SF6 moves comparison: Event Vugar Gashimov Mem 2015

[Ferdy]

After round 4, multipv 3 at 10sec/pos.

Code: Select all

Moves matches between Stockfish 6 64 POPCNT top 3 moves and players from
Event 2nd Shamkir Chess 2015, after round 4

Intel(R) Core(TM) i7-2600K CPU @ 3.40GHz
Physical Cores: 4
Hyper-Threading: ON

Engine id name Stockfish 6 64 POPCNT
Hash 128 MB, Threads 2
MultiPV 3, MoveTime 10.0 sec/pos

MRank                   Players  AveDepth  AllPos   M1   M1Rate   M2   M2Rate   M3   M3Rate    W     L     D  Games  Score  SRate

    1                So, Wesley        22     136   84   61.76%   24   17.65%   12    8.82%    3,    0,    1      4    3.5   87.5%
    2             Mamedov, Rauf        23     146   90   61.64%   25   17.12%   14    9.59%    0,    1,    3      4    1.5   37.5%
    3         Kramnik, Vladimir        24     138   85   61.59%   22   15.94%   12    8.70%    1,    0,    3      4    2.5   62.5%
    4           Carlsen, Magnus        23     153   94   61.44%   30   19.61%   12    7.84%    2,    0,    2      4    3.0   75.0%
    5               Giri, Anish        22     153   92   60.13%   27   17.65%    9    5.88%    0,    1,    3      4    1.5   37.5%
    6   Vachier-Lagrave, Maxime        24     144   85   59.03%   22   15.28%   12    8.33%    0,    0,    4      4    2.0   50.0%
    7          Caruana, Fabiano        25     188  104   55.32%   36   19.15%   18    9.57%    0,    1,    3      4    1.5   37.5%
    8        Anand, Viswanathan        24     152   82   53.95%   33   21.71%   13    8.55%    0,    0,    4      4    2.0   50.0%
    9    Mamedyarov, Shakhriyar        24     168   86   51.19%   40   23.81%   16    9.52%    0,    1,    3      4    1.5   37.5%
   10            Adams, Michael        21     117   50   42.74%   28   23.93%    8    6.84%    0,    2,    2      4    1.0   25.0%

Legend:
MRank      - ranking based on most number of top 1 matching engine moves
AveDepth   - average depth reached by the engine in analyzing those positions
M1, M2, M3 - count of top 1, 2, 3 moves of engine that matches the player move.

Notes:
1. Comparison of moves starts when the time consumed including the elapsed time
   of the move of the current position is more than 2 minutes.
2. The engine is run in MultiPV mode producing best 3 moves

[Ferdy]

Time-pressured moves, these are positions with less than
5 minutes left to next period. Here we can see which players
have good time management. There are also players that
have more positions but were able to make some top engine moves.

Code: Select all

Time-pressured moves matches between Stockfish 6 64 POPCNT top 3 moves and players from
Event 2nd Shamkir Chess 2015, after round 4

Intel(R) Core(TM) i7-2600K CPU @ 3.40GHz
Physical Cores: 4
Hyper-Threading: ON

Engine id name Stockfish 6 64 POPCNT
Hash 128 MB, Threads 2
MultiPV 3, MoveTime 30.0 sec/pos

MRank                   Players  AveDepth  AllPos   M1   M1Rate   M2   M2Rate   M3   M3Rate    W     L     D  Games  Score  SRate

    1   Vachier-Lagrave, Maxime        28       1    1  100.00%    0    0.00%    0    0.00%    0,    0,    4      4    2.0   50.0%
    2        Anand, Viswanathan        27       4    3   75.00%    0    0.00%    1   25.00%    0,    0,    4      4    2.0   50.0%
    3          Caruana, Fabiano        30      21   15   71.43%    4   19.05%    1    4.76%    0,    1,    3      4    1.5   37.5%
    4         Kramnik, Vladimir        28      16   11   68.75%    1    6.25%    4   25.00%    1,    0,    3      4    2.5   62.5%
    5                So, Wesley        27      11    7   63.64%    2   18.18%    1    9.09%    3,    0,    1      4    3.5   87.5%
    6    Mamedyarov, Shakhriyar        30       2    1   50.00%    1   50.00%    0    0.00%    0,    1,    3      4    1.5   37.5%
    7            Adams, Michael        26      18    8   44.44%    3   16.67%    2   11.11%    0,    2,    2      4    1.0   25.0%
    8               Giri, Anish        24       5    2   40.00%    2   40.00%    1   20.00%    0,    1,    3      4    1.5   37.5%
    9             Mamedov, Rauf        28      13    5   38.46%    5   38.46%    3   23.08%    0,    1,    3      4    1.5   37.5%
   10           Carlsen, Magnus        28       5    0    0.00%    5  100.00%    0    0.00%    2,    0,    2      4    3.0   75.0%

Legend:
MRank      - ranking based on most number of top 1 matching engine moves
AveDepth   - average depth reached by the engine in analyzing those positions
M1, M2, M3 - count of top 1, 2, 3 moves of engine that matches the player move.

Notes:
1. Comparison of moves starts when the time consumed including the elapsed time
   of the move of the current position is more than 2 minutes.
2. The engine is run in MultiPV mode producing best 3 moves
3. Only positions with less than 5 minutes left to next period are considered

[Ferdy]

After round 5, multipv 3 at 15sec/pos.

Code: Select all

Moves matches between Stockfish 6 64 POPCNT top 3 moves and players from
Event 2nd Shamkir Chess 2015, after round 5

Intel(R) Core(TM) i7-2600K CPU @ 3.40GHz
Physical Cores: 4
Hyper-Threading: ON

Engine id name Stockfish 6 64 POPCNT
Hash 128 MB, Threads 2
MultiPV 3, MoveTime 15.0 sec/pos

MRank                   Players  AveDepth  AllPos   M1   M1Rate   M2   M2Rate   M3   M3Rate    W     L     D  Games  Score  SRate

    1               Giri, Anish        23     188  124   65.96%   27   14.36%   10    5.32%    0,    1,    4      5    2.0   40.0%
    2           Carlsen, Magnus        24     205  126   61.46%   35   17.07%   16    7.80%    3,    0,    2      5    4.0   80.0%
    3         Kramnik, Vladimir        25     202  123   60.89%   29   14.36%   23   11.39%    1,    1,    3      5    2.5   50.0%
    4                So, Wesley        23     171  102   59.65%   34   19.88%   14    8.19%    3,    1,    1      5    3.5   70.0%
    5             Mamedov, Rauf        24     184  106   57.61%   36   19.57%   11    5.98%    0,    1,    4      5    2.0   40.0%
    6    Mamedyarov, Shakhriyar        25     236  133   56.36%   43   18.22%   24   10.17%    1,    1,    3      5    2.5   50.0%
    7          Caruana, Fabiano        26     224  126   56.25%   42   18.75%   13    5.80%    0,    1,    4      5    2.0   40.0%
    8        Anand, Viswanathan        25     190  105   55.26%   40   21.05%   18    9.47%    1,    0,    4      5    3.0   60.0%
    9   Vachier-Lagrave, Maxime        25     196  108   55.10%   40   20.41%   20   10.20%    0,    1,    4      5    2.0   40.0%
   10            Adams, Michael        23     158   78   49.37%   37   23.42%   12    7.59%    0,    2,    3      5    1.5   30.0%

Legend:
MRank      - ranking based on most number of top 1 matching engine moves
AveDepth   - average depth reached by the engine in analyzing those positions
M1, M2, M3 - count of top 1, 2, 3 moves of engine that matches the player move.

Notes:
1. Comparison of moves starts when the time consumed including the elapsed time
   of the move of the current position is more than 2 minutes.
2. The engine is run in MultiPV mode producing best 3 moves

[Ferdy]

You might want to try. Tested in windows 7. Prepare your engine, I only test this on sf6. If you try other engine use first a fast time say 0.05 sec or 50ms.
The epd file is also included, this is only up to 3 rounds. This epd file is specially generated from pgn with move time info taken from chessbase playchess site.

Customized epd from round 1 to 5.

Download:
http://www.mediafire.com/download/34fad ... r12345.rar

[Ferdy]

Here is a quick analysis at 1 sec/pos.
I calculated the average error like this:
1. Run engine and search the position in 1 sec and record the move and score
2. If the best move found is the same as the move by the player then error for this position is zero
3. If the best move is not the same, then setup the position and make the move of the player, then run the engine for 1sec and record its score.
4. Get the error by pos_error = engine_score - (-engine_score_on_player_move). The latter score was negated because sf6 will return a score from side POV.
It could happen here that the score of the move by the player is actually better than the score of the move suggested by engine because the player move is made on the board before analyzing it, so in this case the pos_error is negative.
5. After analyzing all available position, all pos_error are added and divided by total positions analyzed including those zero error positions where the move are the same, to get the average error.

Other conditions:
1. When the search score of the position is already outside of +/-2 pawns as analyzed by the engine, this position is not included.
2. When error is more than 6 pawns, its error is set to 6 pawns for calculation.
3. The aveError column is in centipawn.
4. Other columns on specific difference in score are also recorded.

Code: Select all

Moves Error calculation based from Stockfish 6 64 POPCNT and players from
Event 2nd Shamkir Chess 2015, after round 5

Intel(R) Core(TM) i7-2600K CPU @ 3.40GHz
Engine id name Stockfish 6 64 POPCNT
Hash 128 MB, Threads 1, MoveTime 1.0 sec/pos

 Rank                   Players  AveDepth  TotalPos  AveError  Err_3.0P  Err_2.0P  Err_1.0P  Err_0.5P    W    L    D  Games  Score  SRate

    1             Mamedov, Rauf        17       177      2.57         0         0         0         0    0    1    4      5    2.0   40.0%
    2   Vachier-Lagrave, Maxime        19       193      3.18         0         0         0         2    0    1    4      5    2.0   40.0%
    3        Anand, Viswanathan        18       180      3.25         0         0         0         4    1    0    4      5    3.0   60.0%
    4    Mamedyarov, Shakhriyar        17       210      3.76         0         0         0         4    1    1    3      5    2.5   50.0%
    5           Carlsen, Magnus        18       188      4.55         0         0         0         1    3    0    2      5    4.0   80.0%
    6         Kramnik, Vladimir        18       181      5.07         0         0         1         3    1    1    3      5    2.5   50.0%
    7               Giri, Anish        17       180      5.33         0         0         0         4    0    1    4      5    2.0   40.0%
    8          Caruana, Fabiano        18       213      5.36         0         0         0         5    0    1    4      5    2.0   40.0%
    9            Adams, Michael        16       148      6.28         0         0         0         2    0    2    3      5    1.5   30.0%
   10                So, Wesley        17       137      6.67         0         0         1         5    3    1    1      5    3.5   70.0%

Legend:
Rank     - ranking based on low average error
AveDepth - average depth reached by the engine in analyzing those positions
AveError - average difference between the engine score and player score
Err_3.0P - count of positions where player has blundered by 3 or more pawns
Err_2.0P - count of positions where player has blundered by 2 to 2.9 pawns
Err_1.0P - count of positions where player has blundered by 1 to 1.9 pawns
Err_0.5P - count of positions where player has blundered by 0.5 to 0.9 pawns

Notes:
1. Analysis of moves starts when the time consumed including the elapsed time
   of the move of the current position is more than 2 minutes.

This would be interesting if time analysis is extended of course. I am looking for the best way to exploit those blunder counts on different amount of pawns to be factored in the determining quality of play,
probably also display the counts and percentage where the error is below half a pawn.

Have a look also here done by Matej Guid and Ivan Bratko.
http://en.chessbase.com/post/the-qualit ... tes-090413

[Ferdy]

Sample comparison for New York 1927 after round 1.
Capablanca has a higher average error compared to players from Shamkir 2015 after 5 games by each player .
Players that has more draws tend to get low average error.

Code: Select all

Moves Error calculation based from Stockfish 6 64 POPCNT and players from
Event New York 1927, after round 1

Intel(R) Core(TM) i7-2600K CPU @ 3.40GHz
Engine id name Stockfish 6 64 POPCNT
Hash 128 MB, Threads 1, MoveTime 1.0 sec/pos

 Rank                   Players  AveDepth  TotalPos  AveError  Err_3.0P  Err_2.0P  Err_1.0P  Err_0.5P    W    L    D  Games  Score  SRate

    1          Vidmar, Milan Sr        22        20      1.25         0         0         0         0    0    1    4      5    2.0   40.0%
    2        Nimzowitsch, Aaron        24        18      5.33         0         0         0         0    3    1    1      5    3.5   70.0%
    3     Capablanca, Jose Raul        22        16      7.19         0         0         0         0    3    0    2      5    4.0   80.0%
    4       Alekhine, Alexander        19        49     12.43         0         0         1         3    1    1    3      5    2.5   50.0%
    5         Spielmann, Rudolf        20        25     13.08         0         0         1         1    0    2    3      5    1.5   30.0%
    6     Marshall, Frank James        21        37     13.38         0         0         1         3    0    2    3      5    1.5   30.0%

Legend:
Rank     - ranking based on low average error
AveDepth - average depth reached by the engine in analyzing those positions
AveError - average difference between the engine and player score, in centipawn (1p = 100)
Err_3.0P - count of positions where player has blundered by 3 or more pawns
Err_2.0P - count of positions where player has blundered by 2 to 2.9 pawns
Err_1.0P - count of positions where player has blundered by 1 to 1.9 pawns
Err_0.5P - count of positions where player has blundered by 0.5 to 0.9 pawns

Notes:
1. Analysis of positions start at move 8.
2. Only positions with engine search score of [-2.0 to +2.0] pawns will be considered

[Ferdy]

In my previous post New York 1927 after round 1, some positions were not analyzed, so here is a complete analysis.

Code: Select all

Moves Error calculation based from Stockfish 6 64 POPCNT and players from
Event New York 1927, after round 1

Intel(R) Core(TM) i7-2600K CPU @ 3.40GHz
Engine id name Stockfish 6 64 POPCNT
Hash 128 MB, Threads 1, MoveTime 1.0 sec/pos

 Rank                   Players  AveDepth  TotalPos  AveError  Err_3.0P  Err_2.0P  Err_1.0P  Err_0.5P    W    L    D  Games  Score  SRate

    1        Nimzowitsch, Aaron        18       134      4.75         0         0         0         2    3    1    1      5    3.5   70.0%
    2     Capablanca, Jose Raul        17       125      5.31         0         0         0         3    3    0    2      5    4.0   80.0%
    3       Alekhine, Alexander        18       178      8.12         0         0         1         5    1    1    3      5    2.5   50.0%
    4          Vidmar, Milan Sr        17       138      8.64         0         0         0         8    0    1    4      5    2.0   40.0%
    5         Spielmann, Rudolf        18       167      9.20         0         0         1         9    0    2    3      5    1.5   30.0%
    6     Marshall, Frank James        18       158      9.71         0         0         2         7    0    2    3      5    1.5   30.0%

Legend:
Rank     - ranking based on low average error
AveDepth - average depth reached by the engine in analyzing those positions
AveError - average difference between the engine and player score, in centipawn (1p = 100)
Err_3.0P - count of positions where player has blundered by 3 or more pawns
Err_2.0P - count of positions where player has blundered by 2 to 2.9 pawns
Err_1.0P - count of positions where player has blundered by 1 to 1.9 pawns
Err_0.5P - count of positions where player has blundered by 0.5 to 0.9 pawns

Notes:
1. Analysis of positions start at move 8.
2. Only positions with engine search score of [-2.0 to +2.0] pawns will be considered

[Dirt]

1. When the search score of the position is already outside of +/-2 pawns as analyzed by the engine, this position is not included.

So, if a player has a winning position and makes a mistake that results in an even score, it doesn't count as an error?

[Ferdy]

1. When the search score of the position is already outside of +/-2 pawns as analyzed by the engine, this position is not included.

So, if a player has a winning position and makes a mistake that results in an even score, it doesn't count as an error?

Yes.

[Vinvin]

Is it possible to get a list of the 20 biggest blunders ?