AI in League of Legends – YamatoCannon League of Legends AMA29

by



RECORDED BEFORE SEMI FINALS

Timestamps:
00:00 Ladies and gentlemen
01:04 Thoughts on using AI’s as scrim partners like in chess?
04:47 Should teams invest in developing their own sandbox tools?
06:06 Why do you think G2 fans always say that G2 are trolling when they lose instead of accepting that the other team was better?
07:51 Thoughts on 30s timeout option for the coach during games to talk with the players?
09:23 Who’s your dark horse to take it all and why? (recorded before Semis)
10:13 What will be the impact of OPL players not taking import spots in the LCS?
12:25 What do you think will be the most impactful changes in the offseason?
14:55 What’s your favourite language?
15:13 Has Jiizuke ever cooked something back in the Vitality days? If so, how was it?
15:43 How did it feel to work in Korea and what was your reason for coming back to Europe?
18:09 Excluding League, Esports, etc. what is something that you want to accomplish in the near future?

#AMA #LeagueofLegends #Worlds2020

SPOTIFY for all podcasts and Lore readings: https://open.spotify.com/show/4NB268pCCGPFltqPBFFoOR?si=T16w0D2xRSO_km1Z3Z3wDQ

JOIN NOW: https://old.reddit.com/r/yamatocannon

FOLLOW NOW: https://twitter.com/YamatoMebdi

SLAP NOW: http://twitch.tv/YamatoCannon

I HATE FACEBOOK: http://facebook.com/YamatoCannonLoL

DISCORD: https://discord.gg/SeAGVXf

I’m part of the official League Partner Program with Riot Games. You can read about the program here: https://partners.leagueoflegends.com/en_GB/

AMAZON AFFILIATE LINKS:

AI in League of Legends – YamatoCannon League of Legends AMA29
AI in League of Legends – YamatoCannon League of Legends AMA29
AI in League of Legends – YamatoCannon League of Legends AMA29

Source

44 thoughts on “AI in League of Legends – YamatoCannon League of Legends AMA29”

  7. Ironically, I feel like whenever calls Suning now as the underdogs or the dark horses to win it's just not the case anymore lol. Because people kept calling them underdogs, it seemed like almost everyone wants Suning to win, and also legitimately can and will win, and with that it seems that Damwon is now the underdogs going into the final lol.

    Edit: I didn't expect to hear Davie504 from this channel lol

    Reply

  13. I hope maybe riot will do like skins for the items: if a player want to play with what he knows, let him do so he can see all the items as the old design (even enemy). If one wants to play with new design he will play, buy and view players items as the new design.

    Reply

  17. chess GM does not scrim against AI. Magnus & numerous other GMs all have said this in their interviews. The assumption of first question is simply wrong.

    Reply

  21. The timeout is a good idea but can only be taken after you have been aced. Therefore you cant just pause momentum, or if you have a 5v5 team fight and you lose it 3-1, you cant call a timeout to talk about it, but if you get aced, call the timeout and have the coach make a plan coming out of the timeout

    Reply

  25. I'm always interested in hearing peoples' thoughts on AI, because it seems like a lot of folks have misconceptions. I'm not an expert but I do really like learning about AI.

    Your first caveat about patches and things changing is valid, but it's a much smaller issue than you might think.

    The idea of testing matchups is excellent, and probably the easiest and most valuable use of a capable AI. If AI Ornn has a 75% winrate against AI Renekton (and we can verify that the AI is playing both champions at the highest possible level), there's obviously something we need to learn about playing that matchup where I assume it's closer to 50% but favoring Renekton. It tells us that we're not using Ornn' kit correctly (and probably that the game needs to be balanced once our skill catches up).

    There are lots of models for AI that are resilient to change, like patches. It does invalidate old data, but it doesn't necessarily make old data or training useless. But, the strength of AI is that it can try and learn from trying more or less infinitely in a very small amount of time. A player can't possibly practice more than 50 games in a day (I'm taking for granted a game takes 30 minutes), and the feedback and learning they get from it is filtered through perspective and bias. A computer has naked input and output data, and its only bias is how successful the relationship between those two things is compared to the desired result.

    Conservatively, if a computer could simulate 500 games a day and learned at 10% the rate a human does from a game, it would still be very easy for it to surpass humans very quickly.

    For most tasks, there's no real question of "if" AI will be better than humans, it's more of a "when." There are exceptions for thing like art and abstract tasks, but for something as simple as a contest with a clear win or lose… The computers are gonna win fam.

    Reply

  28. Fans say G2 is trolling when they lose because it just look like int. They flash forward, go in solo in 5 people. Fighting instead of running etc. It is just their playstyle. They just try things and try to do something from nothing. When it fails, they are usually in a position that they have nothing to do and they just lose objectives and then games.

    Reply

  37. Just a small note. Chess AI's dont know every possible move. Computing every possible chess move would take trillion of years. They learn more or less like we do. just way faster.

    Reply

  42. Data Scientist here. What I would like to see is AI assisted training programs to come out. Tell players what to buy, when to back, where to path, where to ward depending on the game state to maximize their chance of winning. Unfortunately, without an API similar to DOTA, that won't happen anytime soon. With an API, it'd be pretty easy to build something like that out.

    Reply

Leave a Comment