asmFish

[syzygy]

I'm auditing the games now and something interesting is apparent. It was my perception while watching the games that while asmFish was 1,000-2,000 kN/s faster, SF seemed to search deeper. I just shrugged this notion off, thinking "That's impossible!" After all, asmFish IS Stockfish, just an outdated version that's much faster. So if one SF version is faster than the other, surely it's searching to greater overall depths too. But now after auditing the first three games (OK, I've got 97 more to go, so this data is preliminary), I have hard evidence supporting my perception.

So you have learned that the logic behind your "surely" is flawed.

So nps is not everything. The next thing you may learn is that a comparison of search depths reached by different engines is not everything either. (Yes, SF of today is a different engine than SF of last year.)

What is everything?

Playing strength is everything. The only way to measure it is many many games.

(But if you are comparing functionally identical engines, such as asmFish and the corresponding version of Stockfish having identical node counts, then it is enough to look at speed.)

[Stephen Ham]

Thanks for your valued input, Syzygy.

Rather than overall playing strength, I was instead testing tactical acuity. In that 100 game match with long TC, asmFish/PedantFish failed. That was a surprise as the test was skewed in its favor, as previously explained, given the latter's roughly 25% speed advantage. Tactics is generally all about searching faster/more than the opponent. Also, that's the equivalent of 25% more time.

Instead, in a 100 game match with a shared "normal" opening book leading to "standard" chess positions, I expect that SF would have won by a far larger number of games than the 3-to-1 victory rate it scored. In normal chess, there would have been greater opportunities for SF to manifest superior technical ability and evaluation. So again, full credit to SF's programmers.

I'm still auditing the match games. However, I want to report that I did find a couple of games where asmFish/PedantFish's search average was deeper than SF's. But when it happened, it was only by roughly one ply.

My next and last planned match is between the two engines that defeated SF: Raubfisch X40 and Cfish.

All the best,
-Steve-

[MikeB]

Hi Mike,

Thanks for your interest. Like you, I wish the asmFish effort was sustainable since conceptually, its superior speed would make it the strongest Stockfish (SF).

I'm auditing the games now and something interesting is apparent. It was my perception while watching the games that while asmFish was 1,000-2,000 kN/s faster, SF seemed to search deeper. I just shrugged this notion off, thinking "That's impossible!" After all, asmFish IS Stockfish, just an outdated version that's much faster. So if one SF version is faster than the other, surely it's searching to greater overall depths too. But now after auditing the first three games (OK, I've got 97 more to go, so this data is preliminary), I have hard evidence supporting my perception.

In game #1, SF's speed average was 6,309 kN/s versus 7,887 for asmFish. But SF searched to an average depth of 40.5 plies, versus 37.0 for asmFish.

In game #2, asmFish searched at an average of 7,419 kN/s to an average depth of 42.5 plies. But SF's numbers are: 6,018 kN/s and 49.6 plies.

In game #3, asmFish searched at an average of 6,344 kN/s to average depth of 36.9 plies. But SF's numbers are: 5,131 kN/s and 39.5 plies. I'll bet the remaining 97 games confirm my perception from watching some latter games.

Mike, does this suggest that recent SF coding results in more radical pruning to reach greater search depth?

All the very best,
-Steve-

Generally speaking , meaning not always true, but when you see the node count for the bench going down, it points to more pruning going on ( and resulting in deeper depth) , conversely when the bench nodes increases, it would would mean that less pruning ( or perhaps more extensions ) are happening. So your conjecture that "that recent SF coding results in more radical pruning to reach greater search depth" is correct. Quite frankly , that has been magic over the last ten years - such so that even when SF is searching at just 65K nodes per second - it is playing at GM strength today. The smartness in SF's pruning in amazing - but also leads to some well known blindspots that much weaker engines find instantly.

[Leo]

I'm auditing the games now and something interesting is apparent. It was my perception while watching the games that while asmFish was 1,000-2,000 kN/s faster, SF seemed to search deeper. I just shrugged this notion off, thinking "That's impossible!" After all, asmFish IS Stockfish, just an outdated version that's much faster. So if one SF version is faster than the other, surely it's searching to greater overall depths too. But now after auditing the first three games (OK, I've got 97 more to go, so this data is preliminary), I have hard evidence supporting my perception.

So you have learned that the logic behind your "surely" is flawed.

So nps is not everything. The next thing you may learn is that a comparison of search depths reached by different engines is not everything either. (Yes, SF of today is a different engine than SF of last year.)

What is everything?

Playing strength is everything. The only way to measure it is many many games.

(But if you are comparing functionally identical engines, such as asmFish and the corresponding version of Stockfish having identical node counts, then it is enough to look at speed.)

OK. Makes sense.

[leavenfish]

Hello all,

I'm continuing to search for the tactically strongest engine to assist my ICCF performance.

Precisely why true 'Correspondence Chess' is dead. It is no longer 'one on one'...brain vs brain. It is merely...an illusion for people's egos.
I quit it over 15 years ago (about 2399 ICCF) precisely for this reason...it was becoming something completely different.

Anyway, to get back on topic...I do wish asmFish was more viable...maybe a version made every couple of months of the best recent Stockfish....or when a, say, 25pt elo increase was registered in the current best Stockfish. I long used asmFish on my aging laptop OTB opening repertoire precisely because of the speedup.

I don't know about the legality, but maybe charging $25 for a year of updates under such criteria would make it more worthwhile (?) Clearly people will spend $$ on their chess engines...for whatever their reason.

[corres]

Hello all,
I'm continuing to search for the tactically strongest engine to assist my ICCF performance.

Precisely why true 'Correspondence Chess' is dead. It is no longer 'one on one'...brain vs brain. It is merely...an illusion for people's egos.
I quit it over 15 years ago (about 2399 ICCF) precisely for this reason...it was becoming something completely different.

In correspondence chess there was no time when the player could not receive help.
In the old time when chess engines/machines were more weaker or they did not exist correspondence chess player used chess encyclopedias, chess book, databases and helping from chess friends.
I know such a correspondent chess champion who was a leader of a strong chess team and he made analysis with the member of his team...
In general at that old time one needed much more chess knowledge for becoming a champion as nowadays needing. I also know a lot of strong correspondence player who ceased the play. Partly due to the cause of their ideas, partly the cause of materials (for intensive correspondence chess it needs relative much more money than it was in the past), partly they could not use computers.
But the correspondence chess exists at present also.

[leavenfish]

Hello all,
I'm continuing to search for the tactically strongest engine to assist my ICCF performance.

Precisely why true 'Correspondence Chess' is dead. It is no longer 'one on one'...brain vs brain. It is merely...an illusion for people's egos.
I quit it over 15 years ago (about 2399 ICCF) precisely for this reason...it was becoming something completely different.

In correspondence chess there was no time when the player could not receive help.
In the old time when chess engines/machines were more weaker or they did not exist correspondence chess player used chess encyclopedias, chess book, databases and helping from chess friends.
I know such a correspondent chess champion who was a leader of a strong chess team and he made analysis with the member of his team...
In general at that old time one needed much more chess knowledge for becoming a champion as nowadays needing. I also know a lot of strong correspondence player who ceased the play. Partly due to the cause of their ideas, partly the cause of materials (for intensive correspondence chess it needs relative much more money than it was in the past), partly they could not use computers.
But the correspondence chess exists at present also.

Receiving 'help' via historical data: books, games in databases was one thing (OTB players rely on this...and their memory as well), but MOVE GENERATION is quite a different thing. That is why the historically accepted idea of correspondence chess (one on one...just not face to face) is truly dead.

Move generation in any position by (largely flawless) calculation of engines and guiding those engines around a bit, is decidely more along the 'Advanced Chess' OTB concept once advocated (not really...just a temporary way to make $$ off the idea!) by Kasparov. It died...of course.

No, there is no way around it - engines killed the long standing idea of Correspondence Chess. What has been happening since their introduction means it is something fundamentally different. One should be honest and call it 'Advanced Correspondence Chess' or 'Assisted Correspondence Chess' or somesuch.|

There is a local Correspondence GM who has played many an OTB tournament. I have played him twice OTB and have scored 1.5/2 (and the 1/2 was because I failed to convert an endgame edge). Now, if we were to play a modern turn based game, the outcome might well be different. But without the help of his engines, he is merely mortal...roughly 1950 USCF OTB...an "A Player".

[Ozymandias]

the historically accepted idea of correspondence chess (one on one...just not face to face) is truly dead.

It was never one on one, that was the point. It doesn't really matter where the move comes from, it's either yours or it comes from an outside source you trust more. What's funny is, in classical CC, the move didn't come from an engine and yet, outside help was more outrageous than it is nowadays. One, it had to do with who you knew, now it's democratic, everyone can buy a silicon friend. Two, you could just go with the suggested move and be sure it'd be better than anything you could cook up, now with so many "friends" suggesting candidate moves, you have to take an active part.

[Graham Banks]

the historically accepted idea of correspondence chess (one on one...just not face to face) is truly dead.

It was never one on one, that was the point. It doesn't really matter where the move comes from, it's either yours or it comes from an outside source you trust more. What's funny is, in classical CC, the move didn't come from an engine and yet, outside help was more outrageous than it is nowadays. One, it had to do with who you knew, now it's democratic, everyone can buy a silicon friend. Two, you could just go with the suggested move and be sure it'd be better than anything you could cook up, now with so many "friends" suggesting candidate moves, you have to take an active part.

I never used to anybody for advice when I was playing. Can't speak for my opponents, but my openings research and figuring out plans and moves was just me. I would expect that it would have been the same for my opponents too.

[Ozymandias]

the historically accepted idea of correspondence chess (one on one...just not face to face) is truly dead.

It was never one on one, that was the point. It doesn't really matter where the move comes from, it's either yours or it comes from an outside source you trust more. What's funny is, in classical CC, the move didn't come from an engine and yet, outside help was more outrageous than it is nowadays. One, it had to do with who you knew, now it's democratic, everyone can buy a silicon friend. Two, you could just go with the suggested move and be sure it'd be better than anything you could cook up, now with so many "friends" suggesting candidate moves, you have to take an active part.

I never used to anybody for advice when I was playing. Can't speak for my opponents, but my openings research and figuring out plans and moves was just me. I would expect that it would have been the same for my opponents too.

For some of them sure, but the fact that just one of them could do it freely, means you don't really know how many of your opponents were. It's similar to the use of mobile phones today in FIDE games, you assume they aren't using it, otherwise you'd stop playing, but the truth is, cases of cheating abound. The real problem for people, however, isn't cheating per se, but rather who you do it with. If it's a flesh and bones opponent it's ok, but if it's a digital one all the alarms go off.