LeeRoth wrote on 12/20/06 at 15:33:32:


Of course, you also need to understand why you are not doing well. Be careful about abandoning an opening simply because you are not doing well in it. You might simply need to put in some more work on it.

OK, now on the positive side of statistics. IM Greet has published a book on the Worrall Variation of the Ruy Lopez: 1.e4 e5 2.Nf3 Nc6 3.Bb5 a6 4.Ba4 Nf6 5.Qe2. Now I know almost nothing of this, except that the idea is to play Rd1, c3 and d4. So I glanced in my old Pachman book (1980 or so) and found b5 6.Bb3 Be7 7.o-o o-o 8.c3 d5, where IM Brett recommends 9.d3 (I asked him). According to my database Black scores best with 9...Bb7 and especially 9...Re8 (percentage, from White's point of view, less than 50%!). This does not mean, that we should put 8.c3 or 5.Qe2 aside. But guess which chapter I will study first, when I have that book in my hands?
Use statistics as a guideline, nothing more. And remember Polugajevsky's statement: no variation is as strong as one, that's considered bad, but made playable again by an improvement.

I have found the raw game statistics useful when confronting a position that I know nothing about, usually in the later stages of the opening, when there is a branching between two different plans.  Although often, as SF points out, progress has been made and statistically promising lines are no longer viable, this is still a good way to get a sense of the recent development of a line; the most popular lines were the main lines at some point.

The use of statistics to talk about openings in the first couple of moves is pretty much useless, at least as far as using it as a basis for opening selection.  When only two moves have been played, I know more about the position than do databases, and choosing between main lines in a defense (to use MNb's example) is better done by selecting positions that one enjoys playing, or where one has some new ideas to invest than by going by the numbers, which can always been made insignificant by one new move...

The information provided by database statistics is much less useful than expert opinion.  Trying to pick a defense by selecting the most popular lines after 1.e4 is likely to lead to disaster.  I remember the admonishment by I.A.Horowitz in his Chess Openings: Theory and Practice not to try and go through the book looking for forcing lines which were always +/- for White or = for Black -- that advice remains true today.

Just for fun, here's a game played by my database: amogst moves played at least 1% of the time, best scoring move was selected (up to nearest 1%); in the event of a tie, the most popular move was selected.

Motylev,A (2570) - Badea,B (2460) [B06]
Ciocaltea mem Bucharest (9), 12.03.2001

1.e4 g6 2.d4 Bg7 3.Nc3 d6 4.Be3 a6 5.h4 h6 6.Be2 b5 7.h5 g5 8.a3 Bb7 9.f4 gxf4 10.Bxf4 Nd7 11.Nf3 c5 12.d5 Qa5 13.0-0 Bxc3 14.bxc3 Ngf6 15.a4 bxa4 16.Nd2 Qxc3 17.Rxa4 Nb6 ½-½

Note that I transposed to the game from the move order: 1.d4 g6 2.e4 Bg7

With a 5% threshold for selecting moves:

Antreasyan,E - Nilsson,N (2235) [E70]
Copenhagen op2 Copenhagen (1), 1993

1.d4 Nf6 2.c4 g6 3.Nc3 Bg7 4.e4 0-0 5.Be2 c5 6.d5 e5 7.h4 h5 8.Bg5 d6 9.Nh3 a6 10.f3 Nbd7 11.Nf2 Rb8 12.a4 Re8 13.g3 Nf8 14.Bf1 N8h7 15.Be3 Rf8 16.Bh3 Ne8 17.Bxc8 Qxc8 18.Qd2 f5 19.f4 Nc7 20.Nd3 b5 21.fxe5 bxc4 22.Nf4 Bxe5 23.Nxg6 Bxg3+ 24.Bf2 Bxf2+ 25.Qxf2 fxe4 26.Ne7+ Kh8 27.Ng6+ Kg7 28.Nxf8 Qxf8 29.Rg1+ Kh8 30.Qxf8+ Rxf8 31.0-0-0 Rf4 32.Rg6 Ne8 33.Rdg1 Nef6 34.Rg7 Rxh4 35.Ra7 Ng4 36.Nxe4 Nhf6 37.Nxf6 Nxf6 38.Ra8+ Kh7 39.Ra7+ Kh6 40.Rxa6 Rg4 41.Rxg4 hxg4 42.Rxd6 Kg5 43.a5 g3 44.Re6 g2 45.Re1 Nxd5 46.Rg1 Nf4 47.Rxg2+ Nxg2 48.a6 1-0

5% threshold and a minimum rating of 2400 for both players:

Barbero,G (2415) - Cvitan,O (2525) [E12]
Basel Hilton op Basel (6), 04.01.1999

1.d4 Nf6 2.c4 e6 3.Nf3 b6 4.Nc3 Bb7 5.a3 d5 6.cxd5 Nxd5 7.Qc2 Nxc3 8.Qxc3 Be7 9.Bf4 c6 10.e4 0-0 11.Rd1 Nd7 12.Bd3 c5 13.d5 exd5 14.exd5 Bf6 15.Qc2 Re8+ 16.Kf1 g6 17.Bb5 a6 18.Bc6 Qc8 19.Bxb7 Qxb7 20.h4 Rad8 21.Bg5 Bxg5 22.hxg5 Ne5 23.Qe4 Nxf3 24.Qxf3 Re5 25.Rh4 Rexd5 26.Rc1 Qe7 27.g3 Qxg5 28.Rcc4 Qd2 29.Kg2 Rf5 30.Rhf4 Qxb2 31.g4 Rxf4 32.Rxf4 Rd7 33.Qa8+ Kg7 34.Qe8 Rc7 35.Qd8 Qe5 36.Rf3 Qe7 37.Qa8 a5 38.Qb8 Rc6 39.Rd3 Qe4+ 0-1

PS: Nice cat, Pingudon!



(Grrrrrrrrrr)

Having a database program makes it difficult to not look at statistics. I also look at statistics but I try to ignore it as much as possible.

It's very difficult to use statistics in chess. On the one hand, one wants many games to have good statistics, on the other that will make you lump different lines together, which might give completely different type of play and also different statistics. As MNb noted, what were the statistics in the different branches of the Winawer? In the TWiC database, 6...Qa5 7.Bd2 Qa4 variation scores very good. Once you go down different branches, you will have less and less games. Thousands of games are soon hundreds and soon tens. Once you have tens you start looking at games and see that rating differences most often count and that many games are not decided by the opening.

Statistics is one of the most misunderstood and misused tools in and out of chess. It is very hard to take the statistics that emcompass a wide range of ratings and playing conditions and come up with meaningful conclusions for specific persons and situations.

For example, personally I don't care if the bayonet yields the best percentage against the KID. I am not playing the world elite, and I know that with my superior experience with a less widely used minor system, i can score more points.

I am not saying one should play offbeat off unsound systems. But I belive most persons would get better results playing systems that suit their style and with which they are familar regardless of the % performance in Megabase.

Interestingly, IM Larry Kaufman who apparently is a mathematician, made extensive use of statistics in coming up with his recommended repertoire in "The Chess Advantage in Black and White". The recommended lines are primarily based on analysis of performance ratings of the various openings. But look at the result - very dry unenterprising systems that are unlikely to appeal to the vast majority of chess players!

Hi I wonder if statistics are of any use to choose our favourite defence. Let's see the French 3.Nc3.
1.e4 e6 2. d4 d5 3.Nc3 Bb4 4.e5 c5 5. a3 Bxc3 6.bxc3 (385 games from 2005 - 2006)
1/0 36%
1/2 44%
0/1 20%
Now 1.e4 e6 2. d4 d5 3.Nc3 Nf6 4. Bg5 dxe4 5. Nxe4 Be7 (48 games from 2005 - 2006)
1/0 31%
1/2 55%
0/1 14%

Conclution: If you are a French player pray when you see 3.Nc3

Let's compare it with a Sicilian

1.e4 c5 2.Nf3 d6 3. d4 cxd4 4.Nxd4 Nf6 5. Nc3 a6. 6. Bg5 (359 games from 2005 - 2006)
1/0 32%
1/2 38%
0/1 30%

1.e4 c5 2.Nf3 d6 3. d4 cxd4 4.Nxd4 Nf6 5. Nc3 a6. 6. Bc4 (204 games from 2005 - 2006)
1/0 34%
1/2 36%
0/1 30%

Conclusion: Sicilian is great and the Frech sucks!!
Based on statistics can I conclude: everyone should play the sicilian?
I am a French player mysel and I do not know how to handle these results.
Any opinion would be most welcomed!!