why this type of actually useful data was not available since the game is so popular.
Just found this by the algorithm. I don't even play league anymore but I love your style and you seem like a great guy. Very interesting stuff! Makes me want to see this in other games too like TFT !
Petu legit my goat, glad to hear you're doing fine and managing to follow things you're passionate about.
Brother ShenPetu, this is fascinating. I would love more content about data analysis in regards to games.
Man, I really really hope you and your team/project finds success, from hearty I say this.
Being able to see your growth and dedication to this game makes me have some more hope in people haha.
I will try your product knowing that there will be people behind it that actually care about doing something great. I understand that money is the main reason and that's more than fine, makes sense, but the fact that there is passion for the game included makes it special and I'm thankful for that.