I Spent Two Years on a Secret League of Legends Project
Вставка
- Опубліковано 4 вер 2024
- Save Your Time and LP with Coachless: coachless.gg/
Join my Discord server: / discord
⇩ More Information ⇩
// VIEWER SUPPORT
Special thanks to Henry, Cyber XY, Tareyus, Shen, Daemoss420 and all River Shens for supporting the channel! Channel memberships allow you to support me and also receive various perks. If you're interested, click the Join-button for more details.
// LEARN SHEN FROM ME
► How to Climb to Diamond with Shen: • How to ACTUALLY Climb ...
► The Art of Shen (Ultimate Shen Guide): • The Art of Shen - An I...
► In-Depth Shen Guide: www.mobafire.c...
// WHERE TO FIND ME
► Twitch: / xpetu
► Twitter: / xpetu
► Discord: / discord
Business inquiries: xpetu@afkcreators.com
#LeagueOfLegends #Challenger #Shen
You can't trick me Petu, this project is also part of your thesis 😂
What do you mean? That IS his thesis
.
@@nebulista his cruel angel's
Truly the dichotomy of LP gains and loses, as well as their corresponding catalysts, offer immense insight towards the nature of why we spend so much time participating in this hellscape. xD
It felt that way😂
greetings dear petu
Why does ur comment say 5 days ago… what’s going on here????
@@jasonochoa7876 does he know
greetings dear viewer
For some reason my mind read this in Petu's voice.
how is this 5 days old when the video is 15 hours old..
XPetu: I’m too lazy to review my VODs
Also XPetu: I’m going to create a web app from scratch to make VOD reviewing easier
For real! 😂 but I kind of get it. Watching a 20 min replay of a past mistake is painful, but spending 1600 hrs building a new thing is fun.
The human brain is funny.
For real god damn
Literally the curse of all engineers.
The true engineer way, spend 2 months to automate a 20 min task
@@THEJPIndustry
1. He going to use it not once but multiple times
2. He will let the other people to use it so there is more then 200 years of collective time saving
3. He will make mony out of it, so its not only to save his time
0:11 "But how bad is it really?"
*Vsauce music starts*
i non ironically anticipated the sound
Viisaus music
@@ancient5138 love that google correctly translates it to "wisdom music"
"I couldn't help but feel like I wasn't providing enough value for the money they were paying me."
Neace: 👀
Neace started coaching for 20 bucks a session. Supply and demand.
@@andypae true but the point still stands. Supply and demand doesn't justify that price morally for most people. If you are a decent human being you will still feel bad for the client
In neaces defense (his actual skills aside), his prices were never ridiculous for industry standards, only within the perception of non pros *playing a video game*. coaches for pro athlethes or industry outside of sports easily pays thousands per hour for similar percentiles. LS famously doesnt coach anymore because theres no way to make it worth the time for him without bankrupting the client
@@viktorgombos4975 Morality depends on the customer. If you don't want to pay Neace prices, there are plenty of other coaches who cost less. Neace hit a stride by offering easy-to-digest coaching for low level players that many elite coaches/players weren't able or even willing to offer. After all, Neace also uploaded his sessions for everyone to watch for free.
Many people think that streamers shouldn't be millionaires and that a high level professional athlete doesn't deserve to be paid +100 million to kick a ball around.
But if you've managed to make aforementioned things into a day job, anyone of us would take the bag because there's a very high chance that you won't make good income out of your passion for more than a few years.
Instead of complaining about Neace, the coaches should've considered what was the thing Neace succeeded in and make a counter offer to balance the market.
@@viktorgombos4975 morally? Acting like its a life necessity to buy league coaching from Neace wtf you mean morally 🤣 its literally logical, if too many people want coaching you raise the price so you can manage them and only coach people who really want to be coached by you for whatever reason.
Feeling bad for telling them your price and them paying it? Its like saying you should feel bad if someone has the newest iphone because it was too expensive when they can just buy a cheaper one.
Petu just even considering localized cost of living when looking to price your product is WELL MORE than what most companies do. You are achieving as close to a fair price as you can and that's all that matters.
Hell yeah Brother! Praise Jesus!
I mean its pretty much even with every option but argentinia costs 4500$ per month lol
@@tristan3068 That is equivalent to $4.87 in US dollars
@@tristan3068 well, in argentina we have more important issues for resolve than be better at lol, and Xpetu need to can eat from his work. Seems fare to me honestly
His finnish values are showing
if there are concerns around being unable to determine the utility of coachless from too short of a trial period, a potential alternative could be usage based pricing (eg a 10-game analysis package) instead of a time based subscription
I think that would be fair for all instances of the service since some of us play a few games per week while others play several hours daily and so on
We have already made some improvements to the interface and features since recording this video.
If you are interested take a look at the newest version: coachless.gg/
Best of luck on building the platform ❤
Besides opgg, beginner players need some type of assistance - like building the body, or the mind with rigorous study programmes
Maaan. I kinda want to do a UI design for the site. (I am full time UI designer)
So someone can buy VPN and get this for a lower price?
@@fession6448I guess? But then you’d be paying for two services just to cheat on Petu and the Coachless company, who actually care enough about the customers to make the service affordable everywhere.
In that case you would be responsible for setting an example for the industry that a company with a good intention isn't worth it.
What a sad world we are living in.
You should have made a video explaining the tool only playing as you explain feels ridiculous
Also the tool can't ever see everything it's limited by the replay history isn't?
As a simple example a decision to route the jungle in a different way and losing a camp won't show up
Ok you should probably have some kind of description or explanation of why the win probability changes within each game. When I watch your analysis of the game, it's far more sophisticated than what I, a gold toplaner, could deduce on their own. I get that you're creating a better way to replay review, so there will still be thinking involved on the part of the reviewer, however, I think an explanation of the sequence of events, and why the win probability changed, could be a huge boost to understanding how to improve based on the factors that won or lost the game. Maybe the technology is too limited, but seeing a high level breakdown of why the result of the game was what it was, to me, seems more helpful.
Although it might also be true that creating something that perfectly understands the game of league of legends might be impossible to conceive, let alone the technology to have it analyze every individual game and offer insights.
The only way i can think of to solve this, would be that you can send specific parts of the replay to a high elo player that works for coachless. Maybe this costs like a small extra fee, but a limited amount of mini reviews is included in the base service, so you only have to pay extra if you wanna do that often
@@jonkurp6463 What if the application can record your game and use AI to determine the reasoning. Something like Tyler1's AI. Just a though, maybe I'm missing smt important.
@@jonkurp6463 This would not be a small extra fee in my opinion. Yes, the high elo player would only be watching clips of 30 seconds or so but they'd have to provide a detailed analysis on a certain level to what some gold top lane main would think, "Oh this makes sense for this situation." while also keeping the explanation simple enough for anyone to understand. The explanation for each clip would probably take another 1-3 minutes times that by hundreds, thousands, tens of thousands+ of games and people, and the demand for 24/7 support, it would be very expensive to do this service.
A fully explainable model is the dream, I agree with you. But we must start with baby steps. Explainable AI/ML is a super difficult problem; the field is still developing. What the WP graph does well is it allows for quick navigation of a replay, allowing you to focus on the objectively important phases of the game. Another benefit is that you concretely see the effects of your actions, making it impossible (or at least harder) to blame your team, which inturn allows you to focus on improving! Thanks for the feedback.
@@xPetuStep by step creating the equivalent of a chess engine for league...
Stockfish? More like scuttlecrab...
Expanded item stats is the best thing I’ve ever heard.
It’s a pain in the ass to figure out how much damage ardent censer ACTUALLY contributed: you have to go into a replay, click to the end, see how much on-hit damage it added, then back calculate how many times it procced while factoring in average MR values. Ok that’s roughly the number of extra attacks they got from the AS. Then multiply that by an estimate of how much damage an auto attack does, averaged over the entire game. Brutal. I’ve always wished a program could do it for me.
Yes.
Yeah i'm most excited about the advanced items and stats especially as a support player. Not sure if this is really made for anyone outside of top from the way it looks and what he explained but the itemization stuff is just huge
I appreciate the localized pricing. I will try it soon.
Me to from my vpn in turkey
@@zwen6179 the difference is like an euro lol
@@zwen6179 cuz of people like you we are rarely getting localized pricing
@@mateuszkrytyk5711 cause it shouldnt be a thing
the rich area should pay more and the poor areas should pay less? like what.... it should be afordable for what its worth case closed end the business lecture
Bro im just walking right now and listening, i dont even need the gameplay its very calming to listen to you
He’s SO concise with what he’s saying-and while playing somehow?
Unless it’s a voiceover and he’s just talking about the game in the present tense? I can’t even tell.
@Fyrelangs that’s a relief lol. I know my brain is different, but I’d prefer not to think people can do 2 things at once, and do them both *so* much better than I can even when I fully focus on them. 😅
Programmers are lazy. We don’t want to review 1000 games. We want to make a program that will review the games for us.
The reason why I don’t review every game I play is because I don’t know what I do wrong most of the time, hence why I’m low elo. I would probably need to have someone high elo to review and explain to me what am I doing wrong and why is it wrong
Something like this could help immerse you in it while guiding you to moments to consider.
yeah unfortunately a fully explanatory system is probably not possible atm. but think of it this way: 1. you at least get a guide to the most significant moments you did at least something wrong so you can try to figure it out yourself and 2. a high elo friend doesnt have to watch through a 25 minute video to tell you what you did wrong, you can just send them the short clips of your mistakes you dont figure out yourself and they help you! I bet you get a lot of replies in several discords when you just share a 30 second gameplay clip asking for advice
Watch more videos about fundementals. They will help a lot with understanding why stuff that seems insignificant, truly does matter.
I was diamond and master peak and I could find many mistakes in every replay. Replays give you sooo much about the gameplay. A LOT of time you die cause of gank and often these ganks are obvious. With replays you think of yourself "man, I suck". It's really not hard. Usually when you lose a game it's because you die (too often). And it's easy to track why you die. It also depends on the role you're playing. If you are a jungler who goes 0-0 but enemy bot lane went 20-0, you may ask yourself when and how did they get such advantage and how you could change it.
In the end, the game outcome if the result of the gold difference. You just need to keep track of it when watching replay and ask yourself a question - why the enemies got more gold and how I could change it?
Btw, I very rarely actually watched my replays, but when I did, I felt like it really gave a lot of info. As xPetu said - I was lazy. I just wanted to play another game. I didn't even know tens of the new champs.
Hi Petu, thanks for everything you do for the community. There isn't a better example on YT of someone explaining their micro and nuances during gameplay which separates them from not just from elos other than chall, but other chally players too. You're an inspiration to me. God bless from Canada.
Very cool idea looking forward to see how far it goes!
It's a cool idea. The only thing you should keep in mind (in terms of the regional pricing you implimented) that people might VPN to a country to pay less for your Choachless service. I've seen mobile games that had regional prices and people used VPNs to buy the in game currency way cheaper than they were suppossed to. I think it's just fair to point out, if it's not something that got considered beforehand. Best of luck with your Choachless project :)
there are ways to prevent this - one can look up the ASNs of VPN services and blacklist them (or at least the most important ones) so that requests from these IPs are rejected by the server
Chokeless service )
I immediately looked at Baus and I'm seeing some Deaths that are +3 LP, Bausen's law validated with statistics?
This is what happens when a engineer plays too much league. Really a great way for every league player. Thanks petu
Hey Petu,
I think the idea of linking win probability to specific events in the game is intriguing. However, I don't see it as very useful for someone with a basic understanding of league. For example, it’s pretty obvious that dying is bad, getting kills is good, and securing objectives matters-this is common knowledge.
The system might actually lead to worse gameplay by focusing too much on these metrics. For instance, if you die and the system tells you that your death reduced the team's win probability by 10%, you might play overly cautious in the next game. But we both know that being too cautious often isn’t the best strategy.
Also, the system might point out that losing a last teamfight cost the game, but that’s obvious and doesn’t provide actionable advice. It just reiterates basic strategies like getting more kills and securing objectives, which doesn’t help beyond what we already know.
The only real value I see is using the system to analyze specific scenarios retrospectively, like determining whether trading one objective for another is worth it statistically. But even then, anyone with decent game knowledge might already know that securing a dragon for a soul point is generally more valuable than Nash.
An interesting addition could be to show hypothetical win percentages if you had won a certain duel. This way, you could assess whether it's a risk worth taking in future games. For instance, winning might increase your win probability by 5%, but losing could decrease it by 7%.
Overall, while the fundamental concept of the system is fascinating, it doesn’t seem to offer much new insight to anyone who already understands the game basics.
This is very well said, I was thinking the same thing.
I'd like to add another level of criticism to this tool. It only accounts for results-based thinking to evaluate whether the play was good or not.
For example: You're the jungler and you choose to trade off the top side of the map to gank bot lane and end up getting a kill. Lets say the enemy jungler chooses to do the opposite and ganks top side by giving up bot but unfortunately messes up and doesn't get a kill. The tool would tell you that your play was overall worth it and that you gained LP for making this decision, but that may not be the case. You could've never known that the play was wrong or right because it wasn't executed perfectly on both sides and therefore couldn't be a useful metric. If the exact same play happened, but instead, you were the one this time to miss the kill bot, and the enemy jungler got a kill top, the tool would tell you that the play was bad. The issue with this tool is that it doesn't tell you if your decision was right or wrong to make a play, it only tells you if the play did or didn't work out in your favor.
Like this comment above said before, even if you are moderately experienced in the game, you can easily make this distinction yourself and don't need the specific numbers behind it.
I think you are focusing too much on the win probability stat. I don't think the goal is to teach you to only make plays that increase win probability, it's using win probability changes to highlight key moments of the game for analysis and reflection (which might lead you to look earlier than that specific moment, like if a series of poor trades leads to a death in lane or is poor recall timings lead to the early loss of your tower). I think the hypothesis is that people get better when they review they gameplay, but many people don't because it takes too much time; highlighting key moments in the game will make VOD review a faster and simpler process, which makes it more likely to happen which will improve the player.
using win probability to sort of A-B test different decisions/outcomes is a very cool idea though
What you said is what petu said, you just wrote it all again XD, ofc metrics dont speak for themselves, the main idea was to get what a replay overview gives you, without watching the replay, if you look at the replay and you died, then realized oh i died... because i did this and this, this cost me the game or gave an enormous lead to the enemy team or lost us a crucial objective, whatsoever the reasoning behind might be, is the exact thing he wants to achieve, he wants people to know what to look for, that's what he said at the beginning, if you want to improve, its best to know where to improve at, because as obvious as all of those things might be, everyone still gets stuck on silver/gold dying because they ddnt think twice, this is going to be for people who WANT to improve, which would mean, after they see that a play cost them 7% wr overall, they're gonna focus on the play itself, not the % per se, i know you were talking with good in your mind, but i think you went a little over your head trying to prove something that wasnt asked to be proved, the reality is, you cant make a tool for everyone, otherwise we would all be fishers, constructors, doctors etcetc, this is going to be, for the poeple that its meant to be, not for the normie who knows dying is bad and still goes 0-10 on a game.
I feel this tool can lead to something great for sure, if it continue to develop! Step by step, it's a start let's say ;)
But indeed for now, it doesn't really provide anything for someone who have basics of winning condition. I won't go that far as other as said before, but i'll keep it even simple:
The tool (for now) show death / kill / objectifs. Then, the player has to do all the job and watch the replay at those timing and little earlier to see why this situtation. Ok it's great, completly agree! But well, if i open my replay inside league, there's a chronology inside it that already show me without watching the replay, when i die, when i kill, when objectifs where taken. So i can already go to those specific moments and watch only those.
That's where i find a issue with this tool, in his current state. But as i said, it's a start, a good one, but a start. It needs to provide more from my point of view to have a value, provide value to request value!
Still great job @xPetu (as an example, the second approach with gear optimisation i find it way more interesting and reasonable by only mathematics, cause it's literraly only math. Maybe should have / still should focus on this first, from my point of view.
@@mikecurley3070 What you're saying doesn't make sense. If the point is to highlight key moments to review, then the LoL replay and match history in client already do this. The timeline shows you who died, when they died, and how much net gold gained/lost. The only thing Coachless does different from tools Riot already has implemented is that it quantifies the net gold gain/loss into a win rate percentage or a +- LP. If we say that the numbers aren't what to focus on then what is? The only thing that sets it apart is the numbers.
I know Petu mentioned in a part of the video that the site is tailored to how you specifically can impact the game so it'll only show win rate percentage or +- LP on plays you could've impacted but this is no different from other tools. Outplayed not only captures the game, it records all of the crucial moments such as kills, deaths, objectives, and even ultimate usages. In it's current state, I don't think this tool is that effective at what it's trying to accomplish.
The UI looks slick! I particularly like the game listing, with the smooth transition into showing only the champ portraits on click!
Dude this is insanely cool! Excited to be an early adopter.
Questions first:
How much of the tech behind coachless did you guys "invent"? Like do you have patents on anything?
If coachless gets shutdown, would you consider selling a version that is local to ones computer and would require oneself to do the upkeep to have it functional? Asking because i find this too cool to be oost if it shuts down.
Congratulations on the public release. Having tried earlier version, this looks really good and i can't wait for the coming features. Personally totally agree on how you are doing subscriptions and why.
I am already getting ideas of how to put coachless to the test.
The video here itself is also really good just for the philosophy of how to improve and climb, and had points to take from.
1. We do not hold any patents currently and the technology is mostly the application of classic machine learning methods to, in my opinion obvious, problems.
2. We have considered a desktop version but I personally love browser apps because they are less invasive and easier to start using. A desktop application would require maintenance and upkeep just as much as a web application (if not more).
Thank you for your comment!
@@xPetu Hey, did you consider electron for a desktop version? At a base level, you can embed the website directly, and invest further resources on the desktop version as necessary as an incremental process.
@@mrocto329 I'd rather maintain the desktop app myself to not have another electron app installed on my system
@@mrocto329 Yes we have actually talked about this, the idea came internally from our main dev as well. Keeping it as an option for now!
@@xPetu If you go for the desktop route, then you easily get lumped up with other desktop apps that provide all sorts of in-game helpers like objective timers on minimap etc. and then there's more pressure to steer away from your main selling point of your application and to provide things that others have mastered already.
Subscription model is modern but it can feel daunting to pay multiple monthly fees for different apps at the time. One path that I haven't seen tried yet is building up a collab with some other tools and maybe provide a discount to subscribe to multiple apps at once. A sort of bundle price! For example I just tried out iTero Drafting Tool and loved the concept that Overwolf app and watching this video I immediately started thinking that how greatly these two tools could work side-by-side, one before the game and the other right after one. I get it that it might get tough business-wise to find a working model to sort of push each others products but you coming from content creation know that collabs with other creators usually tend to bring out the most exposure and growth, right?
Greetings dear Petu! I must say that a subscription model will immediately turn away many potential customers.
There is a MMO called Guildwars 2 that operates differently to all other MMOs in that it offers a one-time purchase for content rather than a monthly subscription. This is in fact a major selling point for the game because this allows players to stop playing and come back later, and doesn’t force them into playing the game when they don’t want to in order to get value out of their monthly subscription.
I fall into the category of not playing enough League to warrant a subscription, but I would probably pay a one-time larger sum up front if it was an option!
Agree. I am sick of subscriptions full stop.
But I have paid a lot more for services that I get to keep, because i know that i can come back and use them whenever I need them.
Same. Great idea but subscriptions are not for me
I laughed at this because I feel the same way about subscriptions and I'm also a gw2 player, funny that you used it as an example.
Understandable, i'm actually in the same boat but i think he SHOULD do a subscription-based model, OR go free with ads. I suspect he thinks adblockers will reduce a lot of the income and will let him run the site and services but not make enough of a living from it, whereas a subscription-based model allows for better consintent income and then some, to implement new features
I think there target audience and demand for the service, at least as it is now, is very small. I wouldn't mind paying a one-time $50 fee for something like this with the advanced item features later but i still think it wouldn't be a good financial decision for Petu to make it a one-time service fee, unfortunately
I think at this point the potential reach he has with the product is very small. Think how many people would for example spend a one time 15€ fee for this product. Then multiply the price by how many people would pay that price and you will quickly notice that there is very finite money to be made. I think the subscription based model works much better because even it it halves the potential customers they would be a consistent stream of income.
I've recently been looking into what it would take to create a tool that calculates the mathematically perfect builds for each champion, adjusted for allied and enemy team comp, so i think its super cool that youre already on this path!
you can really feel this is an xpetu project
One problem I see straight away is that you can never really tell from data that someone specifically caused the bad play. Sometimes you might not even be involved and that is the reason your team lost that fight since you left them there etc. Sometimes you did good by not involving and running away. Data will never tell you that. There is so many gaps that you just cant fill I hardly think this will be relevant.
Petu, shut up and take my money, I trust you.
You and everyone else involved are AMAZING. Petu, you don't know how much the localized price means to me, and I'm sure many others from Brasil and other 3rd world countries. Having this kind of visibility just made me cry. THANK YOU
Yes!! So many things are so expensive for us living in third world countries so it’s usually only people from first world countries who can easily afford non-essential products, or those with higher income in third world countries who have more disposable income 🥲 Glad to see considerations like this for his viewers who are potentially interested
Make sure that region based pricing isn't abusable (or very hard to). As someone who knows his way around those things, I can tell you people will use it to get a better deal for themselves. (or it be offered through third party services)
thats a pretty cool idea. And it's cool how you are trying to make a product out of it and make an earning. I can see that the feature that sets your product apart from others is the win probability graph. However from what I have seen in the video I have some doubts about the effectiveness of that feature. Your service seems to highlight events that Impacted the win probability of the game. It can detect when you died or if an objective was lost and shows you how much lp it looses you on average. But I think everybody knows that dying or loosing an objective is bad and will reduce the odds of winning that game. The reason people get coached by a person is not so the coach can tell them not to die and not to loose an objective. Instead the coach tells you things you did wrong that is not immediately obvious (should you have roamed?, should you have recalled or go for plates?). It seems your program would be good at pointing out the things that were bad, but it would be bad at pointing out the things that you DIDN'T do that would have been good.
Lets assume you kill your laner and have to decide whether to recall or go for the turret. If in your game you successfully destroyed the turret, the program can then tell you how that affected your win probability. But it can not tell you how going back would have affected your win probability. In some cases you can do something and gain 2lp while having done something different could have gotten you 4lp. I hope you can come up with something to combat this issue since I think it could have potential.
I hope you see this even after the video has been up for a month and I hope you can explain your future plans concerning the issue.
on the vod review topic, i didn't think so little lol players reviewed their games. i'm an overwatch top 500 players and it is really common for players to do it in high elo so i would have thought it would be the same in league. I think coachless looks amazing, i wish we had that in overwatch, keep going for the sake of the community.
So you're one of the last 500 players still playing OW?
are you watching every of your game though?
Bro, the thing about fair pricing is just such a good thing to do. I really love you and coachless for that!
Need a “HOW TO TAKE/APPROACH THIS TRADE BETTER OR NOT AT ALL” feature.
As a fellow mathematics enthusiast ive been adoring your work for a long time. I started to play league more regularly maybe like 9 months ago. I was already a top main but struggled to find a champ that worked for me. I did some research and were fortunate enough to find ur channel. I love that ur incorperating so much math and optimising into your builds AND plays. Im going into my final year of highschool and going to graduate in something along the lines of a science program. I would love to pay for your subscription and i couldnt agree more to what you said about "using/caring more" about a product youve paid for vs a free one. Greetings from Sweden
Ok question. At roughly 9:30 in the video, you very narrowly escape death. I believe this moment would provide tons of learning value in a vod review, but would not be captured through win probably. Do you have statistical models for finding moments like this?
I'm not sure if I understand you correctly but do you mean when you're reviewing someone elses game? Because if you review your own demo, watching your own narrow escape doesn't really improve you, and also it basically shows in your win probability by staying the same, and not dropping since you didn't die.
@@juspe123 that's what he means - the app will treat it as nothing happened, but in reality he was very close to death and risked potential loses. But I think such cases would be really hard to keep track on. Maybe the app could follow flash usages and give some notes just to check why you used it. Maybe it could also keep track of your HP and see why it went very low at certain moments.
8:59 this guy... holy 2014 vibes I love you petu
Welcome back Petu you were missed King 👑
Entrepreneur of the year.
I don't play league so won't use it personally, but I love the approach and the attention to details.
This really deserves to grow big.
for me the only way to watch replays is when i know what i am looking for(i have some sort of excercise) i am a jungler so for me it can be jg tracking i hop into a game(plus POV recording) and i track enemy jungler, while also watching what i was looking at in game when i missed something(like jungler stepping on a ward). Then i hop into late game and try to track myself from enemy perspective. It is only thing that worked for me
Congratulations on your first step on releasing Coachless. really looking forward to this journey of development and i wish you all the best!
Petu, you are a gigachad of the highest order.
Glory.. beyond anything that we have ever witnessed
If this tool works the way I hope it does, it's certainly promising, and I can see a number of uses for it. For example, measuring individual performance, potentially get to use that measure in order to implement fairer LP gains by individual performance, or to implement a functionnal toxicity detection system based on individual performance, to flag excessively poor performances, and exclude good, acceptable performances that shouldn't raise any such questions (like theBausffs playstyle). Etc.
But it does also raise a lot of questions, to determine weather it's really that good or not :
How is the win probability determined ? Is it through statistics taken from the API ? Or is it by reusing that AWS powered tool ? Whatever tools are used or not, which stats are taken account for, to calculate that win probability ? Is it win probability by gold generation at a given game time ? Or through XP generation ? Damage dealt ? Winrate by objectives ? Winrate by KDA ? Etc. Or some combination of those ? And in what conditions are these parameters taken ? At what time in the game ? With which teamcomps ? Against which matchup ? At what elo ? With which objectives taken ? Etc.
And, if we increase the number of critera, how do you get the statistics for one exact composition against another, at a given time in the game, with a given gold/xp/stats lead repartition across teams etc ? The sample size of existing games decreases when you increase the number of criterias to be met, and if you want to account for a large number of criterias in order to provide a more precise prediction, the sample size becomes extrememy small, and hence, the result less precise. If we take the example of the situation at 23:00 , in order to calculate the impact of the baron vs drake trade : What's the number of games with a shen, ekko, anivia, kaisa, janna in team1 VS poppy, xin, sylas, jinx, nami in team2, in master elo, where team2 steals baron and team1 trades it with a drake at around min 29 into the game, with each team their specific gold/item/stats/xp/objective leads, and each player their own specific individual lead at that specific time in the game ? Probably just that one game. Hence, you have to account for these criterias individually, and smartly weigh one against the other to determine how much each matters and contributes to the observed result. It's an incredibly complex problem to wrap your head around, and i'm very curious to know what kind of approach xPetu took to get there.
I did some personal research and statistical/probabilistic calculations on the matter, trying to think of some models, and some AI/algorithm to gradually improve them, but couldn't really find the time to lay the work down in order to test and correct those. I'm really interested in seeing what kind of model xPetu came up with.
🐐
xpetu absolutely mentally destroying darius and pantheon's mental while advertising his mad genius scheme is a mood
but this guy is cooking for sure
You are a scholar and a legend. and this is inclusive of both. I don't even play league anymore but I still watch your videos, hope all the best for your career!!! I'm also about to finish my thesis, congrats, it's quite the slog and I know how much effort it takes to make it that far. Well done on several points :)
peak as always
Hell yeah, brotha! Wish you all the success in life and your new project! Much love from Ohio, friend!
"There are... I think 15 different currency prices"
Huh, I wonder if the Mexican peso is included in those.
"Let's say if you're buying from Mexico"
MEXICO MENTIONED LETS GOOOOOOOOOOOOOOOO
13:00 This cost matching system is just so good. Well done
Outstanding!
i seriously love how much you try to link the things you love! you're a great example in so many ways, absolutely my favorite league streamer. idk how you maintain positivity while being in this community (well actually i do, you full mute LOL), but we're all thankful for you!
I get that Coachless is for a very specifc group of people but i feel like you could offer 'early bird' sort of incentives. I should also note that i am not in marketing but just wanted to add a few thoughts even if it's needed to take them with some grains of salt, so to speak.
Since Coachless is so new and will be improved in the future, why not let everyone know that. Just be transparent and have a lower subscription cost since it is still in early stages. This will also increase it's use as well as give a benefit to players willing to jump on board early. Then perhaps you can notify everyone when the price will increase soon and/or allow the players who already have a subscription to keep the price they are paying for an additional month, or whatever is fair, longer than new subscribers. I believe it would help Coachless grow, encourage people to jump on board, as well as showing a clear, honest, and transparent business model.
Furthermore, it would cool if a system was added that allowed you to simply have Coachless without paying if you had paid enough and/or subscribed for long enough. Or if that is asking too much have a reward system with Coachless where dedicated subscribers get a free month every now and then or something beneficial like that and as a thank you for using the service consistently.
first one might be worth looking into for an actual business, second one is just wishful thinking
@@srensen3081 Thanks for your input.
As much as making it a subscription will turn away potential users (the whole supply/demand model when pricing is involved) I actually like making it a paid product, because already I am weighing the factor that if I pay for this review service, i'm going to make it worth my money ...... and invest the time to learn from it. Which makes my hypothesis that your user base will have a higher percentage of users that improved vs users that didn't improve (success vs failure of product). A little bit of success bias, but in this case, I think having users pay is fantastic incentive for improvement.
Also, what a legend. Keep it up Petu!
Let him cook
Hi xPetu, I think I have a good solution for your program to further analyze objective control and more specifically, it's use. (Mentioned at: 24:00)
Baron: This one is easier to get a read on than dragons. Baron buff lasts 3 minutes. You can have the program check the timestamp from when Baron was taken and when a tower or inhibitor was taken within that period and associate it with the buff as a part of the play based on the time it took to take the obj whilst checking if the buff was active (and slightly after) because a push with Baron tends to go past the time the buff is held.
Herald: The easiest to check is to see whether or not Rift herald was used on the tower when it was taken.
Grubs: [apply a burn for 4 seconds that deals (Melee role 8 − 48 / Ranged role 4 − 24) (based on stacks) true damage every 0.5 seconds, for a total of (Melee role 64 − 384 / Ranged role 32 − 192) (based on stacks).] You can take the total damage to towers for each champion and subtract a percentage of damage dealt each game to create an average for the grubs. A small cop out as it won't give as much detail to how much they impact the game like the rest of the objectives, but it will be a start to researching their total value for a team.
Dragons: These will be the most difficult to analyze as you'd have to factor in each type, which dragon number it is, and what objectives may have been traded. I could give a very detailed explanation on how I think it may be a good way to take on this challenge if you're interested.
To avoid adding too much to this comment I will stop here. Feel free to reach out to me if you'd like to talk it over. I love using numbers and formulas to analyze and solve things. Also, it's a great video!
o
great i hope you will succeed with your project, i always like how i hear the enthusiasm for creating new things in your voice 😂🔥
HE'S BACK, HE'S FINALLY BACK!! Missed your content mate, great to have you back😌🤌
Well petu, i dont really play anymore, but i guess will go again with shen only and your coachless. You are awesome, keep up the good work. Greetings from germany
Man i stumbled upon you channel while trying to sleep. You have such a soothing voice. Now i stay for the quality content
I really like how your commentary is during an actual game you're playing. The little interjections about stuff you're doing in-game makes it very entertaining.
I love how you can just seamlessly jump from explaining how your program works to giving game tips based on whats happening in the background (e.g. reverse flash at around 9:20)
Dude, I'm from Argentina and I was so sad until you said that prices are localized, that's the only way I might be able to buy it. Btw I've been watching for years (I stopped because I stopped playing league a few years, but anyways) It looks great, I'm so happy
I like the idea! Feature suggestion down the line: It would be great to be able to forecast mistakes. Example: Making the decision to push at 4m30 resulted in your vulnerability to the gank where you died. So coachless would highlight the "pre-crime" of pushing at that time.
Potentially the same thing could be done with purchase choice. Example: because you bought doran's blade instead of doran's shield, you were forced to recall instead of the counterfactual situation of having 200 extra hp from healing. (but i bet that would be tricky to get data for)
Your excitement for this project really excites me to try this project too. I feel the same laziness to improve and always shift to watch better player videos because it's more entertaining to me but this project looks extremely interesting.
By the way, as a jungler I am interested to know if this project can notice Junglers Pathing mistake or what could have been better.
In my head, this is Dracula giving me League advice. Now on actual suggestions, I hope there is infrastructure within the coachless service that collects the data from the players. There are some interesting correlations that can be made when you have multiple angles of certain phenomena and being able to then find statistical important parameters with certain outcomes. Not only that but the win-probability calculation itself can always be optimized by the wealth of new data. Also give champion-specific outlines. Given their power-spikes and win-conditions. I always love to see people with passion
9:02 amazing flash, wouldn't have thought of doing that in my life
dear xPetu i am Rokas a lithuanian league player who makes mathamatical league calculations and was inspired by u. i have to say i do rewatch most of my games for 2 or 3 years and it been very educating. i cant even go into a game without geting ready for it, for about 30 to 40 min crafting everything im gona do. thank u deeply for ur work to the league comunity, the game has never been as eficiant as it is today thanks to u and many other great players.
sorry for the messy english.
Good luck with coachless! I hope it turns out well. I checked it out briefly after finishing the video. I noticed that the web app is not perticularly responsive on mobile. Not sure if that is a priority for you, but something to think about. I would personally probably use it quite a bit on the phone if I had the option.
bro keep the content wenderfull i hope i will get better with shenn like u
Very exciting application xPetu! I'm sure you're thinking along these lines, but if you define your optimization objective as maximizing win probability, when you hit data sufficiency you can build a classifier to classify events, and then cluster by the game stage they occur in (early, mid, late). These could then be provided to users as an auto-generated overview of the mistakes and positive plays that had the largest effect on the game outcome, stratified by game stage.
This does look good. I'm also happy to see this isn't a 10-15/month thing. It seems very affordable and valuable. Bravo man.
17:23 lmao nice concept. aatrox vs akali is black cleaver, mhm yeah, tabis next, steraks (actually good but lowest wr in concept), serylda is an option to build with the cleaver now, nice.
im actually here not only to joke around, i know youre a honest man, and i just want to wish you luck with your project. i wanted to start reviewing my own replays for a long time, but exactly, i was too lazy. so i see a lot of hope in your project, gl really
I only watch streaming and don’t play, but I loved your project, I’m fascinated by statistics and math and I can totally understand the pleasure you feel doing this! You’ll make it great!! I’ll keep following your channel, all the best!!
Incredible project!!! The League companion app scene has become stagnant. To see an engineer / mathematician (not sure which) apply problem-solving in this way is extremely inspiring as a software engineer. KEEP GOING !!!
16:20 man actually huge brain on the 3 month subscription, and i'm positive that you should never make it for a month only, there are a lot of products that work like that exactly for the reasons your basing your argument, the 1 month could propably make bad reputation because people want a low effort/high value solution, and might say bad things about a good product because they didn't use it appropriately
I quit League a few years ago so coachless isn't really a thing I need, but I've been watching your videos pretty much without fail all the same. Wishing you all the best in turning your research into a business!
I wish you good luck Shensei xPetu, too bad im from one beautiful country and can't support you, but i can write a couple of nice words, keep it up!
awesome idea ^^ i hope you make it ill look forward for it!!
I just love you for videos like this Petu.
Also love listening to your voice it's so calming. If you'd be doing podcasts I'd spam that sht continuously.
This is also a FANTASTIC way of dealing with tilt. You can see if the game was actually yourself or out of your control and then LET IT GO and move on
i really liked how intelligently you talked about this, its clear you have a vision and a winning idea. I look forward to using this service
I'd be very interested to see how accurate the winrate prediction model is, if a team gets nash they will likely get some turrets and inhibs as well and an accurate model would account for that. I would love to see how this developes!
Also, if two completely distinct plays happen simultaneously, for example me getting solokilled top lane and the enemy adc getting a double kill, from what you have said about the software, it would categorize that as a negative lp, as the winrate does drop. So maybe there is a way to consider how the model judges expected winrate, if you don't tell it about bot lane
And one way the model can become really inaccurate is if certain factors are just not considered, like summoner spells or even wave state.
This is really awesome! I’m looking forward to seeing how this progresses, but I want to give my thoughts on why I don’t think this is worth atm for me since you seem pretty invested in improving coachless:
As a support player, I feel like a big portion of your influence on the gamestate won’t even be tracked (if my understanding of how it works is correct). Vision control is an important part of the support role that can play a massive hand in protecting lanes, whether or not we get an objective, etc that can’t necessarily be quantified by seeing whether or not we have kp or hit the objective. Similarly, if the team secures drag because the support zoned off the jg then that wouldn’t be reflected if the metric is based on us doing damage to the objective (like in game objective assists).
Similarly, if you play too passively then coachless may fail to catch that some players are making mistakes by *not* being involved in fights they should have been. So many mistakes in league are based on the missed opportunities that lead into certain plays being bad. A thoughtful player who is genuinely trying to improve will hopefully be aware of this, but to more casual users their benefit from the app will be negatively impacted by how they approach the information it gives.
I’m also not entirely sold on the win% timeline metric’s usefulness. It’s great as a tool for identifying moments to look at, but I think it leaves a little to be desired for people who really want to improve every aspect of their play. Of course, quantifying every improvable aspect in league is a difficult task and it would be ridiculous to expect that on release. I’m very excited to see if/how you go about tackling that
I wanted to say "all this does is declutter the in game replay view at the expense of not actually showing you the replay" but specifically omitting events (deaths, objectives) that you are not a party does satisfy your objective of "make VOD reviewing easier"
I don't think I need to tell you that the obvious next step is to have some integration with the actual replay itself. You either need a client side executable or you need the server to pull parse and display replay highlights with the massive overhead that entails.
This is very game changing, I am not playing League anymore these days. But I'm still playing MLBB which is a phone MOBA, and even tho I pride myself being Mythical Glory every season (Which is like Master in League), some days my mind just went autopilot mode. I'm wondering if there are some ways to implement this to MLBB.
Good video as always, Petu
Looks like an incredibly cool project, I haven't touched the game in a while so it's not for me, but looks like you're cooking up some really interesting stuff. Excited to see where it goes and learn more about the various features your working on, they sound really interesting!
It would be nice to know to what extent indirect mistakes are also integrated into the programme. Here I mean, for example, warding as a supporter or making decisions as a jungler such as taking objectives or rather taking the teamfight.
If there was a feature looking at the graph and simultaneously getting the vod/replay for that time i would consider subscribing
Looks great, think it might be a good idea to be able to purchase a certain amount of months at a time rather than (or as another option to) a three month rolling subscription. It might help give a kind of goal or deadline to commit to
My man. I've been big fan of you for around 4 years now (I think). The very first video I saw of you was your Legendary Shen Guide with Shending Help.
I've been enjoying your content since then and I'm so happy to see that you may be able to make money out of your passion.
I was here watching when you strugggled for content 2/3 years ago
I was here watching when you had your difficult run to Master
I was here watching when you had your lession in your wrist (if I'm not mistaken)
And I can't wait to keep watching while you absolutly crush it with this new project of you.
Don't give up, keep pushing, take care of your health and drink water.
I'll keep watching... From the shadows.
i would love it if coachless can track ultimate too, for example if you are playing when and the enemy took dragon at minute 12, but you have ult and the enemy is contesting then it could be your responsibility to help out, of course it would take minion wave, enemy took tower, or having teleport etc...
I've subscribed to your service, Petu!
Thanks to your videos I've finally broken out of EUW Emerald and got to Diamond 3. I have a 73% win ratio with Shen and got to rank 36 on the EUW Porofessor page. Thank you!
i dont play league anymore. but i still like to watch so i made an acc even tho i wont be using it to show my support. Gl Petu !
This might be a bad take, but I don't understand how this is useful. It just shows win probability going up or down based on winning team fights. All the spikes are around objectives. It kind of just boils down to winning a team fight around an objective is good. Which... yeah it is. And conversely dying and missing an objective is bad, which... yeah it is. Even in the demo you just looked at objectives taken and your deaths which are already highlighted in the standard vod tool. Now the difference in win probability over time based on that baron fight was pretty cool though. But I don't understand why baron is valued so low either. Baron is usually worth a couple of towers and inhib, but the win probability didn't really change until that happened. Or maybe it was just because everyone died. Fun vid, excited to see how your tool progresses.
No I totally agree. The win probability is so misleading. Steal a baron and your win probability goes up to 70% for 10 seconds before your team gets crushed again.
It's like you said: Every game has a pivotal moment where the momentum is secured, and victory is essentially locked in. That is maybe worth reviewing if you can find it in this app.
Mostly it seems like a waste of time; but I could be wrong.
I'm sure you could encapsulate structures taken during the baron timer and the more you lose during it you could increase the impact that losing the baron had on it.
Not sure if you're able to alter that at all, or if you're relying solely on AWS.
Just something to increase the weight of the baron take based on later happenings. It would still not be perfect but it would definitely be a step in the right direction, and might tell you that losing that baron was a bigger mistake than taking dragon was worth.
Really cool app though! Nice job Petu :)
That part at 8:20 where you went for the assist, got 60~gold and lost 160gold worth of minions under the turret was funny
Adding stuff like Friend plans where you can buy a plan for multiple league accounts sounds so nice would definetly purchase