Komodo Dragon 2.6.1 released

[lkaufman]

Cicada still play weak moves at T/C game in 1 minute but look at the result. Cicada which is rated only 1537 by CCRL can beat 75% of human Fide rated below 1600 at T/C of 1 minute per game.

[pgn][Event "Computer chess game"]
[Site "MININT-UB2PIMJ"]
[Date "2022.01.11"]
[Round "?"]
[White "Dragon-2.6.1-64bit-avx2"]
[Black "Cicada"]
[Result "0-1"]
[BlackElo "1537"]
[ECO "D85"]
[Opening "Grünfeld"]
[Time "18:43:13"]
[Variation "Modern Exchange, 8.Bb5+"]
[WhiteElo "1600"]
[TimeControl "60+0"]
[Termination "normal"]
[PlyCount "70"]
[WhiteType "program"]
[BlackType "program"]



1. d4 Nf6 2. c4 g6 3. Nf3 Bg7 4. Nc3 d5 5. cxd5 Nxd5 6. e4 Nxc3 7. bxc3 c5
8. Bb5+ Bd7 9. Bxd7+ Nxd7 10. O-O e5 11. d5 f5 12. Ng5 Bf6 13. Ne6 Qa5 14.
Rb1 Rb8 15. exf5 Qxa2 16. Be3 gxf5 17. Qh5+ Ke7 18. Qxf5 Qxd5 19. Nc7 Qc6
20. Nb5 a6 21. Na7 Qc7 22. Rfd1 b6 23. Bg5 Bxg5 24. Qxg5+ Kf7 25. Qh5+ Kf8
26. Qf5+ Ke7 27. Qg4 Rbg8 28. Qh4+ Nf6 29. f4 exf4 30. c4 f3 31. Re1+ Kf7
32. g3 Rg4 33. Qh6 Rxg3+ 34. hxg3 Qxg3+ 35. Kf1 Qg2# 0-1[/pgn]

Where did you get that 75% figure? I would expect that an engine with a ccrl blitz rating of 1537 would score something like 97% against humans under 1600 FIDE playing one minute chess. Or did you perhaps mean that Cicada playing game/one minute would score 75% against humans under 1600 playing at some reasonable time control like Rapid? Players under 1600 can't play game in one minute without giving away pieces left and right or flagging. I would think that even a 2200 player would have trouble against such an engine at game in one minute.

[Cornfed]

Just an observation. I don't know anything about Cicada, but 12...Ke7?? - no human would (and no computer should) ever think to make that kind of move (or likely several others in the game, but that's the obvious one), even at 1550 or 1600 elo...or 1200 ( 4...c6. 5...c5. maybe...). But again, this must be the programmers burden when trying to simulate a rating level for 'bad human play'. How the heck does one realistically it? Missing some pretty obvious, sure, poor positional play, sure...but moves like 12...Ke7? I'm not even sure one should test against an engine that does that.

You forgot that Cicada only made that horrible move on T/C of 30 seconds for the entire game, if you allow Cicada at least 2 minutes for the entire game it would NOT make that horrible move and it would constantly beat Komodo Dragon 2.6.1 set at CI Elo of 1600. Try it yourself, and match them.

No, I didn't forget...but seriously, no engine even looking 1 ply deep should offer up a move like 12...Ke7, let alone a human remotely close to that rating.
And for the 2nd time, I do not own Dragon 2.6.1 so there is no need to continue to ask me to test with it.

It does seem like there are better engines to test Dragon against at this level.

I'm just fascinated by how Larry or anyone is able to 'dumb down' a really good engine to play at different levels. I mean there is SO MUCH which can factor into anyone's rating. One pretty stable 1800 player might be so much better in tactical play than another which might be much weaker but have much better positional instincts.

It almost seems like the best one could reasonably do would be to aim for maybe rough 200 rating point ranges on the lower end to 'maybe' 100 point ranges on the top end.

[lkaufman]

Just an observation. I don't know anything about Cicada, but 12...Ke7?? - no human would (and no computer should) ever think to make that kind of move (or likely several others in the game, but that's the obvious one), even at 1550 or 1600 elo...or 1200 ( 4...c6. 5...c5. maybe...). But again, this must be the programmers burden when trying to simulate a rating level for 'bad human play'. How the heck does one realistically it? Missing some pretty obvious, sure, poor positional play, sure...but moves like 12...Ke7? I'm not even sure one should test against an engine that does that.

You forgot that Cicada only made that horrible move on T/C of 30 seconds for the entire game, if you allow Cicada at least 2 minutes for the entire game it would NOT make that horrible move and it would constantly beat Komodo Dragon 2.6.1 set at CI Elo of 1600. Try it yourself, and match them.

No, I didn't forget...but seriously, no engine even looking 1 ply deep should offer up a move like 12...Ke7, let alone a human remotely close to that rating.
And for the 2nd time, I do not own Dragon 2.6.1 so there is no need to continue to ask me to test with it.

It does seem like there are better engines to test Dragon against at this level.

I'm just fascinated by how Larry or anyone is able to 'dumb down' a really good engine to play at different levels. I mean there is SO MUCH which can factor into anyone's rating. One pretty stable 1800 player might be so much better in tactical play than another which might be much weaker but have much better positional instincts.

It almost seems like the best one could reasonably do would be to aim for maybe rough 200 rating point ranges on the lower end to 'maybe' 100 point ranges on the top end.

No engine is going to play quite like a human rated 1800, some just come much closer than others. The rating is supposed to represent the rating that Dragon set to that level would actually get against a randomly selected group of humans with that FIDE rating playing 15' + 10" with some prize money on the line. That doesn't mean that it would be evenly matched with every 1800 human player (even if I got the rating exactly right), some would be better at beating engines than others. I do believe that Dragon set to 1800 elo will play more like a human of that rating than any normal engine of the same strength, because the normal engine will play much worse positionally and much better tactically than 1800, whereas Dragon set to 1800 should play roughly as well on average both positionally and tactically as a human of that rating playing Rapid; at least it should be much closer to human in that respect than the normal weak engine. If it turns out that I got it wrong, for example that Dragon is 2000 positionally but 1600 tactically (or vice-versa), I just have to change some parameters to correct it in a future release. The hard part is measuring its relative abilities in these categories, which are somewhat subjective in any case.

[Modern Times]

I'm happy to play an FRC match at a given Elo level, if that is useful. If so, let me know what level to choose - making sure there are enough opponents around that range.

[Chessqueen]

I'm happy to play an FRC match at a given Elo level, if that is useful. If so, let me know what level to choose - making sure there are enough opponents around that range.

It all depends what is your Rating to be estimated, or if you have an established rating, then you can set the UCI rating in Komodo dragon to that specific rating and play some FRC. My bet is on Komodo Dragon 2.6.1 for two reasons. 1st Most humans are unfamiliar playing FRC positions since the pattern recognition will never present itself in FRC as is does in Standard chess.

Note: Are you ready to accept the challenge and be the first non GM to play versus Komodo Dragon 2.6.1 taking lets say Rook Odds at full strength or playing FRC at a given predetermined Elo Setting ? It would be interesting to see future match of Non GM.

[Modern Times]

I'm happy to play an FRC match at a given Elo level, if that is useful. If so, let me know what level to choose - making sure there are enough opponents around that range.

I was asking this of Larry.

[Chessqueen]

I'm happy to play an FRC match at a given Elo level, if that is useful. If so, let me know what level to choose - making sure there are enough opponents around that range.

I was asking this of Larry.

You should have asked Mr. Kaufman two weeks before he asked GM Finegold to play versus Komodo Dragon 2.6.1, since now you will have to wait at least another 3 months. But I do consider that testing Komodo below GM strength is important even if the majority of the people or members of this forum only wants to see matches versus GMs.

Note: As far as you planning to test Komodo Dragon 2.6.1 against different human Fide Ratings, I also consider the range between 1300 to 1700 to be critical, but only at T/C below 5 minutes since above that T/C those low rated humans have too much time to figure below 2 ply or plies tactics.

[lkaufman]

I'm happy to play an FRC match at a given Elo level, if that is useful. If so, let me know what level to choose - making sure there are enough opponents around that range.

I was asking this of Larry.

It's not clear from your question whether you mean human opponents or engine opponents. If human, I would expect that the Dragon levels would come out ahead even in Rapid at FRC due to human unfamiliarity. If engine, then I suppose blitz time control is appropriate if you use CCRL FRC blitz ratings to set the elo on Dragon, at least in the amateur rating range. The elo range for Dragon simulates the much more compressed human elo range rather than the expanded engine elo ranges, so if you play engines rated below 1500 or so the normal engines will crush dragon under these conditions, whereas if you play engines rated 2700 or so dragon should win. Maybe somewhere in the ballpark of 2400 or so it would be an even match. I suppose the most useful rating to check on Dragon would be 1600, both because it is somewhere around average for our customers (I imagine) and because 1600 just does a one ply search with some eval randomization thrown in. But fair opponents would be much lower than 1600 on CCRL, and such weak engines might not support FRC.

[Modern Times]

Yes sorry I meant an engine FRC match. I'll take a look at the list and see what could be interesting to try.

[JohnW]

Something broke it in the Fritz 17 gui. The "UCI Elo" option doesn't display in the engine parameters screen like it does in version 2.6.
I registered 2.6.1 in Cute Chess and I can see the option there but not in Fritz 17.