When Statistics Lie

A popular and useful function on chess database programs is to check the statistics for particular lines. I like it too but would like to add a word of warning. There are a number of cases in which they can be very misleading.

The first case concerns sample size. If there aren’t too many games with a particular line the results can be heavily skewed by a few accidents, something that becomes much worse if the players concerned are low rated. Of course if your sample size becomes large enough the accidents are likely to get evened out (White and Black making roughly the same number of mistakes) after which we are hopefully left with a reflection of the opening’s value.

The second way that results can be skewed is when the players playing White have a very different average rating to those playing Black. How does this happen? Well solid openings are often employed by underdogs in the hope of making more of a game of it or if they don’t have sufficient faith in their usual repertoire. Unfortunately for them they often end up in more trouble than if they’d just played their usual game, the switch to ‘something solid’ also landing them in a position they just don’t know.

Openings that fall into this category are the London System (1.d4 followed by 2.Nf3 and 3.Bf4) for White and the Queen’s Gambit Declined (1.d4 d5 2.c4 e6) as Black. These are very solid openings but you still need to know what you’re doing.

The third category is somewhat unusual, but some sharp openings might have done quite well until hit by a terminal theoretical novelty (TTN). It could be that only one or two games would feature this new move because its appearance stopped everyone else playing the variation that got terminated. On the other hand the historical results will still appear in any search, making it look quite promising if you go on statistics alone.

Here anyway is an example, take a look at the following game:

Nice, huh? And checking the statistics for the line 1.e4 e5 2.Nf3 Nc6 3.Bc4 Nf6 4.Ng5 on Megabase 2013 shows that 4…Nxe4 the third most successful move (behind 4…d5 and 4…Bc5), weighing in with a respectable looking 42%. Could this be a way to avoid more theoretical lines? Unfortunately not as White has 5.Bxf7+ Ke7 6.d4! after which 6…Nxd4 7.c3 is very strong.

So statistics can be useful but they should be used with discrimination. There are cases when they lie through their teeth!

Nigel Davies

This entry was posted in Articles, Nigel Davies on by .

About NigelD

Nigel Davies is an International Chess Grandmaster living in Southport in the UK. The winner of 15 international tournaments he is also a former British U21 and British Open Quickplay Champion and has represented both England and Wales on several occasions. These days he teaches chess through his chess training web site, Tiger Chess, which has articles, recommendations, a monthly clinic, videos and courses. His students include his 14 year old son Sam who is making rapid progress with his game.