Last Update: 27.01.2022



Hello Chess Friends,

I am pleased to announce that,I organized another new Duel matches...

And as usual,I have good news for Permanent Brain ON's lovers!)
Yes...in the current Duel match, on my Ryzen machine too:Ponder ON 4CPU managed to defeat Ponder OFF 8CPUAlmost same Elo difference as on my 2x E5-2686 resultsAnd again and again,Ponder ON is just born to win...sure under right conditions!!For example..not so good idea using Evalfile for such testingsOtherwise...I can call it simply as cheating..really...because,You may know...in case...gaining from same Evalfile data willLead to not correct, misunderstanding, unfair results...just saying..I mean much better such testings to be played as NON-NNUE..!!Plus, not so good idea to be played with deep opening books tooIn short, we have to avoid giving much assistance...hope helps..Plus, in case...I suggest to be used at least Blitz time controlI mean, Ponder ON is gaining much when during opponents turn!Otherwise, what is the target of running these kind of testings ?)Why I mentioning this,New (Beginner) in testings may run at 20 Sec/Game or 1 Min etc
And as we see again,Ponder ON 4CPU performed 35 Elo better than Ponder OFF 8CPU
1 Ponder ON 4CPU +15/-5/=80 55.00% 55.0/1002 Ponder OFF 8CPU +5/-15/=80 45.00% 45.0/100
Games

Test Conditions:AMD Ryzen 7 4800HFritz GUI, Shredder Classic GUIHyper-Threading ONLarge Page OFFCfish EXT 291120 C40TC 3 Minutes512 MB hashPerfect 2022 book4-MEN
Some Notes:All of the games are played on same machine (in Auto232 mode)Cfish EXT 291120 is used just for comparing, very stable as Brainfish ...No any lost game on time is recorded...this is really very important!Because, at 3 Min time controls not many ones can be so much stable...Both engines are as NON-NNUE, but pretty strong, close to 3600 EloAnd for anyone interested, missed, according to my all previous testings,These engines produce less draw perc. than almost all rest Top ones!Once more, BIG thanks to the programmers...!
Note also that,Ponder ON performs better with Hyper-Threading as disabled e.g:2x E5-2686 v3, plus on my older machines e.g i7-980X, i7-920 etc..But according to my latest HT testings on my AMD Ryzen 7 machinePonder ON via HT ON or HT OFF does not make a BIG influenceIt seems, the latest AMD processor's architecture is improved a lot...------------------------------------------------------------------------------------------------------
And just another Duel, but played on 2x E5-2686 machine:
As usual as we see again,Ponder ON 4CPU performed 36 Elo better than Ponder OFF 8CPU
1 Ponder ON 4CPU +28/-13/=105 55.14% 80.5/1462 Ponder OFF 8CPU +13/-28/=105 44.86% 65.5/146
Games:

Test Conditions:2x E5-2686 machineFritz GUI, Arena GUIHyper-Threading OFFLarge Page OFFBrainfish 151019 BMI2 C40TC 3 Minutes512 MB hashPerfect 2022 book4-MEN
Some Notes:These matches are played on same machine too (in Auto232 mode)Parallel matches, but for each player installed separate folder of engineOtherwise, not recommended playing from same exe, for more DetailsBrainfish 151019 is used, because is very stable in 3 Min gamesNote also Brainfish is more than 2 years old...but not bad..I like it!And no any time forfeits are recorded...Impressive!
On other hand,How NNUE appeared on scene... I noticed also that,Stockfish plus many SF based ones forfeits on time via Ponder OFFStrange, normally newer versions should be better...but just oppositeBtw, Ponder OFF is as caprice LADY ))..about using time controls...Especially at fixed time controls, to not appear a time forfeit...e.g:It is something like to climb successfully to Everest mountain ))Actually there is always solution, even as Ponder OFF, for example:We should give some extra time (e.g seconds per move) otherwise,According to many testings, Ponder OFF suffers with time controls,But via Ponder ON, much more stable... plus stronger for sure...!!
Not so clear via Ponder ON ?Ok..then I am going to give you another, but serious example:Based on latest SCCT book tours games (since 22.08.2020)The data is unbelievable...almost perfect really:..for example:Just only 2 (two) games lost on time (based 156.157 games)
Meanwhile, according to another new testing as Ponder OFFI realized to see the influence as Ponder OFF at 3 Min/GameAnd just only in 12 games played out, 2 games are lost on timeNote:I used (for this test) almost same conditions as CS LIXThe mentioned Ponder OFF games are available: DownloadE.g just imagine if the tours were as Ponder OFF, how manyPlayers would be ranked incorrect and with unfair standings
By the way, for anyone missed, about more forfeits on timeEven as NNUE, some latest SF and based ones suffer: Details
However,I hope the next new SF releases to be fixed...and I think:Any strong engine should be able to play under all conditions!And no doubt that too,Fishtest produces very valuable data...but anyhow, I think thatTheir testings should be played under more various conditions...
As final words over current Duel matches,As I stated before, not so bad idea Ponder OFF to be used:For Ultra fast, Bullet time controls...because we can't expectMuch quality games...many blunder moves can be appeared..Plus if the target is to produce more games (via Parallel matches)Then Ponder OFF, e.g using 1 core...will be a good choice as wellBut please be careful with the time controls, should not be as fixed...E.g recommended to be used such as 1m+1s, 3m+1s, 40/4 etc.
But in case of serious engine competitions, thenNo doubt that MP + Ponder ON seems to be best!!Proved, guaranteed...and once more:Chess is thinking in the opponent's turn!
Hope helps...

Btw, For anyone missed more notes about Permanent Brain: DetailsLearn the real power/strength of pondering ON vs OFF: ResultsAs alternative, Ponder ON vs OFF, 4 Men vs 5 Men etc. Results

Best,Sedat