Some tests to calculate the possible improvement of Sargon 1.01 over 1.00 after (possibly) elimination of
a big part of 3-fold repetitions.

First I played 200 games vs. two rated programs not too far away from Sargon1978 1.00, also listed at CCRL-Blitz.
TC was 2/40, hash for Raven 128MB, ponder off

While the tournament was running I realized that this was not the ideal way to calculate
the possible improvement, because against equal or even weaker opponents, version 1.01
would have more chances to get better positions, which won't end in 3-fold now.

Anyhow it still was interesting though and against the selected opponents the improvement would have been 10-15 rating points.
The games revealed also that in some cases with quite some advantage it still failed to avoid repetition.

From the 11 repetitions counted, this two should be watched.




One game also was interesting, though no 3-fold. Here Sargon could not win with K+R vs. K and it ended in a 50 moves draw.



All games are available here:
https://rwbc-chess.de/Downloads/Debug/20210209.7z

CuteChess 1.2x Sargon1978_Test
RWBC-CAPPUCCINO Win7U64 Q8200 2.33Ghz + Nvidia GT 710, 2021.02.09 - 2021.02.10
------------------------------------------------------------------------------------------
 1: Blikskottel_08  76.5 / 100 (+70 -17 =13)
 2: Raven_030       76.0 / 100 (+74 -22 =4)
 3: Sargon1978_101  47.5 / 200 (+39 -144 =17)
------------------------------------------------------------------------------------------
200 games: +94 -89 =17 (White-Black distribution)
 

Rank	Name	Elo	Points	Games	Score	CCRL
0	Sargon1978_101	-203	47.5	200	23.8%	
1	Blikskottel_08	205	76.5	100	76.5%	1459
2	Raven_030	200	76.0	100	76.0%	1466
						
	(Sargon_1978_100)			        1248
 
Player: Sargon1978_101
   "Draw by 3-fold repetition": 11
   "Draw by adjudication": 1
   "Draw by fifty moves rule": 1
   "Draw by insufficient mating material": 2
   "Draw by stalemate": 2
   "Loss: Black mates": 71
   "Loss: White mates": 73
   "Win: Black mates": 18
   "Win: White mates": 21
 
 

Then I tried another approach by using the games for Sargon1978 1.00 available already from CCRL and CEGT.
I just wanted to recalculate the rating for both, IF it would have been able to win most games with big advantage
in not too complicated positions (considering its strength).

That remained quite a task, because I realized CCRL and CEGT don't even save the final result comment.
(also not the eval/depth/times for their blitz lists, but that I knew already - may be authors can get commented bltz games, but I am not sure)

With some tool chain containing pgn-extract and some tools from Norm Pollock I could extract all games which ended in 3-fold.
This also revealed the usage of a somehow buggy GUI, or buggy settings in a few cases at CCRL, because there were more 3-fold repetitions than games, one game of those even ended in no draw and the 3-fold went unnoticed ;-)

Then I used SCIDs game list export which also has the ability to save the final material for each game and calculated the material diff in eval.
(I had to care also if it was not the opponent who claimed 3-fold in much better position - this also happened not to seldom)
If the diff was very high I assigned a win instead of a draw by 3-fold (by version 1.00). I tried to be quite conservative.
I also checked most positions with very reduced material manually by looking at the final board position.
LucasChess BTW was very helpful here, because it shows a very clean database list AND the board in the same window.

The final result of my estimation and recalculation of the ratings was around +58 for CCRL and +68 for CEGT.
(the diff lies in opponent mix and also that CEGT seems to adjudicate much later, so there were more 3-folds possible even in later stages
with up to +26 material eval)

The complete details for the process of recalculation of possible wins for version 1.01 can be found here:
https://docs.google.com/spreadsheets/d/1...K67vvH0CRuThgq0

I also uploaded the original pgn files (just renamed IIRC) and the extracted game files only containg 3-fold games.
https://rwbc-chess.de/Downloads/Debug/CCRL_Sargon1978_100.7z
https://rwbc-chess.de/Downloads/Debug/CEGT_Sargon1978_100.7z

C:\ChessTools\QualityControl\Sargon_Test\CEGT_Sargon1978_100>pgn-extract -Wepd --nofauxep -s -otemp.epd CEGT_Sargon1978_100.pgn 
C:\ChessTools\QualityControl\Sargon_Test\CEGT_Sargon1978_100>epd3fold temp.epd  
Number of 3-fold repetitions = 430
Number of games with a 3-fold repetition = 430
 
Number of games in CEGT_Sargon1978_100.pgn is 900
 

C:\ChessTools\QualityControl\Sargon_Test\EAP\CCRL_Sargon1978_100>pgn-extract -Wepd --nofauxep -s -otemp.epd CCRL_Sargon1978_100.pgn 
C:\ChessTools\QualityControl\Sargon_Test\EAP\CCRL_Sargon1978_100>epd3fold temp.epd  
Number of 3-fold repetitions = 402
Number of games with a 3-fold repetition = 396
 
Number of games in CCRL_Sargon1978_100.pgn is 1077