Release of my Engines Aggressiveness Statistic Tools

I proudly release my 2 Engine Aggressiveness Statistics Tools (EAS-Tools).
With these tools, the aggressiveness of engines (or human players) can be measured. How do the tools work?
The tools are searching for sacrifices in games (from 1 pawnunit-sacs up to 5+ pawnunit-sacs) and for won games, which are very short (split into games up to 40/45/50/55/60 moves), but only, if the short game ended before the endgame (a low material check is done!). Because a weaker player can be playing aggressive, too, the EAS-Score (= Engine Aggressivenes Score, see explanation below) and all other statistics are build on percents from the won games of an engine/player. So, if an engine has won more games, it must win more short games or win games with sacrifices. A weaker engine, which has won less games, need less wins of short games or win games with sacrifices.

What is the EAS-Score?
EAS-Score is: (percent*10) of the percent-values of the sacs (1-5+ pawnunits) calculated out of the won games by the engine, only. Higher pawnunits-sacs give bonus-points:
1 pawnsac = 3x points *** 2 pawnsac = 10x points *** 3 pawnsac = 25x points 4 pawnsac = 50x points *** 5+ pawnsac = 100x points
Additionally, very short won games (percent*10) with or without sacs give bonus-points: 60 moves or less= 3x points *** 55 moves or less= 10x points *** 50 moves or less= 25x points 45 moves or less= 50x points *** 40 moves or less= 100x points (but only, if the short game ended before endgame (low material check is done!))

Of course, the EAS-Score is not a fixed value. It highly depends on the strength of players and opponents. And in engine-tournaments it depends on thinking-time, opening-sets, PC-speed etc. But, IMO, the EAS-Score can be very helpful especially for engine-developers, which test their engine dev-versions (or new neural-nets) always in the same way and versus the same opponents. In this case, the EAS-Score quickly shows progress or regress in aggressiveness...
And in a RoundRobin-tournament or used on the gamebase of a ratinglist, the EAS-Score ratinglist (see the statistics_EAS_rating.txt file) can be very interesting.
The EAS-Score of humans is much higher, than the EAS-Score of engines, because humans win much more short games, because of blunders, played by the (human) opponent... So never compare EAS-Scores of humans and engines!!!
The EAS-tool for one engine/player writes 5 pgn-files with all found sacrifice games. All short games, which ended before endgame, are stored in short_midgame_wins.pgn. And a statistics.txt -file for the engine, with all values of interest, inccluding the EAS-Score.
The EAS-tool for all engines/players writes a statistics_EAS_rating textfile, where all engines/players are ranked by their EAS-Score. And a second ratinglist below, with all important stats: number of wins, sacrifices and short-wins (before endgame) by the engine. Additionally all interesting games (sac-games and all short games, which ended before endgame, are stored in shorts_and_sacs_games.pgn (without doubles and comments)

Download from my website:
https://www.sp-cc.de/files/engines_aggressiveness_statistics_tool.7z

Here the statistics calculated by the One_Engine_Aggressiveness_Stats-Tool of Stockfish 220205 (last SPCC-tested version before
introducing the new nnue-net architecture) and Stockfish 220210 (first SPCC-tested version with new net arch).
Mention: SF 220210 scored less points and played more draws than SF 220205, but the EAS-Tool reveals very clearly,
that SF 220210 with new net-arch plays much more aggressive! Number of short wins is higher, number of played sacrifices is higher
and the final EAS-Score is clearly higher (2146 to 3260) !!!

********************************************************************************************
*** Evaluated file: SF_220205.pgn ***
*********************************************************************************************
*** Number of games played by "Stockfish 220205 avx2": 7000
*********************************************************************************************
*** Number of draws played by "Stockfish 220205 avx2": 3691 (52.72%)
*********************************************************************************************
*** Number of won games (no doubles) played by "Stockfish 220205 avx2": 3284 (46.91%)
*********************************************************************************************
*** Overall scoring rate by "Stockfish 220205 avx2": 73.2%
*********************************************************************************************
*** Following percent-values calculated on the number of all won games, only (3284)
*** played by "Stockfish 220205 avx2"
*********************************************************************************************
*** Number of won games up to 40 moves: 13 (0.39%)
*** Number of won games up to 45 moves: 15 (0.45%)
*** Number of won games up to 50 moves: 27 (0.82%)
*** Number of won games up to 55 moves: 43 (1.30%)
*** Number of won games up to 60 moves: 24 (0.73%)
*** (all short games, counted here, ended before endgame (low material))
*** Number of all short wins (total) : 122 (3.71%)
*********************************************************************************************
*** Sacrifices found in the won games of "Stockfish 220205 avx2"
*** Number of 5+ pawnunits sacrifices: 8 (0.24%)
*** Number of 4 pawnunits sacrifices : 13 (0.39%)
*** Number of 3 pawnunits sacrifices : 30 (0.91%)
*** Number of 2 pawnunits sacrifices : 141 (4.29%)
*** Number of 1 pawnunit sacrifices : 331 (10.07%)
*** Number of all sacrifices : 523 (15.92%)
*********************************************************************************************
*** EAS-Tool Score points : 2146
*********************************************************************************************
*********************************************************************************************
*********************************************************************************************
*********************************************************************************************
*** Evaluated file: SF_220210.pgn ***
*********************************************************************************************
*** Number of games played by "Stockfish 220210 avx2": 7000
*********************************************************************************************
*** Number of draws played by "Stockfish 220210 avx2": 3754 (53.62%)
*********************************************************************************************
*** Number of won games (no doubles) played by "Stockfish 220210 avx2": 3231 (46.15%)
*********************************************************************************************
*** Overall scoring rate by "Stockfish 220210 avx2": 72.9%
*********************************************************************************************
*** Following percent-values calculated on the number of all won games, only (3231)
*** played by "Stockfish 220210 avx2"
*********************************************************************************************
*** Number of won games up to 40 moves: 22 (0.68%)
*** Number of won games up to 45 moves: 16 (0.49%)
*** Number of won games up to 50 moves: 31 (0.95%)
*** Number of won games up to 55 moves: 38 (1.17%)
*** Number of won games up to 60 moves: 34 (1.05%)
*** (all short games, counted here, ended before endgame (low material))
*** Number of all short wins (total) : 141 (4.36%)
*********************************************************************************************
*** Sacrifices found in the won games of "Stockfish 220210 avx2"
*** Number of 5+ pawnunits sacrifices: 11 (0.34%)
*** Number of 4 pawnunits sacrifices : 28 (0.86%)
*** Number of 3 pawnunits sacrifices : 62 (1.91%)
*** Number of 2 pawnunits sacrifices : 183 (5.66%)
*** Number of 1 pawnunit sacrifices : 390 (12.07%)
*** Number of all sacrifices : 674 (20.86%)
*********************************************************************************************
*** EAS-Tool Score points : 3260
*********************************************************************************************

Here the EAS-ratinglist of my SPCC-ratinglist gamebase. IMO it is spectacular. Danasah 9 for example is on rank 2 in the EAS-ratinglist, but on rank 37 (of 38) in my regular SPCC-ratinglist. So, Danasah 9 plays not so strong, but extremly aggressive. And see, how aggressive Velevet 3.3.0 is! These 2 engines are clearly better than all others (even Wasp on place 3 plays way less aggressive (see the much weaker score!))
And see, how bad the EAS-score of the 3 strong engines Ethereal, Koivisto and Berserk is...


*****************************************************************************
*** Engine Aggressiveness Tool V1.7 Score points Ratinglist
*****************************************************************************
Score is: (percent*10) of the percent-values of the sacs (1-5+ pawnunits)
calculated out of the won games by the engine, only.
So, a weak engine (with a small number of won games) can get
a high EAS-scoring, too, when the percent of sac-games in the won games
is high (and the number of short wins). Higher pawnunits-sacs give bonus-points:
1 pawnsac = 3x points *** 2 pawnsac = 10x points *** 3 pawnsac = 25x points
4 pawnsac = 50x points *** 5+ pawnsac = 100x points
Additionally, very short won games (percent*10) give bonus-points, when they
ended before endgame (a check for low material is done):
60 moves= 3x points *** 55 moves= 10x points *** 50 moves= 25x points
45 moves= 50x points *** 40 moves= 100x points.
*****************************************************************************
*** Evaluated file: All_Stockfish.pgn ***
*****************************************************************************
Rank EAS-Score sacs shorts Engine/player
-----------------------------------------------------------------------------
001 009053 24.08% 12.06% "Velvet 3.3.0 avx2"
002 008480 26.24% 11.53% "Danasah 9.0 avx2"
003 005780 12.83% 11.31% "Wasp 5.20 avx"
004 005440 20.07% 05.37% "Revenge 2.0 avx2"
005 005311 17.89% 09.09% "Tucano 10.00 avx2"
006 004610 17.62% 06.71% "Arasan 23.3 avx2"
007 004458 16.35% 05.70% "Lc0 0.28.0 744706"
008 004074 12.52% 08.86% "Scorpio 3.0.14d cpu"
009 003952 23.77% 03.60% "Coiled 1.1 avx2"
010 003888 08.97% 08.44% "Fritz 18 x64"
011 003775 15.43% 07.40% "Stockfish final HCE"
012 003773 14.57% 07.94% "Igel 3.0.5 popavx2"
013 003694 20.67% 02.31% "Gogobello 3 avx2"
014 003599 12.33% 06.48% "Marvin 5.2 avx2"
015 003529 16.84% 06.55% "Velvet 3.2.0 avx2"
016 003503 10.04% 07.57% "Weiss 2.0 popc"
017 003389 21.24% 04.67% "Stockfish 220319 avx2"
018 003326 18.46% 02.75% "Toga 4 1.11 avx2"
019 003155 18.92% 04.40% "Stockfish 220308 avx2"
020 003133 17.70% 03.25% "Fire 8.NN MCTS avx2"
021 002982 10.37% 06.56% "Combusken 2.0.0 amd64"
022 002893 16.94% 05.43% "Stockfish 14 210702"
023 002817 09.39% 06.16% "Chiron 5 x64"
024 002388 15.26% 03.00% "Slow Chess 2.83 avx2"
025 002347 07.78% 05.02% "Minic 3.18 znver3"
026 002285 18.17% 03.52% "Stockfish 14.1 211028"
027 002202 19.06% 02.78% "KomodoDragon 2.6 MCTS"
028 002194 20.07% 03.63% "KomodoDragon 2.6 avx2"
029 002081 13.61% 02.13% "Fire 8.NN avx2"
030 001950 10.68% 04.12% "RubiChess 2021 avx2"
031 001899 12.17% 04.58% "Nemorino 6.00 avx2"
032 001759 05.17% 05.98% "Clover 3.0 avx2"
033 001726 07.44% 03.45% "Zahak 9.0 avx"
034 001703 07.63% 03.87% "Stash 32.0 popc"
035 001649 08.74% 02.45% "Berserk 8.5 avx2"
036 001206 13.93% 02.16% "Ethereal 13.50 nnue"
037 000773 08.33% 01.25% "Koivisto 8.0 avx2"
038 000767 09.06% 00.90% "Seer 2.4.0 avx2"
****************************************************
*** EAS-tool (C) 2022 Stefan Pohl (www.sp-cc.de) ***
****************************************************

(for this ratinglist, the tool had to evaluate more than 180000 games. That took around 40 minutes on a "normal" PC)

If we look at Toga 4 1.11 (Rebel), we clearly see, that Toga plays a high number of sacrifices (18.46% is a really strong value), but a very small number of short-wins (before endgame) (2.75% is a very bad value). So, here is potential for more aggressiveness in upcoming updates...
And this is the reason, why IMO my new tools will be useful for engine-developers and nnue-net builders!

I added a new site to my website:

https://www.sp-cc.de/eas-ratinglist.htm

There, the EAS-ratinglist and the ranking of the engines in my SPCC-ratinglist can be compared directly. I think, this is helpful...