Alright, I’ve let you make a fool of yourself for long enough. Time to quote the paper:
AlphaStar Supervised and AlphaStar Mid were evaluated starting from an unranked rating on Battle. net for 30 and 60 games, respectively, for each race; AlphaStar Final was evaluated from AlphaStar Mid’s rating for an additional 30 games for each race. The Battle. net matchmaking procedure selected maps and opponents. Matches were played under blind conditions: AlphaStar was not provided with the opponent’s identity, and played under an anonymous account. These conditions were selected to estimate AlphaStar’s strength under approximately stationary conditions, but do not directly measure its susceptibility to exploitation under repeated play.
What I’ve said is consistent with the paper, what you’ve said contradicts it. Time to stop trolling kiddo. You refused to quote it because you clearly haven’t read the paper despite making this big fuss.
I literally quoted the paper in my first few posts. I read enough to look at their sampling method, I read enough to see their estimates for true MMR.
I literally quoted most of the stuff you just linked before mate…
I think you’ve proven yourself incapable of reading this paper well enough if you can’t even figure out the sections where they talk about MMR.
Rabidrone: their sample isn’t large enough to account for repeated play, meta shifts and other mmr fluctuations.
Cheezecake: no it doesn’t, just read the paper
Rabidrone: Ok, so show me where in the paper it contradicts this.
Cheezecake: lmao you clearly don’t understand machine learning enough to read the paper
Rabidrone: Ok, here is the spot in the paper that agrees with my point.
Cheezecake: “I literally quoted the paper” and “you’ve proven yourself incapable of reading this paper”.
Rabidrone: NO GG NO RE.
I don’t understand how you can lie like that when we could just scroll up and see that its not true.
I see Batz is practicing his online debating skills as usual. Couldn’t think of a better skill to level up.
What problems do you have in real life that led you to become such a impulsive liar?
It’s all public record kiddo. You have dozens of posts denying the fact that their sample is too small to allow for things like repeated play, yet here the paper is agreeing that their data doesn’t measure repeated play. You did this while repeatedly harassing me for supposedly never having read the paper and for supposedly having a poor understanding of machine learning/statistics. You’re a clown and anyone who reads the thread will see it.
Like I said, I don’t mind the discussion about the topic but repeatedly attacking me on a personal level in an extremely hypocritical fashion is below the belt and I will not tolerate it.
I’m literally going to quote the very first statement you made on sampling.
You didn’t say “the sample is not large enough”, you literally said it didn’t have the right composition. Your statement cannot be misinterpreted at all.
You are far too kind. This isn’t a debate, it’s a beat down, but that’s always how it’s been with this Shonix guy since the first time he began posting. He couldn’t win arguments so he pretended to be a professor of statistics. It was frankly pathetic.
I cannot wait to see what type of mental gymnastics you are going to use to try to debate your literal own words
Go get nomufftotuff to save you. At least he didn’t completely and utterly butcher basic statistical principles.
You didn’t say “The sample was too small because it was only 100 games”, you literally said "it didn’t have enough games against lurkers "
The REEEEEEEEEEEEEEEEE is strong with this one. He’s quoting the posts that prove him wrong as if they prove him right.
Found another one:
https://i.imgur.com/9NzElnn.png
And another one:
https://i.imgur.com/MZc7ICC.png
And what about AS beating serral 4:1 at blizzcon?
Btw.: Maybe you should read the scientific paper again. There is no “the” Protoss AS, there is no “the” Terran AS and no “the” Zerg AS. They have a lot of different agents playing on lader.