Waving a Yellow Flag

A few years back, in a post titled Advantage in Chess960 Start Positions Revisited, I created a table from CCRL data showing which start positions appeared to be best for White and and which best for Black. The CCRL site, linked to the right of this page under 'Resources', explains,

CCRL means "Computer Chess Rating Lists". We are a club of people inspired by watching computers play chess.

While I'm not at all 'inspired by watching computers play chess', I do accept that the CCRL findings are the single most important collection of chess960 games in existence today. I ended the 'Revisited' post by concluding that the CCRL sample at that time was probably too small to make generalizations. The club now boasts of having played 115.900 games of chess960, or an average of more than 115 games per start position, or 3-4 times more games than at the time of my first post.

How do the current CCRL statistics compare with that first analysis? Here is a table similar to my earlier table. As before, it shows the number of games played (#); the percentage of games won by White, won by Black, and drawn (%W, %B, %D); and the overall score for White (%S).

SP	FEN	#	%W	%B	%D	%S
Highest Overall Score White
408	RBQNBNKR	112	57.1	29.5	13.4	63.8
868	QBBRKRNN	120	54.2	26.6	19.2	63.8
749	RKNBNQBR	116	52.6	25.8	21.6	63.4
730	RKNQBBNR	114	55.3	28.9	15.8	63.2
376	NBRKBRNQ	140	56.4	30.6	12.9	62.9
Highest Overall Score Black
465	BRNBNKRQ	106	34.0	50.0	16.0	42.0
786	BRQKNBRN	114	30.7	47.3	21.9	41.7
888	RBQKBRNN	118	27.1	44.1	28.8	41.5
499	BRQNKNRB	118	29.7	49.1	21.2	40.3
067	BNNRKQRB	116	28.5	49.1	22.4	39.7
Traditional Start
518	RNBQKBNR	135	44.4	34.8	20.7	54.8
Traditional Start (K&Q switched)
534	RNBKQBNR	84	34.5	35.7	29.8	49.4

The first observation is that the 'the overall score for White (%S)' lies in a narrower range than before: 64%-42% now vs. 76%-32% earlier. The second observation is that of the top five percentages for White, only one start position is on both lists (SP868 QBBRKRNN); of the bottom five percentages for White (i.e. top five for Black), the list is completely different. The third observation is that the traditional start position (SP518 RNBQKBNR) is now in line with experience.

On the earlier list, I examined the top position for White (SP024 NBQNBRKR) in A Followup, an Error, and an Insight, and surmised (with the help of another chess960 enthusiast) that 'the dangerous start positions share the common feature of several Pawns unprotected and these can be easily attacked by the Queen and Bishop'. That is clearly the case for SP868, where the 'QBB' on the a/b/c-files rake the 'RNN' f/g/h-files. For the record, SP024 is now no.12 on the list, with a 60.9% success rate for White rather than the 76.2% in the earlier table. It's worth remembering that the CCRL requires both 'Book learning' and 'Position learning' to be 'Off for all engines'. This means that the engines don't improve by playing more games.

Now here's a big, yellow WARNING FLAG: the no.1 and the no.3 entries in the current table (SP408 RBQNBNKR & SP749 RKNBNQBR) are twins -- they have the same sequence of pieces, but in reverse order -- only the castling considerations are different. These two positions, plus the repeat position SP868, need to be examined more closely. I would also like to identify the other top (and bottom) twins to determine how random their CCRL results are. I'll follow-up these actions in future posts.

Ignoring the Positional Handicap

Earlier this year, in No Time for Shelter, I reported on the SchemingMind 2010 Chess960 Dropout Tournament and now I can do the same for the 2011 Chess960 Dropout Tournament. The event finished a round earlier than previous years after the site restricted it to paying members only, which cut the number of entries in half. I wasn't a paid member at the beginning, so I wasn't able to participate.

Two of the strongest players met in the third round and were assigned SP655 RNKRQNBB as their start position. The initial moves were 1.g3 e5 2.f4 exf4 3.gxf4 f5 4.Nc3 d5 5.d4 c6, reaching the position shown in the first diagram.

The first of the talking points in the position is the castling option. Judging by the Pawn structure, it appears that both players have already decided to castle O-O-O. Since the Bishops are on adjacent diagonals aimed at that side of the board, the players have accepted a blocked Pawn center, thereby limiting the scope of the Bishops. Very surprisingly, White has also accepted a backward Pawn on the open e-file. He undoubtedly intends to play a Knight to e5, where Black will be tempted to exchange it, thereby closing the file. White's next move, 6.Nd2, was a step in that direction.

Some moves later the game reached the position shown in the second diagram. The same situation exists on the e-file, giving the impression that Black has been more successful than White in steering the development of the game. Chess positions are not always as simple as they appear and White ignored the positional handicap by playing the tactical 22.e4. Give that move a '!!'. Since both 22...fxe4 23.Bxe4 and 22...dxe4 23.d5 are bad for Black, he played instead 22...f4, trapping the Bishop on g3.

The tactics continued with 23.exd5 cxd5 (if 23...Nxd5, then 24.Bg4), followed by 24.Nxd5 Nxd5 25.Bxd5 Rfe7 26.Bxe6+ Rxe6 27.d5 Rxe1 28.Qc5+, the Queen check untrapping the Bishop for the recapture on e1. I don't often show chess960 middlegame sequences on this blog, especially when, as here, they are indistinguishable from positions that might arise from the traditional start position, but that tactical variation is too impressive to pass up. If you want to check the tactics yourself, here is the PGN, courtesy SchemingMind.

[Event "Chess960: 2011 Chess960 Dropout Tournament, Round 3"]
[Site "SchemingMind.com"]
[Date "2011.07.30"]
[Round "-"]
[White "saxon"]
[Black "Tyler"]
[Result "1-0"]
[Variant "fischerandom"]
[SetUp "1"]
[FEN "rnkrqnbb/pppppppp/8/8/8/8/PPPPPPPP/RNKRQNBB w KQkq - 0 1"]
[WhiteCountry "GER"]
[BlackCountry "SUI"]
[WhiteElo "2496"]
[BlackElo "2479"]

1.g3 e5 2.f4 exf4 3.gxf4 f5 4.Nc3 d5 5.d4 c6 6.Nd2 Ne6 7.e3 Nd7 8.Nf3 Bf7 9.Bf2 Qe7 10.Bh4 Nf6 11.Bg2 Bh5 12.Rd2 h6 13.a4 Bxf3 14.Bxf3 g5 15.fxg5 hxg5 16.Bg3 Re8 17.O-O-O O-O-O 18.Kb1 Qh7 19.Qf2 Rd7 20.Re1 Bg7 21.Rdd1 Rf7 22.e4 f4 23.exd5 cxd5 24.Nxd5 Nxd5 25.Bxd5 Rfe7 26.Bxe6+ Rxe6 27.d5 Rxe1 28.Qc5+ Kb8 29.Bxe1 Rc8 30.Qf2 Bf6 31.d6 Qf5 32.Rd3 Rc5 33.Qf3 Qd7 34.h4 gxh4 35.Qxf4 Qf5 36.Qh6 Qg5 37.Qf8+ Rc8 38.Qf7 Qc5 39.Qb3 Rd8 40.d7 Qg1 41.Qb4 Ka8 42.Ka2 Qg8+ 43.b3 Qg2 44.Qc5 Rb8 45.Bf2 b6 46.Qd6 Bd8 47.Rc3 a5 48.Bxb6 Rxb6 49.Qf8 Qg5 50.Rc8+ Ka7 51.Rxd8 Qd5 52.Ra8+ Qxa8 53.Qxa8+ Kxa8 54.d8=Q+ 1-0

The player of the White pieces went on to win the event.

Onward and Upward

As documented in Shall We Play Fischerandom Chess? (*), a little more than four years ago I started playing chess960. My first rating after that first game played on SchemingMind.com, Chess960? I'm Hooked!, was 1619. Four years later the site has me at no.2 for established players. I thought I'd record the event for posterity in the following image.

The SchemingMind ratings are approximately equivalent to ratings 200 points lower on other correspondence sites like ICCF and LSS, i.e. 2500 here is 2300 there. To avoid succumbing to the chess player's ancient enemy -- hubris -- I should point out that I've played all of the other players in the top five and have a combined record of +0-1=6. In other words, I haven't been able to defeat a single one of them. For that reason, I don't expect to see the no.1 position anytime soon.

Correspondence Chess Ratings and Chess960

On the right side of this page I maintain a list of of online chess960 play sites. A site to be added to the list of 'Correspondence (Turnbased) Chess960' sites is ICCF.com, the International Correspondence Chess Federation. Their page, ICCF Diamond Jubilee 1st Chess 960 World Cup, recently announced that

All preliminary sections are now completed! After 1 year and 522 games, the preliminary rounds are finished.

along with a link to the 'ICCF Games Archive'. I downloaded the file for 'Complete 960 chess until 2012-07-31' and found that it contained 560 games in PGN format.

On top of the required game info -- Event, Site, Date, etc. -- the PGN headers for each game included rating information for the players. Since there was no chess960 rating history on the site, I guessed that the ratings were carried over from the players' experience at traditional chess, making this a good opportunity to check the frequently asked question, 'Does skill in traditional chess carrry over to chess960?' I am almost certain that it does, but it never hurts to check one's assumptions.

I loaded the PGN header info into a database and did some simple calculations. Of the 560 games, there were 488 with ratings for both players. I couldn't think of a clever way to present the data graphically, so I just produced the table shown on the left.

The first column in the table shows the rating difference between White and Black, rounded to the nearest 50 points. The next three columns show a count of the results for that rating difference. For example, the first row shows there was one game where the White player outrated his opponent by about 1050 points, and White won. The last row shows one game where the Black player was higher rated by 1050 points and Black won. In total, White won 216 games, Black won 190, and 82 were drawn.

The table appears to show that the higher rated player of traditional chess does indeed have some advantage when playing chess960. I'm not sure how significant is a sample of 488 games and I'm not sure how well the findings compare to the familiar bell curve of difference in Elo ratings. I'll leave that investigation for another day.