What engine breaks even with GMs in blitz?

[Vinvin]

Chess Genius made some great performance at the Intel Grand Prix in 1994
https://www.chessprogramming.org/Chess_ ... lGrandPrix

My view is that ChessGenius on a Pentium 90 (released in the end of 1994) was already a bit stronger than 2600 in blitz.
May be ChessGenius on a 486-33 was already 2600 in blitz. It's difficult to know without playing 50 games vs GMs. More, at this time, the GMs were discovering the computer (and the anti-computer) style.

Aubervilliers Open Rapid in 1993, Chess Genius made a 2530 perf : Lost against Spiridonov (2420), but beat GM Bachar Kouatly (2505) and IM Lupu (2435). -> http://www.blitzchess.fr/fr/biblio/lapu ... ce-n3.html

Rebel vs Yusupov, June 1997, 5 minutes, 7 games, Rebel won 6-1.
http://www.rebel.nl/june2.htm

Rebel vs Anand, July 1998, 5m+5s, 4 games, Rebel won 3-1.
http://www.rebel.nl/july21.htm

[jdart]

That's pretty impressive given what hardware there was in that era.

--Jon

[lkaufman]

Chess Genius made some great performance at the Intel Grand Prix in 1994
https://www.chessprogramming.org/Chess_ ... lGrandPrix

My view is that ChessGenius on a Pentium 90 (released in the end of 1994) was already a bit stronger than 2600 in blitz.
May be ChessGenius on a 486-33 was already 2600 in blitz. It's difficult to know without playing 50 games vs GMs. More, at this time, the GMs were discovering the computer (and the anti-computer) style.

Aubervilliers Open Rapid in 1993, Chess Genius made a 2530 perf : Lost against Spiridonov (2420), but beat GM Bachar Kouatly (2505) and IM Lupu (2435). -> http://www.blitzchess.fr/fr/biblio/lapu ... ce-n3.html

Rebel vs Yusupov, June 1997, 5 minutes, 7 games, Rebel won 6-1.
http://www.rebel.nl/june2.htm

Rebel vs Anand, July 1998, 5m+5s, 4 games, Rebel won 3-1.
http://www.rebel.nl/july21.htm

That's consistent with my estimated 2550 result on 486-25 for Rexchess around 1990. The Rebel vs Anand match is useful here; Rebel Century is about 2550 on CCRL blitz, so 1998 version being two years earlier maybe 2450, but that's on hardware about a dozen years later than 1998. So perhaps 2300 CCRL for the engine that beat Anand even at slow blitz! Top rated Giraffe (Sep 8, 2015) at 2410 on CCRL list should be much stronger, and running on 5 GHz i7 with a good opening book should make it a clear favorite in 3' + 2" even against Carlsen. Currently in my test it (playing 3' + 2") is beating Komodo 12.3 Level 19 by 97 elo after 44 games. Level 19 is basically ten play with some small randomness. Based on games vs Nakamura and MVL (world number 2 and 3 in blitz) I would think that level 19 is a good match for them at 3' + 2", so it seems reasonable that Giraffe should be favored vs. Carlsen. I don't have any special tie to Giraffe, I just think it would be a good proxy for Carlsen in blitz tests since being self-taught it won't have biases that one would expect if I picked a closely rated A/B engine. Comments, anyone?

[mwyoung]

Chess Genius made some great performance at the Intel Grand Prix in 1994
https://www.chessprogramming.org/Chess_ ... lGrandPrix

My view is that ChessGenius on a Pentium 90 (released in the end of 1994) was already a bit stronger than 2600 in blitz.
May be ChessGenius on a 486-33 was already 2600 in blitz. It's difficult to know without playing 50 games vs GMs. More, at this time, the GMs were discovering the computer (and the anti-computer) style.

Aubervilliers Open Rapid in 1993, Chess Genius made a 2530 perf : Lost against Spiridonov (2420), but beat GM Bachar Kouatly (2505) and IM Lupu (2435). -> http://www.blitzchess.fr/fr/biblio/lapu ... ce-n3.html

Rebel vs Yusupov, June 1997, 5 minutes, 7 games, Rebel won 6-1.
http://www.rebel.nl/june2.htm

Rebel vs Anand, July 1998, 5m+5s, 4 games, Rebel won 3-1.
http://www.rebel.nl/july21.htm

This is the ChessGenius game I remember from 1994. What a great program!! You are bring back memories...

[pgn] [[Event "Intel Chess Grand Prix"]
[Site "London ENG"]
[Date "1994.??.??"]
[EventDate "?"]
[Round "1"]
[Result "0-1"]
[White "Garry Kasparov"]
[Black "Chess Genius (Computer)"]
[ECO "D11"]
[WhiteElo "?"]
[BlackElo "?"]
[PlyCount "120"]

1.c4 c6 2.d4 d5 3.Nf3 Nf6 4.Qc2 dxc4 5.Qxc4 Bf5 6.Nc3 Nbd7
7.g3 e6 8.Bg2 Be7 9.O-O O-O 10.e3 Ne4 11.Qe2 Qb6 12.Rd1 Rad8
13.Ne1 Ndf6 14.Nxe4 Nxe4 15.f3 Nd6 16.a4 Qb3 17.e4 Bg6 18.Rd3
Qb4 19.b3 Nc8 20.Nc2 Qb6 21.Bf4 c5 22.Be3 cxd4 23.Nxd4 Bc5
24.Rad1 e5 25.Nc2 Rxd3 26.Qxd3 Ne7 27.b4 Bxe3+ 28.Qxe3 Rd8
29.Rxd8+ Qxd8 30.Bf1 b6 31.Qc3 f6 32.Bc4+ Bf7 33.Ne3 Qd4
34.Bxf7+ Kxf7 35.Qb3+ Kf8 36.Kg2 Qd2+ 37.Kh3 Qe2 38.Ng2 h5
39.Qe3 Qc4 40.Qd2 Qe6+ 41.g4 hxg4 42.fxg4 Qc4 43.Qe1 Qb3+
44.Ne3 Qd3 45.Kg3 Qxe4 46.Qd2 Qf4+ 47.Kg2 Qd4 48.Qxd4 exd4
49.Nc4 Nc6 50.b5 Ne5 51.Nd6 d3 52.Kf2 Nxg4+ 53.Ke1 Nxh2 54.Kd2
Nf3+ 55.Kxd3 Ke7 56.Nf5+ Kf7 57.Ke4 Nd2+ 58.Kd5 g5 59.Nd6+ Kg6
60.Kd4 Nb3+ 0-1/pgn][/pgn]

[lkaufman]

I ran a twenty game blitz (3' + 2") knight odds match (two ply book for variety, both b1 and g1 knights) between Lc0 11248 (on my 2080) vs. Giraffe (best version, on my 5 GHz i7 laptop), with Giraffe as a proxy for Magnus Carlsen, a good one since it "knows" to simplify when up a piece while some similarly rated A/B engines may not know or appreciate this. "Magnus" won by 12 to 8 (no draws!). So perhaps it's not yet time to bet against the champ if such a match took place, but we're getting close it seems.

[Laskos]

I ran a twenty game blitz (3' + 2") knight odds match (two ply book for variety, both b1 and g1 knights) between Lc0 11248 (on my 2080) vs. Giraffe (best version, on my 5 GHz i7 laptop), with Giraffe as a proxy for Magnus Carlsen, a good one since it "knows" to simplify when up a piece while some similarly rated A/B engines may not know or appreciate this. "Magnus" won by 12 to 8 (no draws!). So perhaps it's not yet time to bet against the champ if such a match took place, but we're getting close it seems.

I am not sure at 3' + 2'' blitz, Lc0 11248 on 2080 is probably close at Knight odds to Carlsen, but my claim here:
http://www.talkchess.com/forum3/viewtop ... =2&t=69956

was that at longer 45' + 15'' a top GM might not win all 10 games out of 10 being a Knight up, more probably 7-9, 1-3 being drawn or even lost by the human. It would be fun to watch such a match, as the top human will be happy having an upper hand most of the times, while enjoying some 1-3 setbacks at Knight odds in 10 games! The prize could be proportional to (Wins - Draws - Losses) or even (Wins - Draws - 2*Losses), to give incentive to the human to not lose any points. Lc0 should be left with some temperature (say 0.5) for the first 4-5 moves, for diversity and for not playing into prepared openings. Or by providing Lc0 with a small, prepared book.

[JJJ]

Maybe for fun a special lc0 should be train to play these handicap match. It would be more useful than an antifish.

[Albert Silver]

In blitz (let's say 3' + 2" or as close to this as possible), the top engines today are far beyond human level. But how far down the list do we have to go to find engines (and specified hardware) that score evenly against GMs, preferably ones with known identities and ratings? I'm sure there is plenty of data to answer this question as countless games have been played online over the years, but does anyone actually have some data, such as "Engine xyz on one thread scored 50% against GMs averaging 2600 FIDE" for example? The question I'd like to answer is: How much would we have to add to CCRL blitz ratings to estimate the FIDE blitz rating of a human GM who would score 50% against it at 3' + 2"?

Hardware is also a big factor. Back in 2001 I gave private computer classes to GM Mecking, whose ICC blitz rating was the no.1. It turned out that he got this rating in great part by beating up on Chess Tiger 13 and Fritz 6 accounts in blitz. I was surprised by this, but saw him play 4-5 games in which he regularly bamboozled them in endgames. However, could he still do this, even with that same skill level, if they ran on a modern machine, even with one core? Maybe, but I am not nearly so confident.

Yes, I'm sure hardware makes a big difference. What do you think the hardware was like that he played against? Also, what time limit? An easy way to raise your blitz rating online is to play something like 12' +4"against engines which is technically "blitz" but of course not really so. Against modern engines it's not easy to even reach an endgame without being obviously lost.

The games were all 5+0. As to the hardware, I would guess whatever was prevalent then. These were all the top scoring engine accounts on ICC back then.

[lkaufman]

Maybe for fun a special lc0 should be train to play these handicap match. It would be more useful than an antifish.

It's a good idea, but training would have to be against either a weaker engine or against itself time-handicapped so that the losing side has a decent chance to win or draw. I was told that the GO engine Leela (not Leela zero) did train on game scores in which top pros gave handicaps to lower ranked pros, and the result is that Leela does play handicap GO pretty well, while Leela zero does not. I think that Lc0 11248 does play handicap chess (at least up to knight odds) pretty well, but since the newer ones are awful at it I suspect that even 11248 doesn't play it nearly as well as it would if trained for it.

[lkaufman]

I ran a twenty game blitz (3' + 2") knight odds match (two ply book for variety, both b1 and g1 knights) between Lc0 11248 (on my 2080) vs. Giraffe (best version, on my 5 GHz i7 laptop), with Giraffe as a proxy for Magnus Carlsen, a good one since it "knows" to simplify when up a piece while some similarly rated A/B engines may not know or appreciate this. "Magnus" won by 12 to 8 (no draws!). So perhaps it's not yet time to bet against the champ if such a match took place, but we're getting close it seems.

I am not sure at 3' + 2'' blitz, Lc0 11248 on 2080 is probably close at Knight odds to Carlsen, but my claim here:
http://www.talkchess.com/forum3/viewtop ... =2&t=69956

was that at longer 45' + 15'' a top GM might not win all 10 games out of 10 being a Knight up, more probably 7-9, 1-3 being drawn or even lost by the human. It would be fun to watch such a match, as the top human will be happy having an upper hand most of the times, while enjoying some 1-3 setbacks at Knight odds in 10 games! The prize could be proportional to (Wins - Draws - Losses) or even (Wins - Draws - 2*Losses), to give incentive to the human to not lose any points. Lc0 should be left with some temperature (say 0.5) for the first 4-5 moves, for diversity and for not playing into prepared openings. Or by providing Lc0 with a small, prepared book.

Since we already know that this same LCO network defeated GM Naroditsky by a wide margin in blitz games averaging around knight odds, and he is probably within a class of Magnus in blitz strength, we already have good reason to think knight odds vs. Carlsen would be close, so this test is mostly just to confirm that we can make meaningful predictions of human results by these simulartions. As for 45' + 15", what CCRL 40/40 rating do you think would come closest to matching Magnus in strength at 45' + 15" ?