This is called an oversimplification fallacy. They wanted to match the skill of pro players through the learning capability of the AI. You left out the most important part of their goal as you blast someone else for doing what you are doing. If the AI is winning games on other merits, then it undercuts their goal.
Again, win-rates vary. A win-rate is a rate of winning over a large sample of games. That’s the definition of a win-rate. You can’t have a win-rate over a small number of games. For the AI to match the skill of pro players, it would need a 50% win-rate over a large number of games vs pro players. That criteria isn’t met.
Yes, which is why a few bots getting to 6k mmr isn’t impressive. The game is very complex, and a few hundred games is nowhere near enough to test the bot in a sufficient spectrum of scenarios.
I’ve also beaten serral. So have many people - he only has an ~70% win-rate at the pro level (most of which are from low-level peons that he 3-0’s at the start of a tournament). If you want to claim that you are as good as Serral, you need a 70% win-rate vs the same opponents that Serral plays against, and a 50% win-rate vs serral himself. Again, a few games doesn’t prove that. You don’t seem to understand how the massive variability in such a complex and volatile system REQUIRES a very large sample to average out that variability. The bot in one game wrecks serral and in the next it doesn’t make observers and dies to a <5.3k mmr Zerg player who went lurkers. Things like this cast serious doubt on whether their sample was large enough to support their claims.
So like, if you read the paper like I’ve been telling you to do this, they come up with a confidence interval for each account where they have an estimate for the true MMR.
But its obvious that either didn’t read the paper or you don’t know enough statistics to interpret the simplest parts of the paper.
Its obvious you are grasping at straws now, you could easily take 1 minute to look at their paper and see for yourself, instead of making crap arguments and then googling until you can find another crappy argument whenever you get proven wrong.
It must be pretty embarrassing to spend hours making an argument only to realize that they already destroyed your argument in the publication. lol.
In other words, you took the existence of a confidence interval as proof that the confidence interval is right without applying even an ounce of skepticism to how the confidence interval was calculated. These are the people who claimed it “mastered” sc2 after it proxied vs TLO’s offrace and are now claiming it is 6k mmr when there are videos of it losing vs 5.3k players. Skepticism is a healthy part of the scientific process, especially when the people involved have a history of inaccuracy.
As I have stated repeatedly, it would be difficult for them to claim a high confidence from such a small number of games. They likely didn’t account for the high variability of opponents, playstyles, and the shifting meta.
The paper is actually quite rigorous and they have already addressed all of your arguments, but I get the feeling that you aren’t reading the paper because its too high level for you.
I am not arguing against the paper, I am arguing against you. If you had a counter-argument, you’d post it. You don’t because you don’t have one so you resort to pretending the paper vindicates you while you refuse to state why. The paper is evidence and whether it supports your position is a fact under contention. Post your argument or admit that you don’t have one and type GG.
As I have stated repeatedly, it would be difficult for them to claim a high confidence from such a small number of games. They likely didn’t account for the high variability of opponents, playstyles, and the shifting meta.
it would be difficult for them
them
their
their
Directly addressed in the paper.
Literally addressed in the paper
In the paper
In the paper
their claim that it is GM level is not substantiated
I’m just using your own words against you, its actually quite funny how you try to pretend you didn’t say things even when im linking you direct quotes.
You’ve done nothing of the sort. We are here to discuss the findings within the paper but you refuse to do so. Saying “it’s in the paper” misses the entire point of a conversation about the paper. Is having a conversation really this much of a challenge to you?
The paper can’t talk. I am not talking to a paper. You are a human who thinks the paper concludes something, but refuses to argue why that is the case.
This is getting boring its clear that you don’t understand machine learning concepts enough to read the paper at all. I suggest you take some more classes before you embarrass yourself again.
A bad strategy losses you the game, period, and it doesn’t matter how good you are at the game. You can have perfect spending, but making defensive zealots from 1 gateway will not win you the game. Mechanics are how well you can execute a strategy and push the strategy to its max potential, but if the strategies max potential is less than that of another strategy it will lose.
Remember, you are the one who outright called the central limit theorem wrong. Take your own advice. You have better things to do with your life than troll a thread about a topic beyond your understanding. Just because you don’t understand it doesn’t give you the right to derail the conversation. Take your insecurities elsewhere. Nobody fully understands the topic which is why there is a need for a discussion about it, and if your insecurities prevent you from contributing to the discussion then move along.