Boris Gelfand’s victory in the Candidates matches gives me an opportunity to present a little known method of choosing your openings. Now most people don’t give him much of a chance against the reigning World Champion, Vishwanathan Anand, in Megabase 2011 Anand has 16 wins compared to Gelfand’s 6 with 45 draws. Yet most of these wins came against Gelfand’s Sicilian Defence where Anand won 9, drew 11 and didn’t lose any.
Take these Sicilians away and a different picture emerges, Anand having 7 wins to Gelfand’s 6 with 34 draws. This is admittedly distorted because Anand then loses a lot of his Whites, but in any case it’s certainly a lot closer. So it looks to me as if Boris shouldn’t play the Sicilian, assuming of course that Anand opens with 1.e4.
Does he have an alternative? Well he’s played the Fort Knox French (1.e4 e6 2.d4 d5 3.Nc3 dxe4 4.Nxe4 Bd7) against Anand in a few games without having lost even one. There are also other solid lines which lead to similar little centre positions, such as the Caro-Kann.
This kind of statistical analysis can be very useful as it can provide a reality check for our often flawed perceptions. I remember performing this on my own games and discovered to my horror that the Modern Defence (1…g6 against anything) was not getting me great results, particularly against strong players. This was quite a surprise because my ‘reputation’ with this opening had convinced me it was one of my best weapons, mistakenly as it turned out.
There are wrong ways to do this as well, for example a player may give up an opening soon after starting to play it because a couple of losses persuade him that the ‘stats’ are bad. But in this case other factors can be in play such as the sample size being too small or the test being done whilst you’re still early in the learning process. So it’s good to learn an opening reasonably well before using it in tournament games and trying it out in friendlies on the internet or elsewhere. After that you should get a decent sized sample of real games under your belt, at least 10-15 and ideally more. If the results are still dire then it might be worth reconsidering its place in your repertoire.