By Steve Giddins
Several friends have reacted to my blog post regarding the tedium of events in Moscow. A brief flurry of excitement on Monday was followed by another bore-draw on Tuesday.
I should perhaps make clear that I am not blaming the players. They are just reacting to the circumstances they find themselves in, and are doing what they think they have to do, to have the best chance of winning the match. Likewise, nor do I give any credence to the claim that all would be rosy, if only the match organisers had imposed the infamous "Sofia rules", to prevent early draw agreements. Having Sofia rules in place for the first two games in Moscow would have forced the players to play another 20 or so moves, before agreeing a draw, but all that would have meant was two largely contentless 40-move games, instead of two largely contentless 20-move games.
It is true that the shortness of the match contributes to the problem, by making the players even more cautious than they would otherwise be. With so few games, a player cannot afford to risk even one loss. Therein lies part of the problem. But even that is only a small factor. The real problem lies elsewhere.
That problem is that computers are killing the game. They have already killed correspondence chess, in all but name, and now classical chess is heading down the same twilight path to oblivion. The computer is now so powerful, that it becomes impossible to out-prepare another top player in the opening. In pre-computer days, Kasparov could analyse so much better than the other top GMs that he could routinely uncork novelties that refuted entire opening variations. Nowadays, though, that is just impossible – everybody is analysing the same opening lines, using the same powerful computers and programs. As a result, everybody is coming to the board, with much the same opening preparation, with the result that nobody can get a serious opening advantage any more.
Imagine the following experiment. Lock Anand and myself in separate flats, for a week, on our own, to analyse a certain opening variation. Even if I work every bit as hard as Anand, or even harder, at the end of the week, he will have analysed the line much better than me – he sees tactics faster, his positional judgement is better, etc. There will be a large gap in the quality of the analysis we each produce.
But now repeat the experiment, only this time, give each of us a powerful laptop and the latest version of Rybka. By the end of the week, Anand's analysis will still be better than mine, but I can assure you that the gap will be very much smaller, especially if the line we are analysing is something fairly sharp and tactical. Despite the enormous disparity in talent and ability between myself and Anand, if I put in the work and use the computer fully, he is not going to be able to out-analyse me to any huge extent, thanks to the levelling effect of the computer.
And this is the crux of the problem in world championship matches. There, we are talking about a very small disparity in strength between the players, which makes the problem even greater. Against me, even if Anand gets nothing from the opening, he will still be able to outplay me over the board, and win. But he cannot do that to a top-class GM, who is only marginally weaker than himself anyway. If he gets nothing from the opening, he will have huge trouble beating a player like Gelfand, and vice versa. The result is a whole series of effectively contentless games, where the players are just checking each other's computer-aided preparation. Once in a while, they will hit on a gap, and get some advantage, but most of the time, there will just be what we have already seen in Moscow – 15-20 moves of preparation, 4-5 more accurate moves, a dead position, and a draw.
So, what is the solution? Sadly, I don't think there is one, at least not without abandoning traditional chess, in favour of Fischer-Random, and I hardly know anyone in the chess world who wants to see that (I certainly don't). It grieves me to say it, but I think classical chess is in its last days.
The worst part is, nobody has really come up with a solid counter-argument to Giddins, and it's only going to get worse. Not only are computers going to get more and more powerful - true, Moore's law is coming to an end, but computers ARE going to become more and more parallel, a situation that chess engines can take advantage of far better than other programs - but also, engines are going to get stronger and stronger. It's likely that Rybka 5 and Komodo MP are going to up the ante even further.
I think that to really drive the point home, we should get one really powerful machine - say, the Rybka Cluster - and have it play a clock simul against the top 10 GMs in the world.
> have it play a clock simul against the top 10 GMs in the world
Sounds great, but I suspect it would cost $10 million or more to get them all to participate! The publicity for Rybka would be terrific; I doubt the Rybka team would require any compensation at all.
There were plently of novelties in the Kramink - Anand match. Furthermore, they went into positions that were equal but had lots of play in them. Whilst things have moved on with comps in 4 years, I don't think it that is the main reason for the difference. The difference is that in 3 out of the 4 games so far Anand and Gelfand are playing openings lines that are level and lead to sterile positions.
But there is no doubt that comps are intruding more and more and opening lines are getting exhausted and it is harder and harder to spring meanigful surprises. Giddins states that he does not want to move to 960 and that other chess players don't either. This is not surprising as pro chess players have a huge personal investment in opening knowledge, and the middle game and even endgame structures that arise from certain openings - it would be pretty painful for them to have this rendered almost meaningless by a wholesale shift to 960. But as a spectator, I would love to see more 960 chess, and would regard it as an excelllent test of chess GMs and any 960 world champ would be a real champ in my eyes, and more so that a champ at trad chess these days.
I would also like to play 960, but it is simply hard to get a game.
In a normal tournament, everyone knows that persistently cautious play cannot possibly result in a tournament victory unless a lot of luck intercedes. In a twelve-game world championship match losing even one game is oftentimes an unrecoverable disaster, because the player with a scoring advantage will do everything in his power to close out the match with a string of lifeless draws. As a result, calculated risk-taking in the form of novelties or ultra-rare sidelines is best deferred to the end-stages of the match. I'm confident this match will follow that pattern; twelve quick draws leading to tiebreaks would be an unthinkable disaster for human chess, the result not of computers but of human game theory.
>If the assertion is that the explosion of opening theory and computer analysis is killing chess all I can say in response is that no human being could possibly be familiar with or memorize more than a tiny fraction of the theory that has been and is being created.
You underestimate GMs ability to memorise the openings that they play and analyse. They incredibly rarely stumble into stuff that has been previously shown to be poor - that is because they remember relevant past games. Past theory is relevant to how chess is played today and is definitely slowly taking the creativity out of chess - the balance of memory verses working stuff out OTB has shifted dramatically, and the stuff that is remembered is now heavily copmuter generated/assisted as opposed to worked out just by humans. There is still plently left to explore but it is much less than there was 10 or 20 years ago. There are many quotes on the internet re this, professional players have expressed this view to me directly, and I heard second hand from a good friend of his that one English GM gave up playing largely because the balance had shifted so much.
Of course it is possible to play fighting chess ragardless of all this - Carlssen has demonstrated that time and time again.
>The problem here is entirely one of risk aversion, which translates in practice to safe, very specific and deeply prepared opening lines.
This I agree wtih and seems to be pertinent to this match which is in marked contrast to the Kramnik-Anand match. Game 5 today was good example of risk aversion. Anand played cxd5 after only a short think. Perhaps he misanalysed something at high speed, but if he really wanted to win he would have looked very deeply at cxd5 and assured himself of with concrete lines. It seems more likely he chose the risk free approach rather than enter the muddier waters of exd5.
But again I think comps have had an impact in this regard. The level of certainty and depth they bring to novelties layered on top of the fact that theory is further into the game before the novelty is sprung means that there is less play after novelties today - i.e. less chance to recover by outplaying your opponent after a strong novelty, and less chance that their novelty is faulty (which many were in the pre-comp days)
>As a result, calculated risk-taking in the form of novelties or ultra-rare sidelines is best deferred to the end-stages of the match.
But the history of matches has often seen the novelties uncorked early, in both the pre and post computer era. In the Kramnik-Anand match Anand uncorked novelties early in the match. So I hope you are right and Anand is not using Grischuk tactics -zero risk at normal speed and relying on his superior blitz skills to win the match.
Regardless, it is so far the dullest world champ match I can remember, and that is due to human elements far more than comp issues.
Chess is not a boring field of study. It's an ongoing black comedy.
>You underestimate GMs ability to memorise the openings that they play and analyse.
They could then memorize the engine moves played by engines outside their preparation and play like an engine, which would be an advantage as engines are much stronger than humans. The problem is that chess space is so huge that no matter how many positions they analyze, they're going to be outside their preparation soon enough in where their ability to memorize doesn't matter.
That's why GMs can only analyze a tiny fraction of theory, no matter how good is their memory.
In the olden days, by the time theory had finished the position would often be level but with plenty of play left in it. There were very few lines 40 years ago that were analysed so far that a draw what almost certain (it is hard to think of many examples).
Nowadays there are still plenty of lines with play in them, but many lines have been analysed so deeply that by the time theory stops the position is not only level but an almost certain draw (we are all aware of hundreds of examples).
Of course this change has happened gradually and some of it before comps. But most of it has been in the past 20 years. It is not just down to engines of course but also the internet and databases. But engines have speeded up analysis and given it a certainty it previously lacked, and I think this has been a massive impact.
> Nowadays there are still plenty of lines with play in them, but many lines have been analysed so deeply that by the time theory stops the position is not only level but an almost certain draw
So then, why can't the GM play such a line against an engine and draw a match against it? The engine would destroy him, which means the engine is doing something right in the opening that these GMs aren't doing against one another.
If the memory is so good and they can predict their opponent's theory moves then they could be playing against the engine, seeing what lines defeat them and go for them against the human. Instead, they are willingly playing into the most drawish theory and I don't think it's the fault of how strong the engines are. Engines aren't only good at refuting theory to a draw, engines are also very good at finding critical positions where the opponent is likely to blunder too. In fact, with engines so strong the playable complex lines you can get into increase as the openings with playable sacrifices are more evident with the way engines can defend them (even if they are drawn, it's unlikely both GMs had analyzed the same unplayed sacrifice). If the GMs don't find the lines, it's their fault for not knowing how to take advantage of engine's strengths.
>So then, why can't the GM play such a line against an engine and draw a match against it? The engine would destroy him, which means the engine is doing something right in the opening that these GMs aren't doing against one another.
> Instead, they are willingly playing into the most drawish theory and I don't think it's the fault of how strong the engines are.
You have indirectly answered your own question. If a comp book maker headed towards drawn positions rather than unbalanced equal positions (or possibly even better posns) then a strong GM could draw a lot of games vs a comp. For example Anand could hold the black sude of positions that arose after about 20 moves in games 4 and 5 of this match against a comp with little difficulty.
Perhaps I should be clear - I am not blaming the passive play in this match on comps. As I said in my first comment in this thread I think Giddin's article overstates the issue, and as I said later in the thread the problem with the match is very largely human made not computer made.
Nonetheless I stand by all my comments about how comps are affecting GM play.
1. As you say, it is possible to find opening lines (for white) where you quickly get a playable equal or unclear position without too much theory. However, people sometimes seem to forget that these players' goal is not to reduce the number of draws, but to win the match. From Gelfand's perspective at least it is much preferable to play opening lines with white where the chances of winning are slim and the chances of losing virtually zero. It is a little different for Anand who is the better player. However, it is understandable that he looks for something in more common lines first, checking Gelfand's preparation. Note that he did play a very sharp line in his second white game and was almost successful. (Just to those who criticize the players: What would you have done in Anand's situation in game 6? You play 1.e4 and are completely surprised by Gelfand's Sveshnikov. Which line do you play to retain winning chances?)
2. As is mostly the case, both extreme positions are most likely not correct. Carlsen for instance demonstrates again and again that you can beat top players without very deep opening preparation, by just cleverly picking lines which don't offer an advantage but are not analyzed to death, either. On the other hand, you can be sure that what Carlsen does is really hard. Firstly, it is not that easy to get these types of positions. And secondly, to be successful in them, you have to be definitely stronger than your opponents. (Grischuk said something along these lines: It's unfair to blame us for all of these draws just because Carlsen wins so often. The rest of us are simply not capable of doing what he does.)
But in Anand's shoes, I doubt Karpov, Kasparov, Topalov or even Karmnik (who BTW I admire greatly and is unfairly criticised as Drawnik) would have played cxd5 in game 5 against a markedly weaker player, at least not as quickly as Anand did. They would have easily seen cxd5 offers nothing and at least carefully assessed exd5. They might even have played Qxd5 which whilst fairly level at least leaves some play in the position. To play cxd5 so quickly really was a bit feeble. Added to which Kasparov and Topalov (and possibly the other two) would never have played c4 and g3. You could add Spassky and Fischer to those lists but I guess that is open to the arguement that they didn't play in the cimputer era. Where is Anand's confidence?
Kasparov thinks Anand lacks motivation. Perhaps the fear of losing is much greater for him against Gelfand than it would be against a stronger opponent - the embarrassment factor of losing the match would be rather high. In such a short match it is possible to play a surprising defence almost every game as black. Is Anand going to duck the critical lines every time?
> In such a short match it is possible to play a surprising defence almost every game as black. Is Anand going to duck the critical lines every time?
I doubt that this is Gelfand's plan (he didn't change his opening against 1.d4, either). But if it is, then he cannot have analyzed (and memorized!) everything equally well. So I think if Anand realizes that Gelfand has prepared several lines against 1.e4, he will take more risks.
By the way, which critical lines did Anand duck? He played either sharp or at least quite fashionable lines in each of his white games.
>By the way, which critical lines did Anand duck? He played either sharp or at least quite fashionable lines in each of his white games.
Well I would regard c4 and g3 as ducking the issue - its been played a bit but is known to give white very little. But your point is fair. Game 1 avoided the sharpest lines and was untimately uninspiring but was a reasonable attempt that Gelfand played very accurately against and Game 3 was just fine.
I thought Kasparov's comments were characteristically undiplomatic and self-aggrandizing, and yet undeniably true. You hate to hear him say what he says because Anand is such a pleasant guy, almost the opposite of Kasparov in that regard, and the match isn't over yet so it is too soon to judge. But yes, who can deny that at the world championship level both champion and challenger have to play like a champions for one of them to deserve the victory? What would you think of a world champion boxer who spent half of a big fight clinching and the other half demonstrating impeccable defense?
> it does seem to me that Anand is the party who is more often increasing the probability of a drawn outcome in game after game
With black you mean?
It's very sad that millions of dollars go into this and the players don't even try to make it interesting.
Something must change imo, either add rapid game with reversed colors after every game or get rid of classical altogether and just play 4 rapid games a day.
There are plenty of interesting positions in chess which computers with their much lower draw rate shows and which players like Carlsen and Nakamura keep demonstrating. The strategy of "let's check his preparation and if it's ok shake hands" is shame for chess :(
> computers with their much lower draw rate
Don't know where you get that idea. The frequency of draws can be visualized on three axes:
quality of the competition
equality between the players
competition rules or external considerations that incentivize or disincentize risk-taking
Thus if the players are of very high quality and practically equal in strength, and the match is of relatively short duration (12 games is very short), at classic time controls and for a world title, you will likely see high draw rates between human players.
Conversely, low-quality players playing blitz in a park who are mismatched will very rarely draw, and when they do it may be an accidental draw, as in a stalemate that could have been easily avoided.
What I see in my databases is that low-grade players have draw rates of 10-15% while grandmasters are more in the 35-45% range. Computers, on the other hand, if equally matched in hardware, software and opening preparation, can produce draw rates of around 60% nowadays at the highest levels.
It's true I was suggested by data I glanced over (CCRL list) not realizing how weak many competitors are thus reducing the drawrate of top competitiors. In any case even 60% draw rate is much lower than what Anand Gelfand achieve with their current mindset and computers don't blunder and remember all the theory there is to remember.
So entities much stronger than Anand and Gelfend, which has unlimited memory for opening theory draw only 60% against each other in such circumstances blaming what is happening in Anand-Gelfand match on computers is imo misguided. Their attitude is to blame and system is to blame for promoting the attitude they display.
I think if these draws continue to be a problem in future championships, they can always go to some type of decisive rapid game following each drawn classical game.
>For these reasons, the championship match should be restricted to extremely attractive women. If the match is tied, the girl with the biggest tits wins!
Classic and I for one can see no faults in this radical idea
Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill