« Public choice and the Nazis | Main | The hottest female professor in the United States? »
Uh-Oh, continued
Thus, out of 5 games – 78% of GM Kramnik’s moves match with the first line of Fritz9 [an advanced chess computer].
That is from an official letter written by Topalov's assistant. Of course just because someone says it, that doesn't mean it is true.
Posted by Tyler Cowen on October 4, 2006 at 08:43 AM in Sports | Permalink
Comments
Why choose that particular chess program? Could you find a statistic like that for every player in the tournament?
Posted by: josh at Oct 4, 2006 8:54:37 AM
I would think that the world's best chess player would make the optimal move over 80 percent of the time, and the world's best computer program would make the optimal move over 80 percent of the time. If you assume 90 percent each, and they are independent, they would make identical moves 81 percent of the time. So this is not exactly smoking-gun evidence.
Posted by: Arnold Kling at Oct 4, 2006 9:36:30 AM
I was thinking the same thing as Kling. The best chess players in the world are choosing the dominant strategy, so you should expect a decent program's recommendations to be highly correlated with the same moves of these players.
Posted by: Jason Voorhees at Oct 4, 2006 9:39:08 AM
I don't know the right answer, but in my view this should be easy enough to test. We can look at the other games of Kramnik, or other top players, and run the same test. 78 percent seems quite high to me, especially in complex strategic settings. But it could be that the Topalov team is cherry-picking results by letting the computer think for varying amounts of time until it selects the moves Kramnik played.
Posted by: Tyler Cowen at Oct 4, 2006 9:41:59 AM
Or, perhaps Kramnik trained against the Fritz9 program, and was adopting the program's strategy (consciously or unconsciously).
Posted by: Independent George at Oct 4, 2006 9:54:08 AM
I would like to see how the same statistics apply to Topalov moves.
Posted by: dsVasques at Oct 4, 2006 10:20:31 AM
While cheating with Fritz would be an extremely unfair advantage, at that level, they're better than Fritz is overall anyway. Just something to keep in mind.
Posted by: Jewish Atheist at Oct 4, 2006 10:23:52 AM
Jewish atheist, first of all what you are saying is not obvious though probaly true.
Remember that Fritz has tied with Kramnik in a match.
Second it would still be a huge advantage to use Fritz to check your proposed move for tactical blunders, an area where humans, even Kramnik, are worse.
Finally, any one besides me who finds it annoying that they call Fritz a chess computer insetad of a chess program?
Posted by: Johan Richter at Oct 4, 2006 10:42:40 AM
Wow, Fritz is really good! 78% of the time it chooses like Kramnik did!
Posted by: luispedro at Oct 4, 2006 11:02:29 AM
I don't know why they just can't solve the problem by searching the players for electronic devices and forbid entry into the bathrooms by anyone else during the game (checking the bathroom before and after)?
This way no electronic devices could be smuggled in and you could be sure of it.
I know this is hard-core, but since electronic devices like handys or PDAs already have significant calculating power, they can be useful assets for cheating players.
However, if he hasn't a good computer installed in the bath, wouldn't it underperform due to lack of time to calculate? (I am no pro in chess programs and their efficiency, but high-quality programs usually take some time to calculate a few moves ahead through all possible moves...)
Posted by: Max at Oct 4, 2006 12:04:16 PM
Kramnik doesn't have to have access to Fritz. He only needs communications with someone who does. Even hand signals could be used to pass the move.
That said. I don't believe he was or is cheating. His moves are often going to match Fritz. The bathroom stunt is probably just eccentric behavior - hardly new.
An analysis of Fritz v. present or past masters would be interesting. But someone would have to come out highest. And that someone is almost certain to be an active top ranked player.
What might be more revealing would be the trend over decades. Did top ranked play steadily get better? Or did it rise and ebb? Did players agree with Fritz more as the game progresses, or less?
Posted by: K at Oct 4, 2006 12:58:11 PM
Some of the comments here reveal that some writers haven't understood what this is about - ie. "applied game theory" in the form of chess engines (defined as computers that can "learn" moves and strategies).
There are only a few chess engines that can challenge a GM: Fritz, DeepJunior, Shreddar and Rybka are the most famous. They all use mathematical devices to compute the material on the board and the position and they also use a function to specifically calculate the safety of the king.
If you let an engine like Fritz or Shreddar "think" about a position for an hour it will only focus on certain developing lines while disregarding others. If you let it analyze a position for an infinite amount of time it will still disregard developments that it finds unreasonable based on its "understanding" of material and positional advantages. It is therefore possible for a GrandMaster to win over a computer with a line that might include a -5 sacrifice but which, 20 moves down the line, forces a compensation.
However, humans are not infallible and chess scholars of all levels tend to value steady development (getting the pieces out in apple-pie order), analytical advantages (to reach positions that are easy to control) and strategical positioning (for example any line that might force a zugswang or establish an outpost). Chess engines, on the contrary, only stay with the equations they've been given. This is why we now have completely new development lines in the chess world that would otherwise never have been addressed. "Computer moves" - as they are called - stem from these AI computations. Humans would not be able to come up with them without the help of computers.
Now to the question raised about Kramnik's suspected moves. Yes, it is true that Grandmasters use the same developing moves as normal computers. (In game seven between Kramnik-Topalov they played a Queen's Gambit Accepted.) Nothing special about that. It's a very logical opening (that has been around long before computers came about) with plenty of theory behind it. Modern chess engines can verify that the opening indeed is sound. But throughout a game there are almost always deviations from the main line. And then deviations from the deviation (minor line). And after watching Kramnik's play in several games Topalov's team couldn't make sense of the moves. Not because Kramnik outsmarted them - because he obviously also made blunders - but because some of the moves simply didn't seem like human moves. When Topalov's second argues that Kramnik is playing Frits' preferred line it is reasonable to believe NOT that this means the standard QGA opening but that Kramnik in almost every critical phase of the game did choose Fritz' number one line. Yes - if Fritz is allowed to think for a week it will come up with several preferred lines but not in positions where the amount of limitations (due to positional and material restrictions) are plentiful, as often is the case in a complex middlegame.
Please don't misunderstand me here - I am not saying that Kramnik is cheating. I am merely saying that given the complaints by Topalov's team we have to look at what is plausible, and not misinterpret the entire debate about computer assistance.
Posted by: Joakim at Oct 4, 2006 10:56:55 PM
I think it's only fair that someone run both Topalov and Kramnik's moves through a documented procedure against a chess engine and show the results while varying certain assumptions.
Posted by: Paul at Oct 4, 2006 11:33:14 PM
Here's a detailed analysis of Game 6 with Fritz 9.
Posted by: Raghav at Oct 5, 2006 4:55:44 AM
数控机床数控机床。。数控改造。。
数控机床。。
数控机床改造。。
数控机床仿真软件数控机床改造。。
数控机床。。
数控机床。。
数控机床。。
数控车床数控车床。。
数控车床。。
华中数控车床。。
大连数控车床。。
无锡数控车床山东数控车床。。
南京数控车床。。
数控车床。。
数控车床。。
广州数控车床数控车床。。
广州数控车床。。
华中数控车床。。
数控车床。。
数控车床北京数控车床。。
数控车床简介。。
教学车床。。
教学数控车床。。
数控铣床南通数控铣床。。
数控铣床。。
数控铣床。。
数控铣床。。
教学数控铣床教学铣床。。
数控机床实训设备。。
加工中心。。
加工中心。。
加工中心加工中心。。
微型加工中心。。
数控加工中心。。
加工中心。。
加工中心加工中心。。
加工中心。。
上海加工中心。。
加工中心
Posted by: 南京北春 at Aug 21, 2007 3:05:32 AM
Mabinogi Gold
Mabinogi online gold
Mabinogi money
cheap Mabinogi gold
buy Mabinogi gold
Posted by: aion kina at Mar 20, 2009 8:39:06 PM
Flyff gold
flyff Penya
flyff money
buy flyff penya
cheap flyff penya
cheap flyff gold
Posted by: aion kina at Mar 20, 2009 11:29:34 PM
