03 Aug 06
3 edits
What Elo would a lifeform have if it played perfectly? For example, if it managed to score 3/4 against Kasparov when he was 2851 it would have an Elo of roughly 3050. If it scored 15/16 its Elo would be roughly 3250. In other words, how difficult would it be for Kasparov (or any other top GM/bot) to scrape a draw with White? (Of course, this assumes that there's no forced win for either side from the initial position.)
03 Aug 06
Originally posted by ThudanBlunderIf depends how close kasparov is to perfection.
What Elo would a lifeform have if it played perfectly? For example, if it managed to score 3/4 against Kasparov when he was 2851 it would have an Elo of roughly 3050. If it scored 15/16 its Elo would be roughly 3250. In other words, how difficult would it be for Kasparov (or any other top GM/bot) to scrape a draw with White? (Of course, this assumes that there's no forced win for either side from the initial position.)
if he is close to complelty mastering the game then 3200 might be enough.
If not, then perhaps to play chess perfectly the elo would be closer to 9, or 20,000
03 Aug 06
2 edits
Originally posted by lauseyI asked what percentage a top GM/bot would score against perfect play, so as to get a ballpark estimate of 'God's Elo'. The answers can vary from '0' to '100'. So there's no need to philosophize about 'Perfection'. Is your answer '0'?
You did ask what someone's Elo would be if it played perfectly. It would be impossible to determine as it is a relative scale, just like asking how close someone is to perfection.
03 Aug 06
Originally posted by ThudanBlunderI would still say it would be impossible to determine. Any "guesses" would be just an arbitrary figure.
I asked what percentage a top GM/bot would score against perfect play, so as to get a ballpark estimate of 'God's Elo'. The answers can vary from '0' to '100'. So there's no need to philosophise about 'Perfection'. Is your answer '0'?
Therefore, my answer is that there is no way anyone could know what the percentage would be. 🙂
The only way we could possibly have any idea is to get two beings to play perfect play against each other and see what the outcome will be (to be able to determine what would be the perfect moves in any position).
03 Aug 06
1 edit
Originally posted by BigDoggProblemWhy so?
Depends on who it plays. It can do no better than a few hundred points above its opponents.
By the general rule that 75% represents roughly a 200 points Elo difference, if you regularly score 15/16 against Kasparov (say) wouldn't your Elo come out 400 points more than his? 63/64 would give you 600 points more, etc.
03 Aug 06
Originally posted by ThudanBlunderSo the real question is "what percentage of games by Kasparov, etc. are played perfectly?" You assume the is no forced win, thus if Kasparov played perfectly, it would be a draw since the other side will never make a mistake by definition. The premise is that chess is a forced draw. If Kasparov makes one mistake it will be a loss for him no matter how small since any move that maintains the forced draw is "correct". This would include mistakes that lead to a foced mate in 100, 200 etc. I don't think any human is there yet. Also, to play "perfectly" an opponent would play moves that attack Kasparov psychologically and would create a situation where Kasparov would make a mistake. A simplified example would be two "Gods" playing and each player perfect except one didn't know how to defend the scholars mate. Well, God one would know this and play the scholars mate every time, thus winning. The branches are too big to Kasparov to avoid this line of attack. I say Kasparov (I use Kasparov as a generic term for the best human player) loses 100%. Therefore, ELO is just a function of how highly rated the opponents are. If I only play 2000 players it will be a function of the opponents rating, not infinity. If you win every game what is your rating is probably a the question you are looking to answer.
I asked what percentage a top GM/bot would score against perfect play, so as to get a ballpark estimate of 'God's Elo'. The answers can vary from '0' to '100'. So there's no need to philosophize about 'Perfection'. Is your answer '0'?
03 Aug 06
8 edits
Originally posted by Gammastyle1) Yes, we must make the very reasonable assumption that there is no forced win from the initial position.
So the real question is "what percentage of games by Kasparov, etc. are played perfectly?" You assume the is no forced win, thus if Kasparov played perfectly, it would be a draw since the other side will never make a mistake by definition. The premise is that chess is a forced draw. If Kasparov makes one mistake it will be a loss for him no matter how n every game what is your rating is probably a the question you are looking to answer.
2) Kasparov (or whoever) need not play perfectly in order to draw, especially with White. (I define 'playing perfectly' as making all moves so as to maximise your chances of winning.) For example, he can chop off pieces at every favourable opportunity in order to draw. (After all, why should he try to win when it is impossible?) There is a big difference between this and playing perfectly.
3) It is not feasible to include subjective factors such as playing the man rather than the board.
4) If a player has a 100% record his Elo is non-quantifiable. He needs to concede at least one draw.
5) Elo is a function not only of your opponents' average Elo but also of your score against them, the latter also being dependent on how many games you play (assuming no 100% scores) - as you cannot score, for example, 99% if you play only 10 games.