EAS-Tool V6.0 - major update

[pohl4711]

I released the new V6.0 of my EAS-Tool. This is a major update, introducing a new early-sacrifice-bonus and a complete overhaul of the scoring system.

"Very cool! I like the EAS-Tool. Also the game selection was awesome!" - GM Vidit Gujrathi (world ranking #26, Indian Chess Olympiad gold medalist in 2024), 2025/07/28 on Lc0-discord

Download:
https://www.sp-cc.de/files/engines_aggr ... ol_v6.0.7z

Whats new in V6.0?

1) A new early sacrifice bonus is introduced. If a sacrifice happens early in the game, the engine gets EAS-points: (percent x 100 / 18)*(percent x 100 / 18). Early sac limit is: Average length of all wins in a database / 2. percent means here: percents of early sacs in the pool of all found sacs of one engine. So, theoretically, values of 0% or 100% can happen (but are very unlikely).

2) The early sac bonus made a complete overhaul of the scoring-system necessary. So, from V6.0 the EAS-scores are very diffferent (higher) from scores, calculated by older version of the EAS-Tool. So, do never compare results of differente EAS-Tool versions!

3) The short-wins move-limit for EAS-points is (after calculating it) no longer rounded to 5 or 10. This increases the accuracy.

4) A 3rd EAS-Ratinglist was added to the statistics file: Here just the EAS-points are listed, no percent values. This can be helpful for engine-developers, because it allows easier optimizations of single EAS-stats in engine-play.

5) The EAS-Tool now builds just 2 pgn-output databases: interesting_wins.pgn and errorgames.pgn

errorgames.pgn contains games with a Termination-Tag that suggests a non-regular game ending. ("anbandoned" for example). These games are sorted out before building the EAS-score, so these games will not lead to distorted results.

interesting_wins.pgn contains:
1) Queen Sacrifices, followed by
2) 5+ PawnUnit Sacrifices, followed by
3) 4 PawnUnit Sacrifices, followed by
4) 3 PawnUnit Sacrifices, followed by
5) 2 PawnUnit Sacrifices, followed by
6) 1 PawnUnit Sacrifices, followed by
7) Very short games, followed by
8) Games, ended before endgame (material) was reached, followed by
9) Games with material imbalance (Rook vs. Bishop and 2 pawns for example)

The games in the output-files are sorted in 2 ways:
First: The games are sorted by categories (category 1 is followed by category 2, 3, ... etc.).
Second: In each category, the games are sorted by length (0-19 moves, followed by 20-24 moves, followed by 35-29 moves... and so on, up to 120 moves and beyond). So, in each category, the shortest wins are at the beginning and followed by the longer wins...
And, there are no double games in one output-file: If a game fits more than one category, it is stored in the lowest category, all other apperances of this game in higher categories are deleted. For example: A game contains a 3 PawnUnit-Sacrifice and is won before the endgame material is reached: This game is stored in category 4 (= 3 PawnUnit Sacrifices) and not in category 8...

And each games gets a new Annotator-Tag, so it is clear, which category the game belongs to.
One of these 9 tags is added to each game-notation:

[Annotator "EAS-Tool: Queen Sacrifice found in this game"]
[Annotator "EAS-Tool: 5+ PawnUnits Sacrifice found in this game"]
[Annotator "EAS-Tool: 4 PawnUnits Sacrifice found in this game"]
[Annotator "EAS-Tool: 3 PawnUnits Sacrifice found in this game"]
[Annotator "EAS-Tool: 2 PawnUnits Sacrifice found in this game"]
[Annotator "EAS-Tool: 1 PawnUnit Sacrifice found in this game"]
[Annotator "EAS-Tool: Game was very short"]
[Annotator "EAS-Tool: Game ended before endgame (material)"]
[Annotator "EAS-Tool: Material imbalance found in this game"]

[pohl4711]

Here the new Top10 of my full EAS-Ratinglist. Rebel Extreme 1.1 is the new #1 here (former: Patricia 3.1)

Code: Select all

                         early           bad  avg.win 
Rank  EAS-Score  sacs    sacs   shorts  draws  moves  Engine/player 
---------------------------------------------------------------------------
   1    748559  52.40%  42.58%  49.75%  08.28%   65   Rebel Extreme 1.1  
   2    712625  51.61%  43.75%  41.43%  05.49%   66   Patricia 3.1 avx2  
   3    660963  50.51%  44.01%  41.89%  06.08%   67   Patricia 250510 a512  
   4    653831  50.65%  39.79%  41.13%  07.20%   68   CSTal 2.1 EAS  
   5    643722  50.79%  41.53%  42.61%  05.85%   67   Patricia 5.0 avx512  
   6    595083  48.39%  44.08%  38.12%  05.82%   67   Cerberus 21125081r9b  
   7    555597  46.96%  40.00%  34.95%  03.46%   70   Patricia 4.0 avx2  
   8    526843  43.56%  41.14%  33.89%  04.21%   68   Cerberus 21124081r81  
   9    428179  37.88%  38.53%  28.30%  08.79%   75   Rebel Extreme avx2  
  10    381300  32.46%  39.90%  28.36%  09.25%   71   Velvet 8.1.0 risky  
  

New EAS-point ratinglist (no percents):

Code: Select all

                             early              bad 
Rank  EAS-Score      sacs     sacs   shorts    draws    Engine/player 
-------------------------------------------------------------------------------------
   1    748559     342030   115696   227832    63001    Rebel Extreme 1.1  
   2    712625     344870   114049   177530    76176    Patricia 3.1 avx2  
   3    660963     308440   109536   171163    71824    Patricia 250510 a512  
   4    653831     297610    93841   195299    67081    CSTal 2.1 EAS  
   5    643722     294220   102900   170426    76176    Patricia 5.0 avx512  
   6    595083     274470   109536   134901    76176    Cerberus 21125081r9b  
   7    555597     261640    84284   123237    86436    Patricia 4.0 avx2  
   8    526843     235170    96984   113464    81225    Cerberus 21124081r81  
   9    428179     196630    55796   112752    63001    Rebel Extreme avx2  
  10    381300     152900    78841    90510    59049    Velvet 8.1.0 risky  

[Rebel]

Thanks for new tool, I can confirm your results my own testing of 3 x 15.000 games of each engine that played, 44 in total.

Code: Select all

                         early           bad  avg.win 
Rank  EAS-Score  sacs    sacs   shorts  draws  moves  Engine/player    *** 3400 elo pool ***
---------------------------------------------------------------------------
   1    693464  50.67%  15.78%  52.81%  08.64%   53   Rebel-Extreme-1.1
   2    656823  52.90%  16.94%  41.48%  04.65%   59   Patricia-3.01  
   3    601303  48.07%  16.58%  44.40%  07.86%   57   CSTal-2.1-EAS  
   4    581864  51.47%  15.93%  39.94%  06.42%   58   Patricia-5.0  
   5    560877  50.26%  13.85%  36.81%  03.39%   60   Patricia-4  
   6    418892  37.83%  13.61%  35.37%  10.13%   61   REBEL-EXTREME  
   7    376994  36.89%  12.06%  35.64%  10.59%   60   Komodo-14.1-Aggressive  
   8    365292  33.10%  13.06%  34.50%  10.41%   59   Velvet-8.1-Risky  
   9    300459  23.25%  12.41%  32.90%  12.08%   58   Stockfish-17.1  
  10    293744  26.80%  10.73%  29.63%  12.33%   63   Rebel-EAS-2.0  
  11    274804  22.08%  11.29%  31.42%  13.50%   59   Fritz-20  
  12     78280  01.94%  15.19%  19.85%  39.61%   66   Clover-3.3  
  13     62157  03.19%  09.09%  13.72%  37.29%   71   Wasp-7.0  
  14     60065  01.86%  15.85%  14.04%  38.41%   68   Pawn-3.0  
  15     57426  02.11%  11.83%  14.14%  39.63%   68   Seer-2.50  
  16     57387  01.66%  11.69%  15.31%  38.70%   67   Black-Marlin-8.0  
  17     55797  01.67%  07.69%  12.67%  34.02%   70   Koivisto-6.16  
  18     53523  02.19%  08.57%  12.34%  37.03%   77   Igel-3.30  
  19     46523  01.99%  18.82%  09.14%  38.67%   73   Halogen-12  
  20     43430  01.55%  22.22%  06.58%  38.09%   75   Altair-6.0  
  21     41580  01.72%  12.31%  08.96%  35.02%   70   viridithas-9.0.0  
  22     36134  02.41%  05.26%  10.14%  39.91%   71   Koivisto-6.0  
  23     33787  02.46%  04.17%  09.83%  42.55%   75   Wasp-7.07  

Code: Select all

                         early           bad  avg.win 
Rank  EAS-Score  sacs    sacs   shorts  draws  moves  Engine/player     *** 3500 elo pool ***
---------------------------------------------------------------------------
   1    698086  49.71%  15.96%  52.33%  09.69%   55   Rebel-Extreme-1.1
   2    626666  47.67%  18.17%  48.58%  08.89%   58   CSTal-2.1-EAS  
   3    589168  51.51%  17.58%  39.43%  06.71%   60   Patricia-5.0  
   4    563347  45.69%  20.28%  45.26%  06.07%   58   Patricia-3.01  
   5    538031  46.03%  12.19%  39.46%  04.32%   61   Patricia-4  
   6    408810  35.71%  18.02%  35.14%  10.74%   65   REBEL-EXTREME  
   7    380389  32.48%  12.53%  37.06%  10.29%   60   Velvet-8.1-Risky  
   8    317037  23.82%  10.75%  35.31%  11.85%   60   Stockfish-17.1  
   9    301890  25.12%  12.60%  29.97%  12.70%   65   Rebel-EAS-2.0  
  10    253273  19.49%  10.24%  30.91%  14.36%   62   Fritz-20  
  11    152425  03.15%  14.42%  31.87%  36.36%   63   Stockfish-final-HCE  
  12    102400  01.43%  12.73%  24.38%  40.63%   65   Seer-2.6.0  
  13     96469  02.94%  13.10%  20.77%  33.80%   70   Slow-Chess-2.9  
  14     85591  01.63%  12.50%  21.26%  37.70%   67   Clover-6  
  15     79062  01.58%  16.42%  18.09%  37.32%   71   RofChade-3.1  
  16     66558  01.43%  19.35%  14.27%  38.83%   72   Caissa-1.12  
  17     63410  01.53%  22.45%  14.08%  38.50%   70   Viridithas-11.0.0  
  18     61089  01.27%  08.82%  14.41%  36.00%   72   Koivisto-9.2  
  19     58479  02.09%  15.07%  12.93%  35.90%   75   Alexandria-5.0.0  
  20     58040  01.83%  11.25%  14.67%  39.02%   73   Starzix-5.0  
  21     57463  02.43%  11.36%  11.90%  32.44%   74   Igel-3.5.0  
  22     55779  01.32%  09.09%  14.00%  35.99%   71   Obsidian-8.0  

Code: Select all

                         early           bad  avg.win 
Rank  EAS-Score  sacs    sacs   shorts  draws  moves  Engine/player     *** 3578 elo pool ***
---------------------------------------------------------------------------
   1    784590  50.85%  26.25%  62.29%  09.39%   53   Rebel-Extreme-1.1
   2    712571  48.83%  21.20%  57.81%  09.51%   55   CSTal-2.1-EAS  
   3    649573  47.57%  20.41%  58.25%  05.82%   58   Patricia-3.01  
   4    646841  53.59%  20.73%  44.61%  07.18%   60   Patricia-5.0  
   5    628215  42.36%  23.89%  51.98%  11.21%   57   Rebel-Extreme  
   6    619339  49.30%  22.86%  44.29%  07.56%   59   Patricia-4-dev  
   7    615864  49.05%  22.84%  45.26%  06.25%   59   Patricia-4  
   8    407376  36.21%  17.29%  37.22%  10.93%   62   Velvet-8.1-risky  
   9    350339  31.13%  15.52%  31.44%  13.78%   67   Rebel-EAS-2.0  
  10    291720  24.02%  17.30%  29.18%  13.79%   65   SF-17.1  
  11    263961  20.95%  13.60%  30.50%  14.54%   65   Fritz-20-AVX512  
  12    103445  01.65%  20.18%  22.04%  37.22%   67   viridithas-14.0.1  
  13    100799  01.77%  20.34%  21.66%  37.01%   68   seer_v2.8  
  14    100164  01.98%  20.38%  20.87%  31.93%   68   Clover.8.0.2  
  15     95392  02.30%  21.03%  18.88%  28.48%   69   Obsidian130  
  16     84145  02.76%  16.39%  16.54%  27.97%   71   Lizard-11_0  
  17     77771  02.37%  20.79%  15.34%  30.57%   72   PlentyChess-2.1.0  
  18     71754  01.88%  13.53%  16.00%  31.69%   71   berserk-13  
  19     71202  02.23%  15.34%  15.74%  32.38%   72   Alexandria-7.0  
  20     70631  01.78%  17.76%  14.80%  33.86%   72   caissa-1.20  
  21     70574  01.94%  22.22%  13.09%  34.76%   75   Titan-1.1  
  22     68722  02.08%  20.93%  12.94%  34.67%   75   Titan  

Just one question, why the sac changed to early sac, isn't a sac not always a sac?

And as a general note, mister Frank Quisinsky is accusing Chris and me of fraude, stealing Fabien source code and call it our own.

[pohl4711]

Thanks for new tool, I can confirm your results my own testing of 3 x 15.000 games of each engine that played, 44 in total.


Just one question, why the sac changed to early sac, isn't a sac not always a sac?

??? There are now 2 sac-stat columns in the EAS-ratinglist (sacs and early sacs) instead of just sacs (old V5.8)... So, now the EAS-ratinglist has an additional stat "early sacs", but the sacs-stat is still there. Of course it is.

The sacs are found as usual. Then all sac-games of an engine are counted, cutted off at the early-sac-movelimit (+8 plies, because my sac-search-algorithm needs 8 additional plies for finding a sac) and then the sac-search is repeated. Finally the percents are build:
Example: Engine X has played 100 sacrifices in a tournament and 35 of them happened before early-sac-movelimit was reached. Then the early-sac percent is 35% (35 out of 100). So, theoretically, early sac bonus percents can go from 0 up to 100, which makes sure, it is spread as wide as possible.

Great, that the new tool works so fine on different Elo-pools for opponnents. Thank you for proving that. This is not a matter of course: Because if the strength of the opponents gets much higher, the number of wins by the aggressive playing engines decreases. But because the stats are all build from percents and not absolute numbers and the short-win movelimit is flexible as the early-sac movelimit is, the EAS-Tool can still manage to find the good stuff!

[Rebel]

Ah, I stand corrected, 2 sacs evaluations...........