The Superfinal

[Lyudmil Tsvetkov]

The problem for stockfish is not to find Rxd4(stockfish can find it very fast but to avoid e4

stockfish needs a lot of time to avoid komodo's e3-e4

[D]1k1r3r/ppq1bpp1/1np1pn2/6p1/3P4/P1N1P2P/1P1BBPP1/2RQR1K1 w - - 0 17

Stockfish_14051921_x64:

35/52 28:13 4,201,289,021 2,481,325 +0.17 e3-e4 Rd8xd4 Nc3-b5 Rd4xd2 Qd1xd2 Qc7-f4 Qd2xf4+ g5xf4 Nb5-c3 Nb6-d7 Rc1-c2 Kb8-c7 b2-b3 Nd7-e5 b3-b4 a7-a6 Nc3-a4 f4-f3 Be2xf3 Rh8xh3 Re1-e3 Nf6-g4 Re3-b3 Ne5xf3+ Rb3xf3 Rh3xf3 g2xf3 Ng4-e5 Kg1-g2 b7-b5 Na4-c5 a6-a5 Rc2-c3 a5xb4 a3xb4 Be7xc5 Rc3xc5 Ne5-d3 Rc5-c3 Nd3xb4 Rc3-a3
36/55+ 36:11 5,350,747,191 2,465,053 +0.23 Qd1-b3 g5-g4 e3-e4 Rd8-d7
36/55+ 42:06 6,301,367,591 2,494,652 +0.29 Qd1-b3 Nb6-d5 Be2-f3 Be7-d6 Rc1-c2 Bd6-h2+ Kg1-f1 Bh2-d6 e3-e4 Nd5xc3 b2xc3 g5-g4 h3xg4 Rh8-h1+ Kf1-e2 Rh1xe1+ Bd2xe1 Bd6-f4 e4-e5 Nf6-d7
36/58 1:14:02 10,952,736,084 2,465,730 +0.23 Qd1-b3 g5-g4 e3-e4 Kb8-a8 Bd2-e3 Rd8-c8 Re1-d1 Nb6-d7 a3-a4 g4xh3 g2-g3 Qc7-b6 Qb3xb6 Nd7xb6 Kg1-h2 Be7-b4 Nc3-a2 Bb4-e7 a4-a5 Nb6-d7 f2-f3 a7-a6 Na2-c3

Latest Stockfish (16 threads, 8 GB hash) is faster, but still takes more than 3 minutes. Also chooses b4 rather than Qb3.

Code: Select all

info depth 33 seldepth 56 score cp 10 nodes 1141203347 nps 16911476 time 67481 multipv 1 pv e3e4 d8d4 c3b5 d4d2 d1d2 c7f4 d2f4 g5f4 b5c3 b6d7 c1c2 a7a5 e1d1 d7e5 f2f3 g7g5 c3a2 f6d7 b2b4 a5b4 a3b4 b7b5 c2c3 b8b7 c3a3 d7f6 g1f2 b7b6 f2e1 b6b7 e1f1 b7b6 f1f2 c6c5 a3a5 c5c4 a2c3 e7c5 f2f1
info depth 34 currmove e3e4 currmovenumber 1
info depth 34 currmove e2d3 currmovenumber 6
info depth 34 seldepth 56 score cp 16 lowerbound nodes 4136208439 nps 18412117 time 224646 multipv 1 pv b2b4 a7a6 b4b5 e7a3 c1c2 g5g4 b5a6
info depth 34 currmove b2b4 currmovenumber 1
info depth 34 currmove e3e4 currmovenumber 2
info depth 34 seldepth 56 score cp 20 nodes 4887853008 nps 18427482 time 265248 multipv 1 pv b2b4 e6e5 c1c2 h8h4 c3b5 c7d7 d4e5 c6b5 e5f6 e7f6 d2c3 d7d1 e1d1 d8d1 e2d1 f6c3 c2c3 h4c4 c3d3 b8c7 g1h2 c4c1 h2g3 g7g6 d1g4 f7f5 g4f3 a7a6 e3e4 c1c2 f3d1 c2a2 e4f5 g6f5 f2f4 g5f4 g3f4 b6c4 d1f3
info depth 35 currmove b2b4 currmovenumber 1
info depth 35 seldepth 56 score cp 26 lowerbound nodes 6792732009 nps 18446380 time 368242 multipv 1 pv b2b4 e6e5 c1c2 h8h4 e2f1 e5d4 e3d4 h4d4 d1c1 e7d6 c3b5
info depth 35 currmove b2b4 currmovenumber 1
info depth 35 seldepth 56 score cp 32 lowerbound nodes 8706311875 nps 18323214 time 475152 multipv 1 pv b2b4 a7a6 b4b5 e7a3 c1c2 g5g4
info depth 35 currmove b2b4 currmovenumber 1
info depth 35 currmove e3e4 currmovenumber 2
info depth 35 seldepth 56 score cp 23 nodes 13057785965 nps 18583178 time 702667 multipv 1 pv b2b4 g7g6 a3a4 g5g4 a4a5 b6d5 c3d5 e6d5 e2g4 a7a6 g4f3 c7d6 d1b3 h8h4 b4b5 c6b5 c1c5 h4h8 d2b4 d6e6 c5c2 e7b4 b3b4 d8c8 c2c5
info depth 36 currmove b2b4 currmovenumber 1
info depth 36 seldepth 56 score cp 29 lowerbound nodes 13474754925 nps 18588484 time 724898 multipv 1 pv b2b4 g7g6 a3a4 g5g4 a4a5 b6d5
info depth 36 currmove b2b4 currmovenumber 1
info depth 36 seldepth 56 score cp 36 lowerbound nodes 13938459658 nps 18566764 time 750721 multipv 1 pv b2b4 g5g4 e2g4 b6c4 g4e2 c4d2 d1d2 g7g5 b4b5 g5g4 d2b2 g4h3 g2g3 c6b5 c3d5 c7d6 d5e7
info depth 36 currmove b2b4 currmovenumber 1
info depth 36 currmove e3e4 currmovenumber 2
info depth 36 seldepth 56 score cp 36 nodes 18500753671 nps 18709779 time 988828 multipv 1 pv b2b4 c7d7 c1c2 h8h4 e2f1 f6d5 g2g3 d5c3 d2c3 h4h8 d1e2 f7f5 c3d2 d8f8 f1g2 f8c8 e1b1 a7a6 c2a2 b6a4 b1c1 e7f6 a2c2 a4b6

I would have chosen b4 myself.

[Henk]

The problem for stockfish is not to find Rxd4(stockfish can find it very fast but to avoid e4

stockfish needs a lot of time to avoid komodo's e3-e4

[D]1k1r3r/ppq1bpp1/1np1pn2/6p1/3P4/P1N1P2P/1P1BBPP1/2RQR1K1 w - - 0 17

Stockfish_14051921_x64:

35/52 28:13 4,201,289,021 2,481,325 +0.17 e3-e4 Rd8xd4 Nc3-b5 Rd4xd2 Qd1xd2 Qc7-f4 Qd2xf4+ g5xf4 Nb5-c3 Nb6-d7 Rc1-c2 Kb8-c7 b2-b3 Nd7-e5 b3-b4 a7-a6 Nc3-a4 f4-f3 Be2xf3 Rh8xh3 Re1-e3 Nf6-g4 Re3-b3 Ne5xf3+ Rb3xf3 Rh3xf3 g2xf3 Ng4-e5 Kg1-g2 b7-b5 Na4-c5 a6-a5 Rc2-c3 a5xb4 a3xb4 Be7xc5 Rc3xc5 Ne5-d3 Rc5-c3 Nd3xb4 Rc3-a3
36/55+ 36:11 5,350,747,191 2,465,053 +0.23 Qd1-b3 g5-g4 e3-e4 Rd8-d7
36/55+ 42:06 6,301,367,591 2,494,652 +0.29 Qd1-b3 Nb6-d5 Be2-f3 Be7-d6 Rc1-c2 Bd6-h2+ Kg1-f1 Bh2-d6 e3-e4 Nd5xc3 b2xc3 g5-g4 h3xg4 Rh8-h1+ Kf1-e2 Rh1xe1+ Bd2xe1 Bd6-f4 e4-e5 Nf6-d7
36/58 1:14:02 10,952,736,084 2,465,730 +0.23 Qd1-b3 g5-g4 e3-e4 Kb8-a8 Bd2-e3 Rd8-c8 Re1-d1 Nb6-d7 a3-a4 g4xh3 g2-g3 Qc7-b6 Qb3xb6 Nd7xb6 Kg1-h2 Be7-b4 Nc3-a2 Bb4-e7 a4-a5 Nb6-d7 f2-f3 a7-a6 Na2-c3

Latest Stockfish (16 threads, 8 GB hash) is faster, but still takes more than 3 minutes. Also chooses b4 rather than Qb3.

Code: Select all

info depth 33 seldepth 56 score cp 10 nodes 1141203347 nps 16911476 time 67481 multipv 1 pv e3e4 d8d4 c3b5 d4d2 d1d2 c7f4 d2f4 g5f4 b5c3 b6d7 c1c2 a7a5 e1d1 d7e5 f2f3 g7g5 c3a2 f6d7 b2b4 a5b4 a3b4 b7b5 c2c3 b8b7 c3a3 d7f6 g1f2 b7b6 f2e1 b6b7 e1f1 b7b6 f1f2 c6c5 a3a5 c5c4 a2c3 e7c5 f2f1
info depth 34 currmove e3e4 currmovenumber 1
info depth 34 currmove e2d3 currmovenumber 6
info depth 34 seldepth 56 score cp 16 lowerbound nodes 4136208439 nps 18412117 time 224646 multipv 1 pv b2b4 a7a6 b4b5 e7a3 c1c2 g5g4 b5a6
info depth 34 currmove b2b4 currmovenumber 1
info depth 34 currmove e3e4 currmovenumber 2
info depth 34 seldepth 56 score cp 20 nodes 4887853008 nps 18427482 time 265248 multipv 1 pv b2b4 e6e5 c1c2 h8h4 c3b5 c7d7 d4e5 c6b5 e5f6 e7f6 d2c3 d7d1 e1d1 d8d1 e2d1 f6c3 c2c3 h4c4 c3d3 b8c7 g1h2 c4c1 h2g3 g7g6 d1g4 f7f5 g4f3 a7a6 e3e4 c1c2 f3d1 c2a2 e4f5 g6f5 f2f4 g5f4 g3f4 b6c4 d1f3
info depth 35 currmove b2b4 currmovenumber 1
info depth 35 seldepth 56 score cp 26 lowerbound nodes 6792732009 nps 18446380 time 368242 multipv 1 pv b2b4 e6e5 c1c2 h8h4 e2f1 e5d4 e3d4 h4d4 d1c1 e7d6 c3b5
info depth 35 currmove b2b4 currmovenumber 1
info depth 35 seldepth 56 score cp 32 lowerbound nodes 8706311875 nps 18323214 time 475152 multipv 1 pv b2b4 a7a6 b4b5 e7a3 c1c2 g5g4
info depth 35 currmove b2b4 currmovenumber 1
info depth 35 currmove e3e4 currmovenumber 2
info depth 35 seldepth 56 score cp 23 nodes 13057785965 nps 18583178 time 702667 multipv 1 pv b2b4 g7g6 a3a4 g5g4 a4a5 b6d5 c3d5 e6d5 e2g4 a7a6 g4f3 c7d6 d1b3 h8h4 b4b5 c6b5 c1c5 h4h8 d2b4 d6e6 c5c2 e7b4 b3b4 d8c8 c2c5
info depth 36 currmove b2b4 currmovenumber 1
info depth 36 seldepth 56 score cp 29 lowerbound nodes 13474754925 nps 18588484 time 724898 multipv 1 pv b2b4 g7g6 a3a4 g5g4 a4a5 b6d5
info depth 36 currmove b2b4 currmovenumber 1
info depth 36 seldepth 56 score cp 36 lowerbound nodes 13938459658 nps 18566764 time 750721 multipv 1 pv b2b4 g5g4 e2g4 b6c4 g4e2 c4d2 d1d2 g7g5 b4b5 g5g4 d2b2 g4h3 g2g3 c6b5 c3d5 c7d6 d5e7
info depth 36 currmove b2b4 currmovenumber 1
info depth 36 currmove e3e4 currmovenumber 2
info depth 36 seldepth 56 score cp 36 nodes 18500753671 nps 18709779 time 988828 multipv 1 pv b2b4 c7d7 c1c2 h8h4 e2f1 f6d5 g2g3 d5c3 d2c3 h4h8 d1e2 f7f5 c3d2 d8f8 f1g2 f8c8 e1b1 a7a6 c2a2 b6a4 b1c1 e7f6 a2c2 a4b6

I would have chosen b4 myself.

I would be surprised if someone that has just learned chess and has an ELO rating which is 1000 lower than mine, starts commenting my chess games.

Maybe the only way to comment on a game of Stockfish is to use Stockfish or another top chess program with much longer time control.

[Lyudmil Tsvetkov]

I might post here brief remarks on some of the most memorable games from the Superfinal. I am sure Martin will not be angry with me.

Anyone contributing with remarks, engine output, etc. very much appreciated. Let us make this a meaningful discussion thread.

Lyudmil, Game #1 and Game #8 really caught my eye (so far) because both revolved around the theme of exchange sacs. It is curious that SF evaluated both sacs correctly while Komodo got both wrong!

In Game #1, Komodo miscalculated the tactic and SF quickly started to gain ground with the pair of rooks. In Game #8, SF sac-ed the exchange purely for positional reasons, and was able to grind White's position down with the knight duo! Amazing games!

Perhaps it's time for you to forget about SF's imbalances and start tuning Komodo's!!

Thanks for participating in the discussion, Arjun!

In games 1 it was actually Komodo that was forced to sacrifice the exchange. That game however did not impress me, as the white advantage was more or less clear, and increasing it did not require special skills. Game 8 is a real gem: nothing concrete, brilliant exchange sacrifice, big depths.

Please do not say that SF has nothing to correct, because the match is still not over. It definitely should correct its imbalance eval. I say the single change of increasing the knight value for the mg by 5-8% will bring 5 elo more.

Lyudmil, now that spsa (an autotuning algorithm) has been integrated into fishtest, I think it won't be long before somebody will post a tuning patch for material values.

I am sure you know a lot about chess, but engine improvements don't work this way. You can't just simply change a parameter by intuition, and magically 5 elo springs out of the hat... It doesn't happen. At least not with an engine like SF which has been tuned to death!

Anyway if you are curious, I ran a quick match on my machine with knight midgame values 5% higher (857) and the result after 600 games is

Code: Select all

137 - 177 - 286 [0.467] 600

-23 elo!

So as you can see, intuition will not get you elo. Also don't forget that the evaluation function is a big beast and so many terms correlate and add up over time, it is hard to keep track. What may seem like a deficiency in "knight value" may not be really. It is probably a function of multiple terms in evaluation, search depth, threading, and god knows what... To find the real problem (if there is any), it will take some time and debugging.

So my suggestion to you Lyudmil is to play to your strength. You know about chess. Talk about chess. Don't bother speculating about values so much. Just talk about the chess aspects of a position and let the developers figure out what the problem is.

But keep the ideas coming!

Now, lets get back to the topic of this thread.

Too few games, Arjun. You should have played a lot more. And also try a 2% and 8% increase separately, to know where things are going. You are right about the interrelation of many terms, however, when there is problem observed in games, it should be fixable.

About CLOP and SPSA. Please do not make me laugh, Arjun. There is hardly a single value those tuners managed to guess right, even a single. Look at the current massive SPSA tuning: all tests fail after 200 000 iterations! On the other hand, human-suggested values succeed much more often.

I didn't play more because it was a pointless endeavour. Random tuning gets you nowhere. I suggest if you want to see the result, test it yourself. I guarantee you it will fail because I have dabbled with material tables before and they are rock solid. You might squeeze 1 elo out of it somehow but only after 100000 spsa iterations or so. I prefer to use my CPU time where it is most useful, rather than playing out someone's fantasy values.

You seem very biased in your observations about things.

For example, you point out the 1 SPSA test that did not work among so many others that did. Just because they weren't done on the public framework, doesn't mean they were not done! Ask Joona, who has practically tuned every large array in stockfish there is. The King PSQT and Knight PSQT for example were done quite recently. All the imbalances and material tables and many others were done long ago. SF is strong today because of months and months of autotuning done in the glaurung and early SF days. You think he intuited all the values? Yeah right!

(Btw, about the recent test, it could be that the values in that test are already rock hard and so there is barely anything to gain or it could be some other reason related to the spsa setup on fishtest, either way, it is just the first test -- an experiment so to say.)

Meanwhile, you claim human intuition has brought the most success. But you conveniently forget the countless tests based on your "intuited" values done (some by me) that have failed. BlockedOn6th? Doubled pawns? Rings a bell? Poor joerg even tried your futile attempt at "intuiting" the imbalance table, and failed obviously because that is just a silly thing to do. We should leave such endeavors to the autotuners.

I suggest you stop laughing at others who know what they are talking about, because clearly you have no idea about what you are talking about and you know very little, if anything, about how an engine works, so take my advice and stick to chess.

Arjun, sorry to post here again, this thread belongs to other things, but I have to. It was not me, however, but you that did raise the imbalance topic here.

You were talking the same about CLOP, and it is a fact that 98% of all CLOP tuning attempts failed in the framework. SPSA might be a bit better, but you see what is happening, there will be more to see. You have to say not that the first attempt at the framework failed, but in order to be correct, that the first 3 attempts failed, is not that right Arjun? Joona tuned some psqt tables, but you know very well that in order to start tuning, you must feed the tuner some pretty accurate values. And those pretty accurate values are human-suggested, Arjun, so that humans again play a more important role.

It is funny that you mention Joerg, but 2 of the 4 npm tests succeeded at STC, something that does not happen often when you want to change imbalance values. Eelco's attempt to tune the pawn values was also successful even at LTC, it needed just a bit.

Blocked on 6th failed because it was not implemented right, you know that you have to exclude storming pawns from the bonus. And about doubled I have to say that this was one of your few creative attempts that went to LTC, so it was not a bad try at all. Repeating the ugly experiment with penalising only h file certainly has much better chances to succeed.

I can assure you there is much to gain from imbalance, really much, much more than in other areas that are overexploited to no avail. But you have to understand a bit how an engine behaves in chess games, not just tune blindly. I would gladly run the knight value test, but as you know, not everyone can do everything at the same time.

After all, this brings us nowhere. I think that chess players' and programmers' cooperation benefits both sides and you only have to expand it.

[Dr.Wael Deeb]

The problem for stockfish is not to find Rxd4(stockfish can find it very fast but to avoid e4

stockfish needs a lot of time to avoid komodo's e3-e4

[D]1k1r3r/ppq1bpp1/1np1pn2/6p1/3P4/P1N1P2P/1P1BBPP1/2RQR1K1 w - - 0 17

Stockfish_14051921_x64:

35/52 28:13 4,201,289,021 2,481,325 +0.17 e3-e4 Rd8xd4 Nc3-b5 Rd4xd2 Qd1xd2 Qc7-f4 Qd2xf4+ g5xf4 Nb5-c3 Nb6-d7 Rc1-c2 Kb8-c7 b2-b3 Nd7-e5 b3-b4 a7-a6 Nc3-a4 f4-f3 Be2xf3 Rh8xh3 Re1-e3 Nf6-g4 Re3-b3 Ne5xf3+ Rb3xf3 Rh3xf3 g2xf3 Ng4-e5 Kg1-g2 b7-b5 Na4-c5 a6-a5 Rc2-c3 a5xb4 a3xb4 Be7xc5 Rc3xc5 Ne5-d3 Rc5-c3 Nd3xb4 Rc3-a3
36/55+ 36:11 5,350,747,191 2,465,053 +0.23 Qd1-b3 g5-g4 e3-e4 Rd8-d7
36/55+ 42:06 6,301,367,591 2,494,652 +0.29 Qd1-b3 Nb6-d5 Be2-f3 Be7-d6 Rc1-c2 Bd6-h2+ Kg1-f1 Bh2-d6 e3-e4 Nd5xc3 b2xc3 g5-g4 h3xg4 Rh8-h1+ Kf1-e2 Rh1xe1+ Bd2xe1 Bd6-f4 e4-e5 Nf6-d7
36/58 1:14:02 10,952,736,084 2,465,730 +0.23 Qd1-b3 g5-g4 e3-e4 Kb8-a8 Bd2-e3 Rd8-c8 Re1-d1 Nb6-d7 a3-a4 g4xh3 g2-g3 Qc7-b6 Qb3xb6 Nd7xb6 Kg1-h2 Be7-b4 Nc3-a2 Bb4-e7 a4-a5 Nb6-d7 f2-f3 a7-a6 Na2-c3

Latest Stockfish (16 threads, 8 GB hash) is faster, but still takes more than 3 minutes. Also chooses b4 rather than Qb3.

Code: Select all

info depth 33 seldepth 56 score cp 10 nodes 1141203347 nps 16911476 time 67481 multipv 1 pv e3e4 d8d4 c3b5 d4d2 d1d2 c7f4 d2f4 g5f4 b5c3 b6d7 c1c2 a7a5 e1d1 d7e5 f2f3 g7g5 c3a2 f6d7 b2b4 a5b4 a3b4 b7b5 c2c3 b8b7 c3a3 d7f6 g1f2 b7b6 f2e1 b6b7 e1f1 b7b6 f1f2 c6c5 a3a5 c5c4 a2c3 e7c5 f2f1
info depth 34 currmove e3e4 currmovenumber 1
info depth 34 currmove e2d3 currmovenumber 6
info depth 34 seldepth 56 score cp 16 lowerbound nodes 4136208439 nps 18412117 time 224646 multipv 1 pv b2b4 a7a6 b4b5 e7a3 c1c2 g5g4 b5a6
info depth 34 currmove b2b4 currmovenumber 1
info depth 34 currmove e3e4 currmovenumber 2
info depth 34 seldepth 56 score cp 20 nodes 4887853008 nps 18427482 time 265248 multipv 1 pv b2b4 e6e5 c1c2 h8h4 c3b5 c7d7 d4e5 c6b5 e5f6 e7f6 d2c3 d7d1 e1d1 d8d1 e2d1 f6c3 c2c3 h4c4 c3d3 b8c7 g1h2 c4c1 h2g3 g7g6 d1g4 f7f5 g4f3 a7a6 e3e4 c1c2 f3d1 c2a2 e4f5 g6f5 f2f4 g5f4 g3f4 b6c4 d1f3
info depth 35 currmove b2b4 currmovenumber 1
info depth 35 seldepth 56 score cp 26 lowerbound nodes 6792732009 nps 18446380 time 368242 multipv 1 pv b2b4 e6e5 c1c2 h8h4 e2f1 e5d4 e3d4 h4d4 d1c1 e7d6 c3b5
info depth 35 currmove b2b4 currmovenumber 1
info depth 35 seldepth 56 score cp 32 lowerbound nodes 8706311875 nps 18323214 time 475152 multipv 1 pv b2b4 a7a6 b4b5 e7a3 c1c2 g5g4
info depth 35 currmove b2b4 currmovenumber 1
info depth 35 currmove e3e4 currmovenumber 2
info depth 35 seldepth 56 score cp 23 nodes 13057785965 nps 18583178 time 702667 multipv 1 pv b2b4 g7g6 a3a4 g5g4 a4a5 b6d5 c3d5 e6d5 e2g4 a7a6 g4f3 c7d6 d1b3 h8h4 b4b5 c6b5 c1c5 h4h8 d2b4 d6e6 c5c2 e7b4 b3b4 d8c8 c2c5
info depth 36 currmove b2b4 currmovenumber 1
info depth 36 seldepth 56 score cp 29 lowerbound nodes 13474754925 nps 18588484 time 724898 multipv 1 pv b2b4 g7g6 a3a4 g5g4 a4a5 b6d5
info depth 36 currmove b2b4 currmovenumber 1
info depth 36 seldepth 56 score cp 36 lowerbound nodes 13938459658 nps 18566764 time 750721 multipv 1 pv b2b4 g5g4 e2g4 b6c4 g4e2 c4d2 d1d2 g7g5 b4b5 g5g4 d2b2 g4h3 g2g3 c6b5 c3d5 c7d6 d5e7
info depth 36 currmove b2b4 currmovenumber 1
info depth 36 currmove e3e4 currmovenumber 2
info depth 36 seldepth 56 score cp 36 nodes 18500753671 nps 18709779 time 988828 multipv 1 pv b2b4 c7d7 c1c2 h8h4 e2f1 f6d5 g2g3 d5c3 d2c3 h4h8 d1e2 f7f5 c3d2 d8f8 f1g2 f8c8 e1b1 a7a6 c2a2 b6a4 b1c1 e7f6 a2c2 a4b6

I would have chosen b4 myself.

I would be surprised if someone that has just learned chess and has an ELO rating which is 1000 lower than mine, starts commenting my chess games.

Maybe the only way to comment on a game of Stockfish is to use Stockfish or another top chess program with much longer time control.

I wasn't aware that Ludmyl is 1000 Elo weaker than you

[Lyudmil Tsvetkov]

The problem for stockfish is not to find Rxd4(stockfish can find it very fast but to avoid e4

stockfish needs a lot of time to avoid komodo's e3-e4

[D]1k1r3r/ppq1bpp1/1np1pn2/6p1/3P4/P1N1P2P/1P1BBPP1/2RQR1K1 w - - 0 17

Stockfish_14051921_x64:

35/52 28:13 4,201,289,021 2,481,325 +0.17 e3-e4 Rd8xd4 Nc3-b5 Rd4xd2 Qd1xd2 Qc7-f4 Qd2xf4+ g5xf4 Nb5-c3 Nb6-d7 Rc1-c2 Kb8-c7 b2-b3 Nd7-e5 b3-b4 a7-a6 Nc3-a4 f4-f3 Be2xf3 Rh8xh3 Re1-e3 Nf6-g4 Re3-b3 Ne5xf3+ Rb3xf3 Rh3xf3 g2xf3 Ng4-e5 Kg1-g2 b7-b5 Na4-c5 a6-a5 Rc2-c3 a5xb4 a3xb4 Be7xc5 Rc3xc5 Ne5-d3 Rc5-c3 Nd3xb4 Rc3-a3
36/55+ 36:11 5,350,747,191 2,465,053 +0.23 Qd1-b3 g5-g4 e3-e4 Rd8-d7
36/55+ 42:06 6,301,367,591 2,494,652 +0.29 Qd1-b3 Nb6-d5 Be2-f3 Be7-d6 Rc1-c2 Bd6-h2+ Kg1-f1 Bh2-d6 e3-e4 Nd5xc3 b2xc3 g5-g4 h3xg4 Rh8-h1+ Kf1-e2 Rh1xe1+ Bd2xe1 Bd6-f4 e4-e5 Nf6-d7
36/58 1:14:02 10,952,736,084 2,465,730 +0.23 Qd1-b3 g5-g4 e3-e4 Kb8-a8 Bd2-e3 Rd8-c8 Re1-d1 Nb6-d7 a3-a4 g4xh3 g2-g3 Qc7-b6 Qb3xb6 Nd7xb6 Kg1-h2 Be7-b4 Nc3-a2 Bb4-e7 a4-a5 Nb6-d7 f2-f3 a7-a6 Na2-c3

Latest Stockfish (16 threads, 8 GB hash) is faster, but still takes more than 3 minutes. Also chooses b4 rather than Qb3.

Code: Select all

info depth 33 seldepth 56 score cp 10 nodes 1141203347 nps 16911476 time 67481 multipv 1 pv e3e4 d8d4 c3b5 d4d2 d1d2 c7f4 d2f4 g5f4 b5c3 b6d7 c1c2 a7a5 e1d1 d7e5 f2f3 g7g5 c3a2 f6d7 b2b4 a5b4 a3b4 b7b5 c2c3 b8b7 c3a3 d7f6 g1f2 b7b6 f2e1 b6b7 e1f1 b7b6 f1f2 c6c5 a3a5 c5c4 a2c3 e7c5 f2f1
info depth 34 currmove e3e4 currmovenumber 1
info depth 34 currmove e2d3 currmovenumber 6
info depth 34 seldepth 56 score cp 16 lowerbound nodes 4136208439 nps 18412117 time 224646 multipv 1 pv b2b4 a7a6 b4b5 e7a3 c1c2 g5g4 b5a6
info depth 34 currmove b2b4 currmovenumber 1
info depth 34 currmove e3e4 currmovenumber 2
info depth 34 seldepth 56 score cp 20 nodes 4887853008 nps 18427482 time 265248 multipv 1 pv b2b4 e6e5 c1c2 h8h4 c3b5 c7d7 d4e5 c6b5 e5f6 e7f6 d2c3 d7d1 e1d1 d8d1 e2d1 f6c3 c2c3 h4c4 c3d3 b8c7 g1h2 c4c1 h2g3 g7g6 d1g4 f7f5 g4f3 a7a6 e3e4 c1c2 f3d1 c2a2 e4f5 g6f5 f2f4 g5f4 g3f4 b6c4 d1f3
info depth 35 currmove b2b4 currmovenumber 1
info depth 35 seldepth 56 score cp 26 lowerbound nodes 6792732009 nps 18446380 time 368242 multipv 1 pv b2b4 e6e5 c1c2 h8h4 e2f1 e5d4 e3d4 h4d4 d1c1 e7d6 c3b5
info depth 35 currmove b2b4 currmovenumber 1
info depth 35 seldepth 56 score cp 32 lowerbound nodes 8706311875 nps 18323214 time 475152 multipv 1 pv b2b4 a7a6 b4b5 e7a3 c1c2 g5g4
info depth 35 currmove b2b4 currmovenumber 1
info depth 35 currmove e3e4 currmovenumber 2
info depth 35 seldepth 56 score cp 23 nodes 13057785965 nps 18583178 time 702667 multipv 1 pv b2b4 g7g6 a3a4 g5g4 a4a5 b6d5 c3d5 e6d5 e2g4 a7a6 g4f3 c7d6 d1b3 h8h4 b4b5 c6b5 c1c5 h4h8 d2b4 d6e6 c5c2 e7b4 b3b4 d8c8 c2c5
info depth 36 currmove b2b4 currmovenumber 1
info depth 36 seldepth 56 score cp 29 lowerbound nodes 13474754925 nps 18588484 time 724898 multipv 1 pv b2b4 g7g6 a3a4 g5g4 a4a5 b6d5
info depth 36 currmove b2b4 currmovenumber 1
info depth 36 seldepth 56 score cp 36 lowerbound nodes 13938459658 nps 18566764 time 750721 multipv 1 pv b2b4 g5g4 e2g4 b6c4 g4e2 c4d2 d1d2 g7g5 b4b5 g5g4 d2b2 g4h3 g2g3 c6b5 c3d5 c7d6 d5e7
info depth 36 currmove b2b4 currmovenumber 1
info depth 36 currmove e3e4 currmovenumber 2
info depth 36 seldepth 56 score cp 36 nodes 18500753671 nps 18709779 time 988828 multipv 1 pv b2b4 c7d7 c1c2 h8h4 e2f1 f6d5 g2g3 d5c3 d2c3 h4h8 d1e2 f7f5 c3d2 d8f8 f1g2 f8c8 e1b1 a7a6 c2a2 b6a4 b1c1 e7f6 a2c2 a4b6

I would have chosen b4 myself.

I would be surprised if someone that has just learned chess and has an ELO rating which is 1000 lower than mine, starts commenting my chess games.

Maybe the only way to comment on a game of Stockfish is to use Stockfish or another top chess program with much longer time control.

Hi Henk.

I learned chess much time before SF's first line of code was born.

[Henk]

When Capablanca was four or five years old he already played better than his father. So it's possible. But ELO wise I don't think so.

[Henk]

The problem for stockfish is not to find Rxd4(stockfish can find it very fast but to avoid e4

stockfish needs a lot of time to avoid komodo's e3-e4

[D]1k1r3r/ppq1bpp1/1np1pn2/6p1/3P4/P1N1P2P/1P1BBPP1/2RQR1K1 w - - 0 17

Stockfish_14051921_x64:

35/52 28:13 4,201,289,021 2,481,325 +0.17 e3-e4 Rd8xd4 Nc3-b5 Rd4xd2 Qd1xd2 Qc7-f4 Qd2xf4+ g5xf4 Nb5-c3 Nb6-d7 Rc1-c2 Kb8-c7 b2-b3 Nd7-e5 b3-b4 a7-a6 Nc3-a4 f4-f3 Be2xf3 Rh8xh3 Re1-e3 Nf6-g4 Re3-b3 Ne5xf3+ Rb3xf3 Rh3xf3 g2xf3 Ng4-e5 Kg1-g2 b7-b5 Na4-c5 a6-a5 Rc2-c3 a5xb4 a3xb4 Be7xc5 Rc3xc5 Ne5-d3 Rc5-c3 Nd3xb4 Rc3-a3
36/55+ 36:11 5,350,747,191 2,465,053 +0.23 Qd1-b3 g5-g4 e3-e4 Rd8-d7
36/55+ 42:06 6,301,367,591 2,494,652 +0.29 Qd1-b3 Nb6-d5 Be2-f3 Be7-d6 Rc1-c2 Bd6-h2+ Kg1-f1 Bh2-d6 e3-e4 Nd5xc3 b2xc3 g5-g4 h3xg4 Rh8-h1+ Kf1-e2 Rh1xe1+ Bd2xe1 Bd6-f4 e4-e5 Nf6-d7
36/58 1:14:02 10,952,736,084 2,465,730 +0.23 Qd1-b3 g5-g4 e3-e4 Kb8-a8 Bd2-e3 Rd8-c8 Re1-d1 Nb6-d7 a3-a4 g4xh3 g2-g3 Qc7-b6 Qb3xb6 Nd7xb6 Kg1-h2 Be7-b4 Nc3-a2 Bb4-e7 a4-a5 Nb6-d7 f2-f3 a7-a6 Na2-c3

Latest Stockfish (16 threads, 8 GB hash) is faster, but still takes more than 3 minutes. Also chooses b4 rather than Qb3.

Code: Select all

info depth 33 seldepth 56 score cp 10 nodes 1141203347 nps 16911476 time 67481 multipv 1 pv e3e4 d8d4 c3b5 d4d2 d1d2 c7f4 d2f4 g5f4 b5c3 b6d7 c1c2 a7a5 e1d1 d7e5 f2f3 g7g5 c3a2 f6d7 b2b4 a5b4 a3b4 b7b5 c2c3 b8b7 c3a3 d7f6 g1f2 b7b6 f2e1 b6b7 e1f1 b7b6 f1f2 c6c5 a3a5 c5c4 a2c3 e7c5 f2f1
info depth 34 currmove e3e4 currmovenumber 1
info depth 34 currmove e2d3 currmovenumber 6
info depth 34 seldepth 56 score cp 16 lowerbound nodes 4136208439 nps 18412117 time 224646 multipv 1 pv b2b4 a7a6 b4b5 e7a3 c1c2 g5g4 b5a6
info depth 34 currmove b2b4 currmovenumber 1
info depth 34 currmove e3e4 currmovenumber 2
info depth 34 seldepth 56 score cp 20 nodes 4887853008 nps 18427482 time 265248 multipv 1 pv b2b4 e6e5 c1c2 h8h4 c3b5 c7d7 d4e5 c6b5 e5f6 e7f6 d2c3 d7d1 e1d1 d8d1 e2d1 f6c3 c2c3 h4c4 c3d3 b8c7 g1h2 c4c1 h2g3 g7g6 d1g4 f7f5 g4f3 a7a6 e3e4 c1c2 f3d1 c2a2 e4f5 g6f5 f2f4 g5f4 g3f4 b6c4 d1f3
info depth 35 currmove b2b4 currmovenumber 1
info depth 35 seldepth 56 score cp 26 lowerbound nodes 6792732009 nps 18446380 time 368242 multipv 1 pv b2b4 e6e5 c1c2 h8h4 e2f1 e5d4 e3d4 h4d4 d1c1 e7d6 c3b5
info depth 35 currmove b2b4 currmovenumber 1
info depth 35 seldepth 56 score cp 32 lowerbound nodes 8706311875 nps 18323214 time 475152 multipv 1 pv b2b4 a7a6 b4b5 e7a3 c1c2 g5g4
info depth 35 currmove b2b4 currmovenumber 1
info depth 35 currmove e3e4 currmovenumber 2
info depth 35 seldepth 56 score cp 23 nodes 13057785965 nps 18583178 time 702667 multipv 1 pv b2b4 g7g6 a3a4 g5g4 a4a5 b6d5 c3d5 e6d5 e2g4 a7a6 g4f3 c7d6 d1b3 h8h4 b4b5 c6b5 c1c5 h4h8 d2b4 d6e6 c5c2 e7b4 b3b4 d8c8 c2c5
info depth 36 currmove b2b4 currmovenumber 1
info depth 36 seldepth 56 score cp 29 lowerbound nodes 13474754925 nps 18588484 time 724898 multipv 1 pv b2b4 g7g6 a3a4 g5g4 a4a5 b6d5
info depth 36 currmove b2b4 currmovenumber 1
info depth 36 seldepth 56 score cp 36 lowerbound nodes 13938459658 nps 18566764 time 750721 multipv 1 pv b2b4 g5g4 e2g4 b6c4 g4e2 c4d2 d1d2 g7g5 b4b5 g5g4 d2b2 g4h3 g2g3 c6b5 c3d5 c7d6 d5e7
info depth 36 currmove b2b4 currmovenumber 1
info depth 36 currmove e3e4 currmovenumber 2
info depth 36 seldepth 56 score cp 36 nodes 18500753671 nps 18709779 time 988828 multipv 1 pv b2b4 c7d7 c1c2 h8h4 e2f1 f6d5 g2g3 d5c3 d2c3 h4h8 d1e2 f7f5 c3d2 d8f8 f1g2 f8c8 e1b1 a7a6 c2a2 b6a4 b1c1 e7f6 a2c2 a4b6

I would have chosen b4 myself.

I would be surprised if someone that has just learned chess and has an ELO rating which is 1000 lower than mine, starts commenting my chess games.

Maybe the only way to comment on a game of Stockfish is to use Stockfish or another top chess program with much longer time control.

I wasn't aware that Ludmyl is 1000 Elo weaker than you

I was referencing my niece when she was ten years old and wanted to learn chess for there was a chess tournament at school.

[Dr.Wael Deeb]

The problem for stockfish is not to find Rxd4(stockfish can find it very fast but to avoid e4

stockfish needs a lot of time to avoid komodo's e3-e4

[D]1k1r3r/ppq1bpp1/1np1pn2/6p1/3P4/P1N1P2P/1P1BBPP1/2RQR1K1 w - - 0 17

Stockfish_14051921_x64:

35/52 28:13 4,201,289,021 2,481,325 +0.17 e3-e4 Rd8xd4 Nc3-b5 Rd4xd2 Qd1xd2 Qc7-f4 Qd2xf4+ g5xf4 Nb5-c3 Nb6-d7 Rc1-c2 Kb8-c7 b2-b3 Nd7-e5 b3-b4 a7-a6 Nc3-a4 f4-f3 Be2xf3 Rh8xh3 Re1-e3 Nf6-g4 Re3-b3 Ne5xf3+ Rb3xf3 Rh3xf3 g2xf3 Ng4-e5 Kg1-g2 b7-b5 Na4-c5 a6-a5 Rc2-c3 a5xb4 a3xb4 Be7xc5 Rc3xc5 Ne5-d3 Rc5-c3 Nd3xb4 Rc3-a3
36/55+ 36:11 5,350,747,191 2,465,053 +0.23 Qd1-b3 g5-g4 e3-e4 Rd8-d7
36/55+ 42:06 6,301,367,591 2,494,652 +0.29 Qd1-b3 Nb6-d5 Be2-f3 Be7-d6 Rc1-c2 Bd6-h2+ Kg1-f1 Bh2-d6 e3-e4 Nd5xc3 b2xc3 g5-g4 h3xg4 Rh8-h1+ Kf1-e2 Rh1xe1+ Bd2xe1 Bd6-f4 e4-e5 Nf6-d7
36/58 1:14:02 10,952,736,084 2,465,730 +0.23 Qd1-b3 g5-g4 e3-e4 Kb8-a8 Bd2-e3 Rd8-c8 Re1-d1 Nb6-d7 a3-a4 g4xh3 g2-g3 Qc7-b6 Qb3xb6 Nd7xb6 Kg1-h2 Be7-b4 Nc3-a2 Bb4-e7 a4-a5 Nb6-d7 f2-f3 a7-a6 Na2-c3

Latest Stockfish (16 threads, 8 GB hash) is faster, but still takes more than 3 minutes. Also chooses b4 rather than Qb3.

Code: Select all

info depth 33 seldepth 56 score cp 10 nodes 1141203347 nps 16911476 time 67481 multipv 1 pv e3e4 d8d4 c3b5 d4d2 d1d2 c7f4 d2f4 g5f4 b5c3 b6d7 c1c2 a7a5 e1d1 d7e5 f2f3 g7g5 c3a2 f6d7 b2b4 a5b4 a3b4 b7b5 c2c3 b8b7 c3a3 d7f6 g1f2 b7b6 f2e1 b6b7 e1f1 b7b6 f1f2 c6c5 a3a5 c5c4 a2c3 e7c5 f2f1
info depth 34 currmove e3e4 currmovenumber 1
info depth 34 currmove e2d3 currmovenumber 6
info depth 34 seldepth 56 score cp 16 lowerbound nodes 4136208439 nps 18412117 time 224646 multipv 1 pv b2b4 a7a6 b4b5 e7a3 c1c2 g5g4 b5a6
info depth 34 currmove b2b4 currmovenumber 1
info depth 34 currmove e3e4 currmovenumber 2
info depth 34 seldepth 56 score cp 20 nodes 4887853008 nps 18427482 time 265248 multipv 1 pv b2b4 e6e5 c1c2 h8h4 c3b5 c7d7 d4e5 c6b5 e5f6 e7f6 d2c3 d7d1 e1d1 d8d1 e2d1 f6c3 c2c3 h4c4 c3d3 b8c7 g1h2 c4c1 h2g3 g7g6 d1g4 f7f5 g4f3 a7a6 e3e4 c1c2 f3d1 c2a2 e4f5 g6f5 f2f4 g5f4 g3f4 b6c4 d1f3
info depth 35 currmove b2b4 currmovenumber 1
info depth 35 seldepth 56 score cp 26 lowerbound nodes 6792732009 nps 18446380 time 368242 multipv 1 pv b2b4 e6e5 c1c2 h8h4 e2f1 e5d4 e3d4 h4d4 d1c1 e7d6 c3b5
info depth 35 currmove b2b4 currmovenumber 1
info depth 35 seldepth 56 score cp 32 lowerbound nodes 8706311875 nps 18323214 time 475152 multipv 1 pv b2b4 a7a6 b4b5 e7a3 c1c2 g5g4
info depth 35 currmove b2b4 currmovenumber 1
info depth 35 currmove e3e4 currmovenumber 2
info depth 35 seldepth 56 score cp 23 nodes 13057785965 nps 18583178 time 702667 multipv 1 pv b2b4 g7g6 a3a4 g5g4 a4a5 b6d5 c3d5 e6d5 e2g4 a7a6 g4f3 c7d6 d1b3 h8h4 b4b5 c6b5 c1c5 h4h8 d2b4 d6e6 c5c2 e7b4 b3b4 d8c8 c2c5
info depth 36 currmove b2b4 currmovenumber 1
info depth 36 seldepth 56 score cp 29 lowerbound nodes 13474754925 nps 18588484 time 724898 multipv 1 pv b2b4 g7g6 a3a4 g5g4 a4a5 b6d5
info depth 36 currmove b2b4 currmovenumber 1
info depth 36 seldepth 56 score cp 36 lowerbound nodes 13938459658 nps 18566764 time 750721 multipv 1 pv b2b4 g5g4 e2g4 b6c4 g4e2 c4d2 d1d2 g7g5 b4b5 g5g4 d2b2 g4h3 g2g3 c6b5 c3d5 c7d6 d5e7
info depth 36 currmove b2b4 currmovenumber 1
info depth 36 currmove e3e4 currmovenumber 2
info depth 36 seldepth 56 score cp 36 nodes 18500753671 nps 18709779 time 988828 multipv 1 pv b2b4 c7d7 c1c2 h8h4 e2f1 f6d5 g2g3 d5c3 d2c3 h4h8 d1e2 f7f5 c3d2 d8f8 f1g2 f8c8 e1b1 a7a6 c2a2 b6a4 b1c1 e7f6 a2c2 a4b6

I would have chosen b4 myself.

I would be surprised if someone that has just learned chess and has an ELO rating which is 1000 lower than mine, starts commenting my chess games.

Maybe the only way to comment on a game of Stockfish is to use Stockfish or another top chess program with much longer time control.

I wasn't aware that Ludmyl is 1000 Elo weaker than you

I was referencing my niece when she was ten years old and wanted to learn chess for there was a chess tournament at school.

[Rebel]

Nice strangulation game in the style of Chigorin.

[JJJ]

Super Super final will be the winner vs itself. Probably Stockfish vs Stockfish. 666 games will be played.