Question for Engines testers ?

Discussion of computer chess matches and engine tournaments.

Moderators: hgm, Harvey Williamson, bob

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
Chessqueen
Posts: 154
Joined: Wed Sep 05, 2018 12:16 am
Full name: Nancy M Pichardo

Re: Question for Engines testers ?

Post by Chessqueen » Sat Nov 10, 2018 8:41 pm

Chessqueen wrote:
Fri Nov 09, 2018 5:21 am
tpoppins wrote:
Thu Nov 08, 2018 4:44 pm
Since CCRL has been mentioned...

Four games mimicking CCRL testing conditions:
TC: 40/2 (CCRL's 40/4 equivalent on this hardware)
5-men EGTBs
512 MB hash
Threads=1
Fritz 15 x64 GUI

1) Deep Fritz 14 64-bit (AKA Pandix) 3077 Elo
2) Rybka 4 64-bit 3101 Elo
3) Spike 1.4 Leiden 2921 Elo
4) Fritz 11 2887 Elo


Try testing by replacing Deep Fritz14 with Fritz15 and the rest all the other three programs that lost to Komodo 12.2, Komodo might draw at least to two of these programs; with a knight odd, or this position;
[D]rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/R1BQKBNR w KQkq - 0 1
OR
[D]1rbqkbnr/1ppppppp/8/8/8/8/PPPPPPPP/1NBQKBNR w KQkq - 0 1

Four games mimicking CCRL testing conditions:
TC: 40/2 (CCRL's 40/4 equivalent on this hardware)
5-men EGTBs
512 MB hash
Threads=1
Fritz 15 x64 GUI

1) Fritz 15 64-bit
2) Rybka 4 64-bit 3101 Elo
3) Spike 1.4 Leiden 2921 Elo
4) Fritz 11 2887 Elo

Chessqueen
Posts: 154
Joined: Wed Sep 05, 2018 12:16 am
Full name: Nancy M Pichardo

Re: Question for Engines testers ?

Post by Chessqueen » Sat Nov 10, 2018 9:36 pm

Chessqueen wrote:
Sat Nov 10, 2018 8:41 pm
Chessqueen wrote:
Fri Nov 09, 2018 5:21 am
tpoppins wrote:
Thu Nov 08, 2018 4:44 pm
Since CCRL has been mentioned...

Four games mimicking CCRL testing conditions:
TC: 40/2 (CCRL's 40/4 equivalent on this hardware)
5-men EGTBs
512 MB hash
Threads=1
Fritz 15 x64 GUI

1) Deep Fritz 14 64-bit (AKA Pandix) 3077 Elo
2) Rybka 4 64-bit 3101 Elo
3) Spike 1.4 Leiden 2921 Elo
4) Fritz 11 2887 Elo


Try testing by replacing Deep Fritz14 with Fritz15 and the rest all the other three programs that lost to Komodo 12.2, Komodo might draw at least to two of these programs; with a knight odd, or this position;
[D]rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/R1BQKBNR w KQkq - 0 1

I just realized that the 2nd position is less than 1 point difference from the starting position, therefore Komodo had an easy WIN versus the three weak programs. therefore try the Knight odd and see what happens!
[D]1rbqkbnr/1ppppppp/8/8/8/8/PPPPPPPP/1NBQKBNR w KQkq - 0 1

Four games mimicking CCRL testing conditions:
TC: 40/2 (CCRL's 40/4 equivalent on this hardware)
5-men EGTBs
512 MB hash
Threads=1
Fritz 15 x64 GUI

1) Fritz 15 64-bit
2) Rybka 4 64-bit 3101 Elo
3) Spike 1.4 Leiden 2921 Elo
4) Fritz 11 2887 Elo

lkaufman
Posts: 3197
Joined: Sun Jan 10, 2010 5:15 am
Location: Maryland USA
Contact:

Re: Question for Engines testers ?

Post by lkaufman » Sun Nov 11, 2018 6:28 am

I'll try to summarize answers to your questions here.
1. I played lots of handicap games between earlier versions of Komodo against many weaker (but still Carlsen level or above) engines) at various handicaps and time limits, none with recent Komodo though. I don't have the games any more. In general Komodo could defeat stronger engines at handicaps than it would be able to against a human world champion, probably because the human knows to avoid complications while the engine won't generally do so unless it has a very sophisticated Contempt. But time limit matter a lot, especially for humans but even for engines; the faster the game, the larger the "fair" handicap. Handicaps like the Exchange or two pawns or "f" pawn and two moves were reasonable against such Carlsen + level engines, while knight odds was only appropriate against engines in the GM range below Elite level, talking about blitz or near-blitz level games.
2. In the opening, a knight is worth about four pawns, while the Exchange is a bit less than two pawns. So knight odds is huge, more than double Exchange odds.
3. Earlier versions of Komodo were fairly even giving knight odds to GMs around 2600 FIDE in blitz (3' + 2") games. Maybe current Komodo could give knight odds in blitz to Hou Yifan or a similarly rated GM. But in (fast) standard chess, meaning 45' + 15" increment, older versions of Komodo scored only around 25% giving knight odds to FMs in the 2100 to 2200 FIDE range. Yet the same versions were undefeated (about half draws) giving rook for knight (with both colors) to GMs, including the Nakamura game. Also Komodo was never defeated giving f7 pawn or even f7 pawn plus second White move to GMs at this level, again about half draws. Even giving two pawns (both colors, various pawns) to GMs Komodo has a huge plus score, over 70% in many games at this level. So Komodo can give handicaps evaluated around the two pawn mark to fairly strong GMs at fairly serious time limits, and even to top ten GMs (Nakamura and MVL so far) at levels around the 15 min mark, but knight odds in serious games is out of the question with GMs. I often play Komodo myself with knight odds, and the results depend mostly on the time limit; if it's something like 4' + 2" I tend to lose, but if it's 5' + 5" or slower I become the favorite. Although I'm a GM, I'm 70 years old and I certainly don't play at GM level now, especially in fast games, although I am the top rated U.S. player age 70 or above. My son Raymond, an IM around the same level as I am now, has similar results.
4. Against mid-level GMs at the 45' + 15" tc, I think the fairest handicap now is something like N for pawn (remove b1 and c7,d7, or e7 pawn). Or f7 pawn plus a second non-edge Black pawn. Or rook for knight plus the "a" pawn. Or f7 pawn and three moves (e4,d4, and White to move).
Komodo rules!

Chessqueen
Posts: 154
Joined: Wed Sep 05, 2018 12:16 am
Full name: Nancy M Pichardo

Re: Question for Engines testers ?

Post by Chessqueen » Sun Nov 11, 2018 5:47 pm

lkaufman wrote:
Sun Nov 11, 2018 6:28 am
I'll try to summarize answers to your questions here.
1. I played lots of handicap games between earlier versions of Komodo against many weaker (but still Carlsen level or above) engines) at various handicaps and time limits, none with recent Komodo though. I don't have the games any more. In general Komodo could defeat stronger engines at handicaps than it would be able to against a human world champion, probably because the human knows to avoid complications while the engine won't generally do so unless it has a very sophisticated Contempt. But time limit matter a lot, especially for humans but even for engines; the faster the game, the larger the "fair" handicap. Handicaps like the Exchange or two pawns or "f" pawn and two moves were reasonable against such Carlsen + level engines, while knight odds was only appropriate against engines in the GM range below Elite level, talking about blitz or near-blitz level games.
2. In the opening, a knight is worth about four pawns, while the Exchange is a bit less than two pawns. So knight odds is huge, more than double Exchange odds.
3. Earlier versions of Komodo were fairly even giving knight odds to GMs around 2600 FIDE in blitz (3' + 2") games. Maybe current Komodo could give knight odds in blitz to Hou Yifan or a similarly rated GM. But in (fast) standard chess, meaning 45' + 15" increment, older versions of Komodo scored only around 25% giving knight odds to FMs in the 2100 to 2200 FIDE range. Yet the same versions were undefeated (about half draws) giving rook for knight (with both colors) to GMs, including the Nakamura game. Also Komodo was never defeated giving f7 pawn or even f7 pawn plus second White move to GMs at this level, again about half draws. Even giving two pawns (both colors, various pawns) to GMs Komodo has a huge plus score, over 70% in many games at this level. So Komodo can give handicaps evaluated around the two pawn mark to fairly strong GMs at fairly serious time limits, and even to top ten GMs (Nakamura and MVL so far) at levels around the 15 min mark, but knight odds in serious games is out of the question with GMs. I often play Komodo myself with knight odds, and the results depend mostly on the time limit; if it's something like 4' + 2" I tend to lose, but if it's 5' + 5" or slower I become the favorite. Although I'm a GM, I'm 70 years old and I certainly don't play at GM level now, especially in fast games, although I am the top rated U.S. player age 70 or above. My son Raymond, an IM around the same level as I am now, has similar results.
4. Against mid-level GMs at the 45' + 15" tc, I think the fairest handicap now is something like N for pawn (remove b1 and c7,d7, or e7 pawn). Or f7 pawn plus a second non-edge Black pawn. Or rook for knight plus the "a" pawn. Or f7 pawn and three moves (e4,d4, and White to move).

I see that if the White pieces that has the disadvantage of the Knight odd moves first in the opening, a knight is worth slightly less than 4 pawns :


[Event "Computer chess game"]
[Site "DESKTOP-OFQ3C0P"]
[Date "2018.11.11"]
[Round "?"]
[White "Stockfish_9_x64"]
[Black "Stockfish_9_x64"]
[Result "*"]
[BlackElo "3200"]
[Time "11:22:22"]
[WhiteElo "3200"]
[TimeControl "80/120:0"]
[SetUp "1"]
[FEN "rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/R1BQKBNR w KQkq - 0 1"]
[Termination "unterminated"]
[PlyCount "7"]
[WhiteType "program"]
[BlackType "program"]

1. d4 {(d2-d4 Ng8-f6 Ng1-f3 d7-d5 c2-c3 e7-e6 Bc1-f4 c7-c5 e2-e3 Nb8-c6
Bf1-e2 Nf6-e4 Nf3-d2 c5xd4 Nd2xe4 d5xe4 c3xd4 Bf8-b4+ Ke1-f1 O-O h2-h4
Bc8-d7) -3.36/19 7} Nf6 {(Ng8-f6 Ng1-f3 d7-d5 c2-c3 e7-e6 Bc1-f4 c7-c5
e2-e3 Nb8-c6 Bf1-e2 Nf6-e4 O-O g7-g5 Bf4-g3 Bf8-e7 Be2-d3 Ne4xg3 h2xg3 O-O
Bd3-c2 Bc8-d7) +3.78/17 1} 2. Nf3 {(Ng1-f3 d7-d5 c2-c3 e7-e6 Bc1-f4 c7-c5
e2-e3 Nb8-c6 Bf1-e2 Nf6-h5 Bf4-g5 Qd8-b6 Ra1-b1 h7-h6 Bg5-h4 g7-g5 Nf3-d2
Nh5-g7 Bh4-g3 c5xd4 c3xd4 Ng7-f5 O-O) -3.47/18 4} b6 {(b7-b6 Bc1-g5 Bc8-b7
e2-e3 e7-e6 Bf1-d3 d7-d6 O-O Bf8-e7 Bg5xf6 Be7xf6 c2-c4 Nb8-d7 Qd1-c2 c7-c5
Bd3-e4 Bb7xe4 Qc2xe4 O-O Qe4-g4) +3.89/18 2} 3. Bg5 {(Bc1-g5 Bc8-b7 e2-e3
e7-e6 Bf1-d3 Bf8-e7 c2-c3 h7-h6 Bg5xf6 Be7xf6 e3-e4 d7-d6 Qd1-a4+ Nb8-d7
O-O c7-c5 Ra1-e1 O-O) -3.64/16 1} Bb7 {(Bc8-b7 e2-e3 e7-e6 Bf1-d3 Bf8-e7
O-O O-O Bg5-f4 d7-d6 c2-c4 h7-h6 h2-h3 Nb8-d7 Bf4-g3 c7-c5 a2-a3 Kg8-h8
Bg3-h2 Kh8-g8) +4.07/17 2} 4. Bxf6 {(Bg5xf6 e7xf6 e2-e3 d7-d5 g2-g3 Nb8-d7
Bf1-g2 Bf8-e7 O-O O-O Nf3-d2 f6-f5 c2-c4 Nd7-f6 Ra1-c1 g7-g6 c4xd5 Bb7xd5
Bg2xd5 Nf6xd5 e3-e4 f5xe4 Nd2xe4) -3.72/18 1} *


But if the side with the advantage of the knight odd in this case if the black pieces moves first the advantage is worth slighty more than 4 pawns:

[Event "Computer chess game"]
[Site "DESKTOP-OFQ3C0P"]
[Date "2018.11.11"]
[Round "?"]
[White "Stockfish_9_x64"]
[Black "Stockfish_9_x64"]
[Result "*"]
[BlackElo "3200"]
[Time "11:50:30"]
[WhiteElo "3200"]
[TimeControl "80/120:0"]
[SetUp "1"]
[FEN "rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/R1BQKBNR b KQkq - 0 1"]
[Termination "unterminated"]
[PlyCount "1"]
[WhiteType "program"]
[BlackType "program"]

1. ... e5 {(e7-e5 e2-e3 d7-d5 d2-d4 Nb8-d7 d4xe5 Nd7xe5 Ng1-f3 Bf8-d6
Qd1-d4 Ne5xf3+ g2xf3 Ng8-f6 Rh1-g1 c7-c5 Qd4-h4 O-O Bc1-d2 Bc8-f5 O-O-O
Bf5-g6 Bd2-c3 d5-d4 e3xd4 Nf6-d5 Qh4xd8 Rf8xd8 Bc3-a5 b7-b6 d4xc5 Bd6xc5
Bf1-c4 Nd5-f6) +4.37/27 136} *

lkaufman
Posts: 3197
Joined: Sun Jan 10, 2010 5:15 am
Location: Maryland USA
Contact:

Re: Question for Engines testers ?

Post by lkaufman » Sun Nov 11, 2018 8:53 pm

Right, the first move is worth roughly a quarter pawn, so White a knight down should be about 3 3/4 pawn down and Black a knight down should be about 4 1/4 pawn down. But anyway you are quoting Stockfish evals, which may not be accurate. Stockfish plays very well, but whether its evals are realistic is another question entirely. I think that Shredder may have the most objective evals in general among top engines. In this particular case though I agree with the Stockfish evals. Note that Komodo evals are somewhat closer to zero because a score of +1.00 is more typical of a middlegame pawn than an opening pawn. The numbers may be lower, but on average an extra knight in the opening is worth about four times as much as an average pawn in the opening.
Komodo rules!

Chessqueen
Posts: 154
Joined: Wed Sep 05, 2018 12:16 am
Full name: Nancy M Pichardo

Re: Question for Engines testers ?

Post by Chessqueen » Mon Nov 12, 2018 4:14 am

tpoppins wrote:
Thu Nov 08, 2018 4:44 pm
Since CCRL has been mentioned...

Four games mimicking CCRL testing conditions:
TC: 40/2 (CCRL's 40/4 equivalent on this hardware)
5-men EGTBs
512 MB hash
Threads=1
Fritz 15 x64 GUI

1) Deep Fritz 14 64-bit (AKA Pandix) 3077 Elo
2) Rybka 4 64-bit 3101 Elo
3) Spike 1.4 Leiden 2921 Elo
4) Fritz 11 2887 Elo



Too bad Cute Chess won't accept the FEN, I have hundreds of engines installed in it compared to a couple dozen in Fritz. The earliest Fritz version I got installed is Fritz 11, though.

Probably asking to test Komodo versus those four programs with a knight odd was too much to ask, since the initial position has an equivalent of 4 pawns, but I found out that in this position the difference is just -2.39 pawns where Komodo has a very good chance of at least winning 2 games.
It could also be a good position to test versus GM Nakamura, since he drew with the Knight versus Rook exchange game in 45 minutes with an early version of Komodo 9.xx something.
[D]r1bqkbnr/pppppppp/8/8/8/8/P1PPPPPP/1NBQKBNR w Kk - 0 1

FEN: r1bqkbnr/pppppppp/8/8/8/8/P1PPPPPP/1NBQKBNR w Kk - 0 1

Stockfish_9_x64:
1/1 00:00 20 7k -1.41 e2-e4
2/2 00:00 68 23k -2.26 e2-e4 e7-e6
3/3 00:00 196 65k -1.96 d2-d4 Ng8-f6 Nb1-d2
4/4 00:00 712 142k -2.47 Ng1-f3 Ng8-f6 d2-d4 d7-d5
5/5 00:00 923 154k -2.28 Ng1-f3 Ng8-f6 d2-d4 d7-d5 Qd1-d3
6/6 00:00 2k 206k -2.12 e2-e4 Ng8-f6 e4-e5 Nf6-e4 d2-d4 d7-d5
7/7 00:00 5k 303k -2.30 e2-e3 Ng8-f6 Nb1-c3 d7-d5 d2-d4 e7-e6 Ng1-e2
8/8 00:00 9k 308k -2.40 d2-d4 c7-c6 Nb1-c3 d7-d5 e2-e4 e7-e6 Qd1-d3 Qd8-a5
9/10 00:00 16k 316k -2.40 e2-e4 d7-d5 e4xd5 Qd8xd5 Nb1-c3 Qd5-e6+ Qd1-e2 Ng8-f6 d2-d4 c7-c6
10/15 00:00 32k 286k -1.91 e2-e4 d7-d5 e4xd5 Ng8-f6 c2-c4 e7-e6 Qd1-e2 c7-c6 d5xc6 b7xc6 Nb1-c3
11/16 00:00 72k 337k -2.18 e2-e4 d7-d5 e4-e5 d5-d4 Ng1-f3 Qd8-d5 Bf1-e2 Bc8-g4 Bc1-b2 Bg4xf3 Be2xf3 Qd5xe5+ Qd1-e2
12/18 00:00 149k 382k -2.21 e2-e4 d7-d5 e4xd5 Qd8xd5 Nb1-c3 Qd5-a5 Ng1-f3 e7-e6 Qd1-e2 c7-c6 Qe2-e5 Qa5xe5+ Nf3xe5 Bf8-d6 d2-d4
13/20 00:00 233k 403k -2.10 e2-e4 d7-d5 e4xd5 Qd8xd5 Nb1-c3 Qd5-a5 Bf1-d3 Bc8-d7 Ng1-f3 Ng8-f6 Bc1-b2 e7-e5 O-O Bf8-d6 Qd1-e2 O-O Nc3-e4
14/23 00:00 398k 420k -2.27 e2-e4 d7-d5 e4xd5 Qd8xd5 Nb1-c3 Qd5-c5 Ng1-f3 Ng8-f6 Nc3-b5 Bc8-d7 d2-d4 Qc5-b6 c2-c4 Bd7xb5 c4xb5 e7-e6 Bf1-d3 Qb6-a5+ Bc1-d2 Qa5xa2
15/26 00:01 560k 428k -2.21 e2-e4 e7-e6 e4-e5 d7-d5 Ng1-f3 c7-c5 Bf1-e2 Ng8-e7 c2-c4 Bc8-d7 O-O d5xc4 Be2xc4 Qd8-b6 Nb1-c3
16/26 00:02 949k 441k -2.42 e2-e4 e7-e6 Ng1-f3 d7-d5 Nb1-c3 c7-c6 Bf1-d3 Ng8-f6 O-O Bf8-e7 Bc1-b2 O-O e4-e5 Nf6-d7 Bd3-e2 c6-c5 d2-d4 c5xd4 Nc3-b5 Qd8-a5 Qd1xd4 Qa5xa2
17/25+ 00:03 1,494k 443k -2.35 c2-c4
17/25+ 00:03 1,523k 444k -2.27 c2-c4
17/25+ 00:03 1,583k 444k -2.16 c2-c4
17/25 00:04 1,980k 432k -2.16 c2-c4 Ng8-f6 Ng1-f3 e7-e6 d2-d4 d7-d5 e2-e3 c7-c5 c4xd5 e6xd5 d4xc5 Bf8xc5 Bf1-d3 O-O O-O Rf8-e8 Nb1-c3 Bc8-g4 h2-h3 Bg4-d7 Qd1-b3
18/25- 00:06 2,538k 416k -2.23 c2-c4 e7-e5
18/25- 00:06 2,665k 417k -2.31 c2-c4 e7-e5
18/25+ 00:07 2,999k 417k -2.23 c2-c4
18/27 00:07 3,191k 419k -2.20 c2-c4 e7-e5 Nb1-c3 Ng8-f6 Ng1-f3 Bf8-b4 Bc1-b2 Bb4xc3 Bb2xc3 d7-d6 e2-e3 O-O Bf1-e2 Bc8-f5 d2-d3 Rf8-e8 Nf3-h4 Bf5-d7 O-O c7-c5 Nh4-f3
19/27 00:09 4,015k 424k -2.22 c2-c4 e7-e6 Ng1-f3 d7-d5 e2-e3 Ng8-f6 Bf1-e2 Bf8-d6 O-O O-O Nb1-c3 b7-b6 Nc3-b5 Bc8-b7 Bc1-b2 Bd6-e7 Bb2-e5 d5xc4 Nb5xc7
20/26- 00:11 4,804k 421k -2.30 c2-c4 e7-e6
20/26 00:12 5,301k 415k -2.24 c2-c4 e7-e6 Ng1-f3 d7-d5 e2-e3 Ng8-f6 Bf1-e2 Bf8-d6 O-O b7-b6 Nb1-c3 O-O d2-d4 Bc8-b7 Bc1-b2 a7-a6 c4xd5 e6xd5 Nc3-a4 Nf6-e4 Be2-d3 Qd8-e7 Nf3-e5
21/25 00:15 6,339k 420k -2.24 c2-c4 e7-e6 Ng1-f3 Ng8-f6 e2-e3 d7-d5 Bf1-e2 Bf8-d6 O-O b7-b6 Bc1-b2 O-O d2-d4 Bc8-b7 Nb1-c3 a7-a6 c4xd5 e6xd5 Nc3-a4 Nf6-e4 Be2-d3 Qd8-e7
22/28 00:24 10,188k 417k -2.30 c2-c4 e7-e6 Ng1-f3 Ng8-f6 e2-e3 Bf8-e7 d2-d4 d7-d5 Nb1-d2 O-O Bf1-d3 b7-b6 O-O c7-c5 d4xc5 d5xc4 Nd2xc4 Be7xc5 Bc1-b2 Bc8-b7 Nc4-e5 Bc5-d6
23/36 00:35 14,762k 414k -2.34 c2-c4 e7-e6 Ng1-f3 Ng8-f6 e2-e3 Bf8-e7 d2-d4 b7-b6 Bf1-d3 Bc8-b7 O-O c7-c5 d4xc5 Be7xc5 Nb1-d2 O-O Bc1-b2 h7-h6 Qd1-b3 Ra8-c8 h2-h3 d7-d5 Rf1-d1 Bc5-e7 c4xd5 Bb7xd5
24/31 00:47 19,500k 411k -2.32 c2-c4 e7-e6 Ng1-f3 Ng8-f6 Nb1-c3 Bf8-b4 Bc1-b2 O-O e2-e3 d7-d5 Bf1-e2 b7-b6 O-O Bc8-b7 c4xd5 e6xd5 h2-h3 c7-c5 a2-a3 Bb4-a5 Qd1-c2 Nf6-e4 Rf1-b1 Ba5xc3
25/31 01:02 25,382k 406k -2.32 c2-c4 e7-e6 Ng1-f3 Ng8-f6 Nb1-c3 d7-d5 e2-e3 Bf8-e7 Bf1-e2 O-O O-O b7-b6 Bc1-b2 Bc8-b7 Qd1-c2 c7-c5 c4xd5 e6xd5 d2-d4 Ra8-c8 Qc2-a4 a7-a5 d4xc5 Be7xc5 Rf1-d1 Bc5-b4
26/30- 01:16 31,227k 406k -2.39 c2-c4 e7-e6
26/33- 01:22 33,782k 408k -2.47 c2-c4 e7-e6
26/34+ 01:30 36,706k 405k -2.39 c2-c4
26/34 01:32 37,335k 405k -2.39 c2-c4

lkaufman
Posts: 3197
Joined: Sun Jan 10, 2010 5:15 am
Location: Maryland USA
Contact:

Re: Question for Engines testers ?

Post by lkaufman » Mon Nov 12, 2018 6:26 am

It makes more sense to remove a2 pawn rather than b2 pawn when also giving Exchange odds, because the a2 pawn removal doesn't affect any of the pieces' legal moves. We played this way twice so far, beating IM Danny Rensch in a 45' + 15" game but losing to super-GM MVL (Maxime-Vachier LaGrave) at 15' + 2". The second game was with Komodo MCTS. In general, the challenge in setting up handicap matches is to choose setups that are as similar as possible to the normal one while giving the weaker player a fair chance.
Komodo rules!

Post Reply