Ozymandias wrote: ↑Mon Dec 14, 2020 11:23 am
draws are present at all levels and proliferate more at the top, yet she keeps winning/losing not matter what her level or that of her opponents is supposed to be.
That's because the way she's represented. every single time the game would have been a draw she just kept pushing and pushing and swindling until she reached a position where draw wasn't possible anymore. And that caused her to lose. So it makes sense (either she's superior enough that she'll beat her opponent, or she reaches an equal position and then gets desperate to win and busts the game.)
That works sometimes, but if the result of your play is a win/loss in every situation, your play isn't that good at all, and she's supposed to play near perfect chess (at some point, they imply that perfect machine-like chess ends in a win, not a draw).
jdart wrote: ↑Sun Dec 13, 2020 12:07 pmdraws scarcely seem to be part of her playing record
At least as shown, and as far as I can recall, completely non-existent.
She draws against Benny Watts.
Can you give me the episode and minute:seconds point at which the scene begins? The way I remember it, she lost to tie for the tour?
She is co-champion with Benny Watts at the Las Vegas U.S. Open Chess Championship and I interpreted that as that they must have drawn in the finals.
I looked it up and I think I misunderstood that part. There wasn't a "finals" game, as that is not how the tournament is setup, instead she lost her head-to-head match against him (which we saw) but he drew other games so in the end they had the same amount of points.
Minimal Chess (simple, open source, C#) - Youtube & Github Leorik (competitive, in active development, C#) - Github & Lichess
Ozymandias wrote: ↑Thu Mar 18, 2021 8:01 amThat works sometimes, but if the result of your play is a win/loss in every situation, your play isn't that good at all
I disagree, I think someone could easily program a "Beth Harmon" that plays Stockfish NNUE's moves most of the time, except when the game is about to be a draw, then it switches to some ProDeo personality that avoids draw no matter what (or sacrifices a piece for a pawn, or the queen for 3 pieces, etc.) so either it recovers and wins or you can beat it.
This would produce an extremely strong opponent (lest we'd be getting reports of humans drawing against Stockfish NNUE) that would behave like Beth (throwing drawn games), you'd just need to switch SNNUE for a weaker engine (one that'd lost against the likes of Bobby Fisher.)
Ozymandias wrote: ↑Thu Mar 18, 2021 8:01 amThat works sometimes, but if the result of your play is a win/loss in every situation, your play isn't that good at all
I disagree, I think someone could easily program a "Beth Harmon" that plays Stockfish NNUE's moves most of the time, except when the game is about to be a draw, then it switches to some ProDeo personality that avoids draw no matter what (or sacrifices a piece for a pawn, or the queen for 3 pieces, etc.) so either it recovers and wins or you can beat it.
This would produce an extremely strong opponent (lest we'd be getting reports of humans drawing against Stockfish NNUE) that would behave like Beth (throwing drawn games), you'd just need to switch SNNUE for a weaker engine (one that'd lost against the likes of Bobby Fisher.)
It already exists, it's called armageddon chess and you don't need a composite engine for it.
Ozymandias wrote: ↑Thu Mar 18, 2021 8:01 amThat works sometimes, but if the result of your play is a win/loss in every situation, your play isn't that good at all
I disagree, I think someone could easily program a "Beth Harmon" that plays Stockfish NNUE's moves most of the time, except when the game is about to be a draw, then it switches to some ProDeo personality that avoids draw no matter what (or sacrifices a piece for a pawn, or the queen for 3 pieces, etc.) so either it recovers and wins or you can beat it.
This would produce an extremely strong opponent (lest we'd be getting reports of humans drawing against Stockfish NNUE) that would behave like Beth (throwing drawn games), you'd just need to switch SNNUE for a weaker engine (one that'd lost against the likes of Bobby Fisher.)
If you switch to an inferior engine, you get away from perfect chess, not closer. In any case, the series is about a person (the machine reference was an attempt on my part to understand where they're coming from).
Ozymandias wrote: ↑Thu Mar 18, 2021 8:01 amThat works sometimes, but if the result of your play is a win/loss in every situation, your play isn't that good at all
I disagree, I think someone could easily program a "Beth Harmon" that plays Stockfish NNUE's moves most of the time, except when the game is about to be a draw, then it switches to some ProDeo personality that avoids draw no matter what (or sacrifices a piece for a pawn, or the queen for 3 pieces, etc.) so either it recovers and wins or you can beat it.
This would produce an extremely strong opponent (lest we'd be getting reports of humans drawing against Stockfish NNUE) that would behave like Beth (throwing drawn games), you'd just need to switch SNNUE for a weaker engine (one that'd lost against the likes of Bobby Fisher.)
If you switch to an inferior engine, you get away from perfect chess, not closer. In any case, the series is about a person (the machine reference was an attempt on my part to understand where they're coming from).
This is not always true. A much stronger engine usually prunes a lot. Sometimes, a more brute force approach will find a better solution faster for certain classes of problems. The engine might be 2000 Elo weaker, but for a few very special problems, it can find the perfect solution in less time.
Taking ideas is not a vice, it is a virtue. We have another word for this. It is called learning.
But sharing ideas is an even greater virtue. We have another word for this. It is called teaching.
Ozymandias wrote: ↑Thu Mar 18, 2021 8:01 amThat works sometimes, but if the result of your play is a win/loss in every situation, your play isn't that good at all
I disagree, I think someone could easily program a "Beth Harmon" that plays Stockfish NNUE's moves most of the time, except when the game is about to be a draw, then it switches to some ProDeo personality that avoids draw no matter what (or sacrifices a piece for a pawn, or the queen for 3 pieces, etc.) so either it recovers and wins or you can beat it.
This would produce an extremely strong opponent (lest we'd be getting reports of humans drawing against Stockfish NNUE) that would behave like Beth (throwing drawn games), you'd just need to switch SNNUE for a weaker engine (one that'd lost against the likes of Bobby Fisher.)
If you switch to an inferior engine, you get away from perfect chess, not closer. In any case, the series is about a person (the machine reference was an attempt on my part to understand where they're coming from).
This is not always true. A much stronger engine usually prunes a lot. Sometimes, a more brute force approach will find a better solution faster for certain classes of problems. The engine might be 2000 Elo weaker, but for a few very special problems, it can find the perfect solution in less time.
If you throw away 99 draws to win an extra game, you're father away from perfect chess (99 out 100 times).
Ozymandias wrote: ↑Thu Mar 18, 2021 8:01 amThat works sometimes, but if the result of your play is a win/loss in every situation, your play isn't that good at all
I disagree, I think someone could easily program a "Beth Harmon" that plays Stockfish NNUE's moves most of the time, except when the game is about to be a draw, then it switches to some ProDeo personality that avoids draw no matter what (or sacrifices a piece for a pawn, or the queen for 3 pieces, etc.) so either it recovers and wins or you can beat it.
This would produce an extremely strong opponent (lest we'd be getting reports of humans drawing against Stockfish NNUE) that would behave like Beth (throwing drawn games), you'd just need to switch SNNUE for a weaker engine (one that'd lost against the likes of Bobby Fisher.)
If you switch to an inferior engine, you get away from perfect chess, not closer. In any case, the series is about a person (the machine reference was an attempt on my part to understand where they're coming from).
This is not always true. A much stronger engine usually prunes a lot. Sometimes, a more brute force approach will find a better solution faster for certain classes of problems. The engine might be 2000 Elo weaker, but for a few very special problems, it can find the perfect solution in less time.
If you throw away 99 draws to win an extra game, you're father away from perfect chess (99 out 100 times).
True. A weaker engine plays less perfectly. But sometimes it may play better than the stronger engine.
My claim was not that a weaker engine plays more perfectly. It was that once in a while it might play more perfectly.
Taking ideas is not a vice, it is a virtue. We have another word for this. It is called learning.
But sharing ideas is an even greater virtue. We have another word for this. It is called teaching.
Ozymandias wrote: ↑Thu Mar 18, 2021 8:01 amThat works sometimes, but if the result of your play is a win/loss in every situation, your play isn't that good at all
I disagree, I think someone could easily program a "Beth Harmon" that plays Stockfish NNUE's moves most of the time, except when the game is about to be a draw, then it switches to some ProDeo personality that avoids draw no matter what (or sacrifices a piece for a pawn, or the queen for 3 pieces, etc.) so either it recovers and wins or you can beat it.
This would produce an extremely strong opponent (lest we'd be getting reports of humans drawing against Stockfish NNUE) that would behave like Beth (throwing drawn games), you'd just need to switch SNNUE for a weaker engine (one that'd lost against the likes of Bobby Fisher.)
If you switch to an inferior engine, you get away from perfect chess, not closer. In any case, the series is about a person (the machine reference was an attempt on my part to understand where they're coming from).
This is not always true. A much stronger engine usually prunes a lot. Sometimes, a more brute force approach will find a better solution faster for certain classes of problems. The engine might be 2000 Elo weaker, but for a few very special problems, it can find the perfect solution in less time.
So getting worse in 99.9999% of positions but getting better in 0.0001% of positions for you is being closer to perfect chess????
This reply of yours should stand in a textbook as the most obvious example of a strawman.
Taking ideas is not a vice, it is a virtue. We have another word for this. It is called learning.
But sharing ideas is an even greater virtue. We have another word for this. It is called teaching.