Reports | February 10, 2011 21:43

Free Houdini beats commercial Rybka 23.5-16.5

Free Houdini beats commercial Rybka 23.5-16.5In what could be loosely described as the unofficial world championship of chess engines, free chess engine Houdini beat commercial engine Rybka. The match lasted forty games and the final score was 23.5-16.5. Organizer Martin Thoresen: "I think open source is the way of the future."

For years and years it was Fritz. Then, when top GMs analyzed their games for a chess magazine, sometimes they would mention other engines, like Shredder, Hiarcs or Junior. Then, around 2005, Vasik Rajlich's Rybka started to rule the chess engine scene like Kasparov had done in regular chess. Eventually, in July 2008, Chessbase had to admit that Fritz wasn't the strongest anymore, and started selling Rybka.

With one engine clearly standing out, the chess fan's life was easy. But, as Arne noted in his March 2010 piece, things are much more complicated these days. Several new engines have appeared on the scene, and we hear names such as Firebird and Stockfish. More importantly, many of these engines are open source and so don't cost a dime.

Shortly after Arne wrote his piece, in May 2010, version no. 4 of Rybka appeared on the market. Many experts believed that programmer Rajlich had managed to incorporate the best aspects of other existing engines, and had again created the strongest engine around.

Coincidentally in the same month Robert Houdart's Houdini saw its first release - the first stable release appeared on January 15th, 2011. This engine is not open source (i.e. the programmer's source code is not available), but free to use for non-commercial purposes. The choice of the name, after the famous magician and escapologist, is explained by programmer Houdart on the Houdini website:

The name Houdini was chosen because of the engine's positionaly style, its tenacity in difficult positions and its ability to defend stubbornly and escape with a draw – sometimes by the narrowest of margins. On the other hand Houdini will often use razor-sharp tactics to deny its opponents escape routes when it has the better position.

Houdini has been the subject of controversy as well. For example, it doesn't appear on the rating lists of the Computer Chess Rating List (CCRL) or the Chess Engines Grand Tournament (CEGT) due to possible derivation from the Ippolit series, allegedly plagiarized from Rybka. Houdart admits the influence of other engines on the Houdini website:

Without many ideas from the open source chess engines Ippolit/Robbolito, Stockfish and Crafty, Houdini would not nearly be as strong as it is now.

Update 20:58 CET
Houdini does appear on the CEGT list. Martin Thoresen wrote to us:

Just a note though: CEGT is now testing Houdini. They have it listed on the blitz page. Although they have only tested version 1.5 with 1 core, let's compare it to Rybka with 1 core:

Houdini 1.5 x64 1CPU 3254 (2300 games)
Rybka 4.0 x64 1CPU 3180 (4842 games)

Whatever its origin, one cannot deny that Houdini is a terribly strong chess engine. In his recent Crestbook interview, top GM Peter Svidler agreed:

Russianchessfan: If you had to pick one player to represent earth in a chess game vs. aliens, which active player would it be? You can pick different active players for the white and black side, if you feel it necessary.


And it looks like the latest version of Houdini is even stronger than the latest version of Rybka. In a recent, forty-game match between these two engines, Houdini won with a score of 23.5-16.5. Even more interesting perhaps was that Houdini showed almost frightening human chess, by playing positional pawn sacrifices in many games.

We asked organizer Martin Thoresen, a 29-year-old chess fan from Norway, a few questions:

Why did you organize this match?
TCEC is my first own chess project, so I have not organized anything similar in the past. I started the TCEC project simply because I love chess. I am by no means a strong player myself, but I have been following Magnus Carlsen since late 2005 when he played in the World Cup. Later in 2006 I joined the team of one of the largest computer chess engine rating lists and helped them test various engines for a couple of years or so. I gradually lost interest in the pure statistics and decided to create a project where the time control was longer and giving chess fans all over the world a chance to watch it all live on the internet.

Martin Thoresen

Martin Thoresen

The first TCEC event was broadcasted in the summer of 2010. Around that time I experimented a bit with both time control and various engine parameters, but it has since then evolved to what it is now - a more structured project with a fixed set of rules.

The Elite Match is the last event in a TCEC season. First there are regular divisions playing, much like how Tata Steel has divisions, except in TCEC they are played one at a time. The two participants for the Elite Match was selected from the top two finishers of Division I, or the highest division (in season 2 it is called TCEC A).

For all the current rules valid for season 2, I can recommend reading the information page on my website.

How do you do it? Can you describe the set-up?
Basically, there's nothing special about the set-up. The hardware used is a standard Intel core 980x, a six-core processor which is over clocked to make it even more powerful.

Then I set up the games in ChessGUI and each two minutes a scheduled ftp transfers the PGN to my website. The live pages on the website are automatically updated each 30 seconds, so they will display the new moves relatively fast when the PGN has been updated.

All this is of course not very relevant for the viewers, as it is perceived to be quite seamless. I have had some great help from Paolo Casachi who is the programmer of pgn4web (excellent tool for displaying chess games on the net)and Matthias Gemuh, programmer of the excellent ChessGUI.

Martin Thoresen's computer

Martin Thoresen's computer

There's a lot of discussion about copyright issues with all these different engines, some of them open source. What's your opinion on this?
My basic opinion is that I will continue to include the engines I want to include, unless any of them are proven to be violating copyrights or otherwise proven to be a clone in a court of Law. I am not a programmer so I am in no way qualified to say that "engine A" is a clone or "engine B" contains too much code taken from "engine C" and so on.

I think open-source is the way of the future - an engine like Stockfish is considered (by the community) to be perfectly legal and it is super-strong, too. I am sure quite a few programmers have learned a lot from reading the source code of this engine.

What, do you think, could be interesting for regular (tournament) players, regarding this match?
I would think that they could get some tips from the games played with (to them) familiar openings or by studying some of the endgames. If not any of this, then the overall quality of the games from a pure entertainment point of view.

Do you think top GMs should start using Houdini for their analysis? Can this make an important difference?
I would never recommend just using a single engine for analysis, but I am sure that with this latest version of Houdini any top player could get a few valuable, additional ideas of an opening or a position. Worth to note is that in the match, Houdini won 12 games and lost 5. If one were to study the patterns in the games Houdini lost you might get a good idea of which positions it cannot handle that well and vice versa for Rybka.

Inside Martin Thoresen's computer

Inside Martin Thoresen's computer

Can you make a selection of the most interesting games for replay in our game viewer?
First I will present game 1 of the match (note that all book moves end after the 10th move). Here we have a Sicilian where both kings are pretty exposed. Houdini first sacrifices a pawn, then another and then finally a third until Rybka suddenly finds itself in heaps of trouble. All this was done with no queens or any mating threats on the board. I still have problems understanding this game but it was surely one of the greatest games I have seen in a while. And that it wasn't played by a human makes it even more mystifying.

The second game is game 26 where Rybka in the end wins with 2R+2P vs Q. This is also a very nice game in my opinion.

The last game is game 37 where Houdini, as White, plays a move which really made the chat go warm. The move 43.c6!! was not really seen by anyone and I am still trying to grasp the depth of this move.

All in all there were several more great games, but also a few not-so-interesting draws.

Three games selected by Martin Thoresen

Game viewer by ChessTempo


Peter Doggers's picture
Author: Peter Doggers

Founder and editor-in-chief of, Peter is responsible for most of the chess news and tournament reports. Often visiting top events, he also provides photos and videos for the site. He's a 1.e4 player himself, likes Thai food and the Stones.


noyb's picture

Sorry, but there's no chance of this at all and the reason is quite simple. You can't make money! Nobody can afford to do all of this for free.

Delinquncy's picture

I agree with you. Unless you get enough government goons to confiscate money from others ("taxes"), all you are going to get is a bunch of computer hobbyists, rather than a professional class. Not that hobbyists are bad, Thoresen is a good one!, but progress will inevitably lag.

Rajlich took the Fruit code of Fabien Letouzey back in 2005, and via professional efforts raised it a few hundred ELO in his Rybka. Letouzey made a small attempt to go commercial, but he was really a guy who had better things to do than write chess engines. He even said that making an evaluation function would be a boring task. Rajilch evidently agreed, because he barely touched the one from Fruit, and hired Kaufman to modify the Fruit evaluation function for Rybka 3.

Houdini is donateware, which seems to work better than "shareware" two decades ago, but unless Chessbase signs him up, he'll still need a day job. Not like the "pros" really do much chess development though. You need to test 100s of thousands of games to measure minor ELO changes, so most "development time" is just waiting for the mass runs to finish. Guys like Shredder and Hiarcs who have been around for a decade probably spend as much time on handling customers as engine improvements, they know that just having a high ELO engine is not a selling point anymore.

Gavin's picture

I think that you have a misconception about what open source is. A good example to keep in mind is a car. As the owner of a car you can take apart the engine figure out how all the bits work put it back together in different ways. You can higher someone else to do this, even if they did not make the car in the first place.
With open source programs people still make money for the same reason that car mechanics make money it is much easier and safer to high a professional to solve a complicated problem than to do it your self.
Also I would like to point out that Linus Torvalds gets paid!

Guillaume's picture

You should try open source softwares. They're pretty good.

PP's picture

So? Open source software is all around you! I do not see why this would be impossible with chess engines.

One does not need a single paid program to have a fully working computer nowadays! This is first hand, because I'm typing from such an environment at this moment.

test's picture

Reasonable assumption but (as mentioned) reality has proven different. There are open source (and obviously free) versions of pretty much everything you can think of. (Most of which are even better than their paying versions, see this article for an example.)

Chess database: Scid.
Chess engine: Stockfish.

Office: OpenOffice.
Email: Thunderbird.
Browser: Firefox.

Photoshop: Gimp.
Media player: VLC.

Operating system: Ubuntu. (Linux)

Encyclopedia: Wikipedia.

What this shows is that there are many people with (a lot of) free time on their hands who are willing to release stuff for free, which is not really that surprising considering that there are now roughly 2 billion people connected to the internet.

Kint's picture

"Most" open source better than closed-source is hard to prove. But it IS FREE!

Many of your examples are nowhere near the original - see Openoffice vs office, thunderbird vs outlook, firefox? (chrome beats this), gimp vs photoshop, ubuntu vs. windows 7.

That said, I very much like the open source model and idea and I think in a few fields it's headed in the right direction.

Calvin's picture

If someone makes code available to anyone to view, they give up the right to copyright, regardless of international law. If they do not want anyone to view and use thier code, compile and release. Then if someone wants the code they must break into the application and steal the code, like one will do if they wanted money out of a bank. If you read several books on a particular subject, are you forbidden to use that information in your own writing? Ofcourse not, however you must cite your sources. So why does people seem to think reading code is any different and should be treated differently from information in a book. Isn't that code information?

Ken's picture

Open source does not equal free code to do with as you please. In addition to giving credit to the original author, there may be other stipulations in the license of the open source code in question. These licensing agreements explain (often in great detail) what may or may not be done with the code in question.

Felix's picture

Well, there are several points to make:

First, Mr. Thoresen is playing those games on a private computer and just publishes the games, everyone can do this. It's not a world championship in any way. You should definitely look at the rating lists, as a 40 game match is insignificant.
Second, the fact that the engine is free is due to the fact that the person who did the real work isn't being payed. It's like if you would post some links to a piracy website and say "oh, look, it's free! great!". If someone would take Chessvibes Openings and publish it with some extra comments and name it "Houdini's Openings", you wouldn't be too happy I guess. Even if the extra comments are good.

One of the reactions to the disassembling stuff is the Rybka Cluster which was launched 10 days ago:
The user gets a connection software and the engine runs on a super fast server. At the moment only the expensive 40 and 100 core offers are available, but a way to offer calculation time and number of cores more flexible is in work. That not only protects the code, but also offers a super fast engine for laptop or mobile users, who e.g. are at a chess tournament and would like to have a fast machine for analysis. That may be the future for the engine market.

Robert's picture

You sound very much like Felix Kling, the administrator of the Rybka forum.
Congrats for denigrating this well organized and very interesting tournament.
Congrats for not having looked at the rating lists, which all show Houdini 1.5 at least 50 points ahead of Rybka 4.
Congrats for using this comment area as advertizing space for the Rybka cluster.

Titu's picture

Sound? It's obvious from his link and the name Felix. He is not trying to hide anything as opposed to all the rybka cloners.

Graham B's picture

Every visitor from the Rybka Forum would tell you that Felix Kling would never use the forbidden H****** word that is banned from his own forum.

Felix's picture

You are the one hiding his identity, I linked to my website as pointed out by "Titu". I'm not the administrator of the Rybka forum, but the Rybka webmaster (I'm just a moderator at the forum). I also created the design of the Cluster website, so it's natural that I link to it I think.

Congrats to you for using those clones. Luckily there are still people paying for original software.

Guillaume's picture

Felix, what makes you think Houdini is a clone of Rybka exactly? On the outside, looking at the number of ply per second it looks like they are using an entirely different approach to select the best move.

Kint's picture

You mean the original, *DISQUALIFIED* software?

Luckily, indeed...

Delinquncy's picture

No, it *can't* be the real Felix. He would say "You should definitely look at the professional rating lists like CCRL (:DEL: and CEGT :DEL:), as a 40 game match is insignificant."

Now that Fabien Letouzey has announced his intent to contact the Free Software Foundation about Rybka stealing Fruit's code, the true author might finally get payed.

And Does anyone actually believe the code and sessions on Rybka Cluster are really secure and totally safe? It's not hack proof I suspect, and you also need to trust insiders like Vasik Rajlich, with the whole Fruit copying mess now coming down on his head. Just to list a scenario (hypothetical), If there really are any super GMs buying this, how much under the table would Radjabov be willing to pay to find out Kramnik's session details? Chess pros can be pretty paranoid about their contacts, and this one is too dodgy IMO.

nep's picture

Has Letouzey announced intent to seek the FSF about Rybka? Could you provide the link with this news?

alpha123's picture

He might have. He recently came back to computer chess and said that Rybka 1.x (and likely 2.x and parts of 3 and 4) are indeed modified from Fruit (2.1).

IIRC, the Free Software Foundation is the current copyright holder of Fruit 2.1. If I can confirm this, it wouldn't be hard to report the violation ( I would hope that Mr. Rajlich would simply release Rybka under the GPL and offer source code to anyone who asked, and no lawsuits would ensure.

Incidentally Felix, what's your response to CEGT deciding to test Houdini? They used to be mentioned in your Rybkaforum rules as a place to check for clones or not, and I couldn't help but notice you removed their link as soon as they started testing Houdini....

Delinquncy's picture

His response to ChessVibes publishing this TCEC article was to remove the "Independent Chess News" link from RybkaChess to here.

JM's picture

I got the feeling we've had this discussion before. Basically, the point of contention boils down to the following:

"My basic opinion is that I will continue to include the engines I want to include, unless any of them are proven to be violating copyrights or otherwise proven to be a clone in a court of Law." (Thoresen)

You may not agree with the guideline "innocent until proven guilty", but many people seem to disagree with you, considering it is one of the main principles of almost all legal frameworks in the western world.

On a sidenote: I think it is rather ironic that you include an advertisement (or something very close to it) in the same post in which you accuse Chessvibes of morally unjust behaviour.

Bert de Bruut's picture

Reverse engineering has been around for as long as there is technology, though it has become very noticeable in the present, as the distances in time and place have become increasingly small. After the introduction of the DVD, several years ago, the CEO of one of the innovaters, Philips (Electronics), stated that the exclusive (and lucrative) period after the introduction of something new, had already been reduced to three months. After that tiny window, clones already started to appear. Like the big corporations that invest billions in research and development, programmers and software developing companies can feel cheated when others almost immediately use their codes and even improve upon it. But it is a fact of businesslife, that they, like companies everywhere, have to live and deal with.

Delinquncy's picture

The main thing is that "computer chess" dudes have always considered their hobby a "competition", where ethics applies, so no RE. But when you start turning it into a business, then RE is totally legit. Once Rajilch sells his Rybka product to me, I can inspect it to find out what's going on, and more.

He's now under investigation from using too much Fruit code (Fabien Letouzey) when in 2005 Rybka first jumped almost 1000 ELO from a no-name to #1, and the major difference is that Letouzey provided source code for everyone, sparing Rajlich the effort of disassembly. He even said this himself back when: "Let me put it like this: every aspiring computer chess programmer has been very strongly tempted to try his hand at disassembling. When I started computer chess, Shredder was the king. We all wanted to know what he was doing. And nobody found out. Not Chrily Donninger. Not Frans Morsch. And not any of the then-amateurs." And "disassembling for purposes of finding information is legal and cannot be prevented. It would be legal (though incredibly hard) for someone to disassemble one of the commercial programs and publish his findings." His opinions are suddenly changed due to prevailing conditions it seems.

calvin amari's picture

I've heard about plans for the cluster and am glad to see to see it now available. A great idea.

Is there some analysis available for the layman to help him make some informed judgment about your assertions of intellectual property theft? I should stress that I have no opinion about this at this point and have no basis for such an opinion.

Has Rybka undertaken judicial action? Will it ?

If Houdini is very substantially based on Rybka, how is it someone can simply tweak it to make it marketedly stronger. It seems that the performance differential at this lofty level entails some more fundemental differences. Said another way, it seem logical to assume that if one could simply tweak Rybka marginally to get such a bump in performance, the Rybka folks would have done so.

There is much about this that remains very much in the dark to a chessplayer who is not also a computer geek (yes, such an animal actually exists) and perhaps even to chessplayers who are. I will say, however, that I strongly believe that Rybka and the other top commercial programs are amazingly cheap for what you get and, if there were compelling evidence of theft, I would not view Houdini as an alternative to patronizing commercial programs.

Delinquncy's picture

"Has Rybka undertaken judicial action? Will it ?"
Raljich specifically previously said no, but that was with Ippolit, because tracking down anonymous guys would take forever. Now that excuse is gone. The Fruit author (Fabien Letouzey) just reappeared after a Rip van Winkle rest of 5 years, and after finding out the facts of Rybka using Fruit code, is readying to contact the Free Software Foundation about legal action against Rajlich.

Rajlich is keeping his cards hidden, and has never given the slightest ounce of actual evidence about Ippolit, even though sites wouldn't even mention the "clones" for 6 months solely on his say-so. There was a long technical document published independently about Ippolit and Rybka, and the conclusion I got was that the Ippolit "cloners" had done RE on Rybka and then rewrote the whole thing rather completely. They gained 30-70 ELO, depending on whether you measure at fast time controls on a single cpu, or long time controls on a quad. Rybka's lackeys then said it was due to various non-chess reasons, like better time management, or removing imaginary time-consuming "code obfuscation" (guess *that* didn't work), or using a better compiler, or any excuse except reality. Which was I think, they dumped some of the more bloated parts of Larry Kaufman's evaluation to make it faster, and improved the pruning to make go more than a ply deeper in search w/o losing tactical skill. Houdart has now replicated their feat, and more. Progress ever onward!

calvin amari's picture

As I said, I remain neutral (mostly ignorant) on this question, but I have noticed that defenders of Houdini frequently allege that Rybka borrowed heavily from Fruit. This has been a constant drum-beat recently, but I do not see why. It almost seems as if the implicit argument being put forward is that it is ok if Houdini borrowed heavily from Rybka because Rybka borrowed heavily from Fruit. But this posture assumes that Houdini in fact borrowed from Rybka, no? Otherwise why is this allegation particularly relevant?

kingliveson's picture

Felix, it amazes me that you have the nerve to come here and advertise Rybka cluster on this site after removing chessvibes link from Rybka website because of this article.

You also talk about originality of a program, and the programmer not being paid...are you talking about Rybka being a derivative of Fruit, and Fabien being the original author?


kingliveson's picture

Felix, it amazes me that you have the nerve to come here and advertise Rybka cluster on this site after removing chessvibes link from Rybka website because of this article.

You also talk about originality of a program, and the programmer not being paid…are you talking about Rybka being a derivative of Fruit, and Fabien being the original author?


calvin amari's picture

That was my take on what was most notable. With amazingly powereful programs, no simplistic tactical errors, no forgetting of opening move order subtleties, etc., we still get more decisive games (17/40) than one would get at a top GM tournament. Fascinating.

Nakamura has done very well in utilizing a second who is very long in computer skills but far less so in OTB chess skill. Perhaps there is a role for Thoresen on GM Carlsen's team?

Delinquncy's picture

Yes, another fact is how often the evaluations differed, even quite radically, particularly as it refutes the party line that Houdini is just a "clone" of Rybka. In not a few games, Houdini preferred a piece rather than 3 pawns, and Rybka the opposite. Sometimes the piece won, sometimes the pawns, sometimes a draw. Houdini misevaluated a classical 3 vs 3 on K-side and a-pawn on Q-side rook ending to lose one. Both misevaluated some opposite bishops endgames. They value passed pawns rather differently too. The real question for Season 2 is whether Stockfish or Critter will gain enough before their bracket starts to beat out Rybka for the #2 spot.

Delinquncy's picture

"the first stable release appeared on January 15th, 2011. "
I think this is inaccurate. The most recent stability patch for 1.5 was released then. Houdini 1.03 was pretty stable from July 2010.

As noted by another, CEGT changed its policy 2 weeks ago, and rates Houdini now.
The TCEC was quite nice, and the 17/40 decisive games of the Elite Match were a bit of surprise. I agree "Chess is not dead yet", at least at these time controls!

Paul V's picture

Thanks for this!

Pretty sure that Chessbase will not follow up with a similar story :)
Can someone please explain how this hardware compares with those used in the WCCC? Why not turn the opening books completely off?


nep's picture

Without opening books, it is quite probable that you would have only a handful of different games. As the evaluation of the engines is deterministic (but parallel processing introduces random variation), starting from the same initial position would lead to identical games. What could introduce variation is parallel processing and random CPU time due to competition for computer resources with other programs.

JM's picture

I think you're looking at it from a slightly wrong angle.

Consider the following scenario: you and I play a game of chess in a tournament and you beat me. Afterwards, I become angry and start accusing you of cheating. I do not file an official complaint against you, but I do start whispering to many other participants about your bad behaviour. You would probably become quite annoyed with me and it is not hard to imagine that you (or your friends) would be slightly vindicative when it suddenly turns out that I have to face cheating allegations myself! This particular feeling of vindication by no means implies that you were, in fact, guilty of cheating.

In the case of the allegations of code stealing amongst Houdini/Rybka/Fruit things are quite the same. Defenders of Houdini feel amused that the Rybka team which has been accusing others without proof, is suddenly under scrunity itself. This does not mean that they feel the accusations against either Houdini or Rybka are in fact true.

KingTal's picture

There´s a mistake, Houdini appears on CEGT now. There are also some heating debates about derivates now, because it appears that the first Rybka was a derivate of Fruit.

Ray's picture

Great match to follow live (or semi-live)!
Looking really forward to future editions. Will we see a Rybka comeback?! Is the future really free access / open-source?
Cheers. Charley

P.S. Thx to Peter for this great site too. Keep up the good work!

noone's picture

It is pleasant that computers still make mistakes humans can understand. All backward moves are not good my dear engines.

Raj's picture

These games between software programs are bereft of feelings ! Nothing like top ranked players against each other - games would be richer and full of emotion !

ZL's picture

If two humans had played for example. game 1, you would be praising them for their positional insight

Zeblakob's picture

Nice report, nice games and great job by Martin. Those tourneo show that chess in not dead yet.

e4e6's picture

All will understand one day that open source is the only way in the future of online production. And Houdini is one of the proofs of this.

Bert de Bruut's picture

Mating treats are nice, mating threats are nicer.

szoker's picture

Impressive !

i always thought that free is better than commercial ! :P

john's picture

matches like this show how deep chess is and prove there is room for vast improvement still. If chess was just a draw for 3000+ elo players we might be disappointed, but no, still plenty of ways to win!

Delinquncy's picture

Pal Larkin is doing a similar Long Time Controls test. After 111 games, Houdini has won 35, and Rybka only 7 (and 69 draws), with swapped openings to equalize chances. Two of those were almost "book wins" for each. Complete domination by Houdini 1.5 over Rybka 4.

Correspondence matches though are heading toward all draws. I don't know of a recent top level FICGS one that had more than one decisive game from 8 or more games.

Bert de Bruut's picture

This sad message is vented too frequently, since it is still far to early to declare CC to have passed away from a "Death by Draws". For starters, FICGS is not top level correspondence chess, for that have to look at ICCF, for instance the Olympiad team events:

Few decisive results at the top boards indeed, where the tournaments are up to category 15. But this was always the case at top level CC, though not as outspoken as at the present. At the lower boards, with still strong tournaments up till category 10, there is not a trace of a "draw death" of CC. And mind you, CC knows no elo inflation, ratings are on average the same as ten years ago and linger some 100 points behind over-the-board chess (only two active CC-players are over 2700 atm, though one more virtually is). That rating gap makes up for a difference of 4 more categories, so in relation to OTB the categories mentioned above are 19 and 14 respectively.

Delinquncy's picture

I was talking about *matches*. Round robins and team events (olympiads) are comparative garbage for WLD stats and draw death, as everyone has different goals. After the first few results come in, those who do well put in a harder effort, and those who don't often simply seek to finish their games quickly. Usually it comes down to whether a tail-ender will care enough to try to optimize results with those in contention. Matches get rid of all this, for you are put head-to-head rather than in a mixed bag, and you don't take unnecessary risks. This is where you can truly measure Death by Draws.

And it makes no sense to compare ratings on any absolute basis. I don't see any evidence that "top level" ICCF guys could beat someone like Eros Riccio *in a match*, though they might know better how to beat the guys who overreach in an round robin. FICGS matches have become typically +1-0=7 or +0-0=8 and a tiebreak. The last major match with both sides winning games is from like 2008.

Just looked at your data. On Board 6 (lowest), among the top 8 players (of 13), there are 23 draws and 5 games in progress. ZERO decisive games!

Bert de Bruut's picture

Eros Riccio is an ICCF guy himself....

Bert de Bruut's picture

And I examined some of the FICGS data as well. For one thing, it seems you were not quite accurate about the last FICGS-match with more then one decisive game in it. In fact, the second "FICGS world championship" match was decided not too long ago with the remarkable score 12-0:

Zero undecided games is worth something perhaps, though there is little reason to call such a match one for the "world championship". Obviously there is a major difference between the levels of play in FICGS and ICCF, to begin with the rate of plays of 30 days plus 1 day increment for FICGS versus ICCF's usual 50 days per ten moves.

John O's picture

Game 1 (Sicilian) the computer plays the same variation the computer who first beat Kasparov played. Interesting.


Latest articles