Dear Chess Friends,
On this news page,
You can read my postings about Human-Like Rating
13.09.2022:
Here are current standings and soon new updates are comingNote also that each player will contain minimum 400+ gamesSure (for a better conclusion), we need more games to be outAnyhow, I expected to be similar to their Elo, but not so much..As we see, current rankings are not too far away from reality...On other hand, we can't expect same peak Elo...due to many Reasons..e.g Engine, Book strength, used Blitz conditions etc.In other words, minus plus 50 Elo diff. can be counted normal...Another similarity, according to latest GMs Blitz tours (2600+):The overall draws (based on thousands of Human games) 27%Really shocking news.e.g same draw stats as current tourney!The Leader Kasparov + Rybka - Congrats to Kasparov, Rybka team!11.09.2022:
I am pleased to announce that,Another new competition is coming very soon:Human-Like Engine Rating will be out there!In short, first ever most humanized eng rating!1st of all, Here are the planning players:
World Champions (by peak rating):
Carlsen 2882 + Stockfish 1.4 x64 Kasparov 2856 + Rybka 1.2fAnand 2817 + Naum 2.2 x64Kramnik 2817 + Bright 0.4aTopalov 2816 + Junior 2010Fischer 2785 + Zappa MexicoKarpov 2780 + Shredder 10.1 x64Ponomariov 2764 + Hiarcs 10Botvinnik 2730 + Tornado 4.4 x64Capablanca 2725 + Chess Tiger 2007Kasimdzhanov 2715 + Spike 1.1Tal 2705 + The King 3.50 x64Khalifman 2702 + Daydreamer 1.75 x64Alekhine 2700 + Gandalf 7 x64Lasker 2700 + Gaviota 0.84 x64Spassky 2690 + Pawny 1.0 x64Petrosian 2660 + Ktulu 7.5Steinitz 2650 + Pharaon 3.5.1Smyslov 2630 + Ruffian 2.1.0Euwe 2600 + Fruit 2.0
Some Notes:The target will be to test firstly all the World Champions,Which will be played by Engines close to 2600-2900 EloThat means via new tourney, the openings will play lessInfluence, e.g in case of matches 2650 Elo vs 2800 EloBut if av. 2800 vs 2800 Elo: the openings will play BIG role..However, I think that it's worth to give a try....at least nowWe will have chances to see regarding which engines areMore close to theirs strength... because there are suchParticipants, which are still not tested... as examples:GMs opening books via engines of different strength etc.Btw, some time ago, I ran similar tourneys: VIIIh / VIIIgWhere now different engines (in strength) will be played...For example in older tours: the engines were much closerIn strength e.g minus plus 25-50 Elo difference, but nowThe Elo difference is minus plus 300 Elo...for this reason,I named current tour as Human-like (not as Book tourney)Not sure, maybe later, I may run another new competition, In case, GM books will be played by Cfish 030820 engineAnd if...the strongest books will play the biggest influence!
And without to not mention this I can't as well,No any Engine seems to be same as Human's playing style!E.g one of main difference between Human vs Engine is that,As usual, Humans prefer to play more various opening lines,Where mostly of the Engines prefer same or similar openings,Due to mostly engine's opening code is quite limited and strict..Sure I mean in case of eng testings as without using any bookBut since the planning tour will be via books...not so important..As other difference: Humans offer draw (if the game is draw),Where Engines are continuing to play (in draw positions) till The outputs of Engines are starting to show as equal (0.00)And as other difference is that, Humans play more frequentBlunder moves than the chess engines, especially in fast TC Sure there are other differences, between Human and EnginesE.g as usually, the chess engines are stronger in end-gamesBut in mid-games: Humans are more close to Engines strength
Sure, after all,We can't expect 100% identical moves, actually nothing strange..For example, this happens even if we replay the same game via Same engine versions as well...however I think that many engineMoves can be counted close or similar to Humans played moves...At least I am talking about under current planning conditions!
Continuing...
To be playing as Human-like I have done my best about that,The planning rating to be close or to be similar to Humans stylesFor examples, NNUE engines out (the Evalfiles are not allowed)Plus in this tour, EGTB (Nalimov, Syzygy...) are not allowed too,Because Humans (when playing..) don't gain from external files!Otherwise, many chess friends may call it simply as cheating...As another important criteria, I created GMs books, optimized toPrefer theirs played openings, e.g Carlsen book tuned to preferOnly those opening lines, which are played by Magnus CarlsenAnd I have picked (for Carlsen player) Stockfish 1.4 x64 engineE.g the current SF is about 2850-2900Elo (according to ratings)
Just I'd like to point out this too,I believe that it would be wrong if Carlsen to be played byStockfish 15 NNUE, which is close to 3750-3800 Elo pointsOr Carlsen to be via X engine, e.g close to 1000-1800 EloSo Stockfish 1.4 (av. 2850-2900 Elo) seems to be not bad..As other example: Euwe will be played by Fruit 2.0 engine,Which is close to 2600 Elo and Fruit will be played by book,Optimized to prefer only Max Euwe's played opening linesSame words are going for all of the rest participants as well
A few words about Time Management Control, Ponder etc.Which are important factors regarding Human-like styles too!So... I ran some beta testings about the most stable engines,Plus which are more close to GM strength.. e.g many are out, Simply due to they are stronger or weaker...or not capable to Use properly Blitz TC, e.g some of them playing too fast...and As usually half of the time is used, normally you may know too,Mostly Human players prefer to think more time on openings,Mid-games than in end-games...e.g mainly up to 40-50 moves...Plus some engines are out too, because not capable to ponderFor example, Humans prefer to think (during opponent's turn)!
To be more clear,There are some engines e.g Chessmaster, Rodent, Prodeo, Sjeng...Which offer to play many various GM, Human personalities...yes,Sounds not bad, but I realized to stay far away of such testingsThe main reason due to STRENGTH, e.g Elo difference between CM11 Anand vs CM Euwe should be at around 200 Elo: ResultsOther test, mostly CM11 settings are close to 50 Elo diff. ResultsAs other main reason is that,Some of them (Prodeo, Rodent etc) has Time management issues...
Btw, Lc0 v0.29.0 CPU + Maia-1900 is much over than 1900 Elo E.g according to Maia's author: Maia-1900 weight is trained onOnly 1900 Elo Human games, but as far as I noticed is stronger!Actually no BIG surprise, some perform diff. than expected: ResultsHowever, later I hope to test Lc0 CPU+ Maia-1900 via more games,Then all this mystery will be cleared (regarding its real Elo strength)ShashChess 24 via Capablanca and Tal setting is another mystery!I mean much stronger, beyond than any GM strength..not sure why?!As you may see, via Human-like rating: many details will be cleared!
In other words, mainly as Engine players:As far as possible I picked UCI engines (to avoid using adapters)And the ones, which are very close to GMs playing level strengthPlus those engines, which are stable e.g capable to ponder, no time Management issues, no time forfeits etc. What we can need more ?)Note that many of the selected engines were as Number One (in past)That means, mostly of them are based mainly on own original ideas!Because no any one can be on Top by the work of others, right !?
As Final words (via planning new rating)We will have possibility to see all World Chess Champions openings!After checking closely, Champions Elo diff. between 1900s - Present:Is minus plus 300 Elo and it's clear that Humans Elo progress is notSo huge as Machines Elo progress (for example since 2006 - 2007):I mean in 15 years, the Engine improvement is about 1000 Elo), orStockfish (since 2020) managed to be improved 250+ for Results but I have to admit that, without Humans, the Machines would not exist !)And to be honest, I'm very impressed with World Chess Champions!Yes... theirs combative, tactician and attacker styles make them as One of the greatest Human chess players of all time! Note also that, I was quite careful about choosing the right strength chess engines..And I've done my best about tuning the books (for theirs GM styles)In short, the opening books will play as main KEY to be Human-like!
Anyhow, just a fun competition and enjoy...!)
Ok..that's all for now and thanks for your interest...
Best,Sedat