Reckless 0.10.0 in test suites

[Jouni]

300 difficult positions:

Reckless 0.10.0__240
Reckless 0.9.0__221
Stockfish 18__208
PlentyChess 7.0.37__185

[Damir]

Those test suites are of no importance. If Reckless is so much better than Stockfish, why than is it not leading in TCEC? It can't find the best moves, while Stockfish has no problems to find them.

[reflectionofpower]

Good point

[Frank Quisinsky]

Hi Damir,

In most cases, test suites or test positions are designed to identify the best move. In this sense, there is usually a tactical twist hidden behind a test position. And this is precisely where the problem lies with Stockfish. Broadly speaking, this also reflects the level of aggressiveness in the early stages of a game. Here, too, Stockfish does not play as aggressively as other programs. For me, all of this fits into a picture that is formed from such small pieces of the puzzle. It may be that, overall, Stockfish plays the strongest chess, but the “wow” factor during a game falls short compared to many of the other programs. The mere fact that the average number of moves is significantly higher than in PlentyChess, Reckless, and Obsidian is another clue.

The question is what we await from a number 1.
Such a chess Karpow is playing or such a chess Ivantchuk are able to play in his best years. For me Ivantschuk is the greatest chess player in history, he plays the most impressive games. Unfortunately, often terrible msitakes. But who cares about the mistakes these days? The highlights are more interesting.

If Karpow ... Stockfish does everything right.

Best
Frank

[peter]

6 threads of the 16x4.3GHz CPU, RTX 5070 GPU, the known 325 positions with 30"/pos. and EloStatTS (Frank Schubert):

Code: Select all

    Program                                    Elo   +/-  Matches  Score   Av.Op.   S.Pos.   MST1    MST2   RIndex
  
  3 Reckless0.10.0-6t-MuPV4                  : 3547    2  21116    56.9 %   3499   266/325    4.3s    9.0s   0.67
  
  7 Reckless0.9.0-6t-MuPV4                   : 3537    2  20862    55.5 %   3499   260/325    4.7s    9.8s   0.64
 
 27 Lc0v0.32.1-6147500PT-MuPV4               : 3521    2  20733    53.0 %   3500   237/325    4.1s   11.1s   0.59
 
 33 Stockfish-260307-6t-MuPV4                : 3517    2  19653    52.5 %   3499   234/325    4.3s   11.5s   0.62
 
 35 Stockfish18-6t-MuPV4                     : 3517    2  19677    52.5 %   3499   232/325    4.2s   11.6s   0.64
 
 41 Lc0v0.32.1-6147500PT-MuPV1               : 3510    2  20296    51.4 %   3500   224/325    3.8s   11.9s   0.56
 42 PlentyChess7.0.37-6t-MuPV4               : 3510    2  20266    51.4 %   3500   228/325    4.6s   12.1s   0.56
 
 44 Stockfish17.1-6t-MuPV4                   : 3502    2  19563    50.3 %   3500   227/325    5.2s   12.7s   0.54
 45 Reckless0.10.0-6t-MuPV1                  : 3495    2  19772    49.1 %   3501   222/325    5.2s   13.1s   0.49
 
 50 Reckless0.9.0.3-6t-MuPV1                 : 3491    2  19583    48.6 %   3501   213/325    4.8s   13.5s   0.50
 51 Stockfish-260307-6t-MuPV1                : 3490    2  19100    48.4 %   3501   211/325    5.1s   13.8s   0.56
 
 58 Stockfish18-6t-MuPV1                     : 3478    2  18867    46.7 %   3501   204/325    5.4s   14.5s   0.50
 
 64 PlentyChess7.0.37-6t-MuPV1               : 3454    3  18791    43.1 %   3502   174/325    5.0s   16.6s   0.44

 75 Dragon3.3-6t-MCTS-MuPV6                  : 3367    3  17977    31.2 %   3505   117/325    6.8s   21.6s   0.10

MST1  : Mean solution time (solved positions only)
MST2  : Mean solution time (solved and unsolved positions)
RIndex: Score according to solution time ranking for each position

[Jouni]

Thanks Peter for confirming my result. Mate solving is now so bad, that I stopped test after 100 positions .