14

In article in wiki about chess960, it is stated that:

It has been argued that two games should be played from each starting position, with players alternating as White and Black, since some initial positions may offer White a bigger advantage than in standard chess. For example, in some Chess960 starting positions White can attack an unprotected black pawn after the first move, whereas in standard chess it takes two turns for White to attack and there are no unprotected pawns

But what are these positions? Is there a comprehensive study of some of them?

3
  • There could potentially be many dozens of these positions. What kind of an answer are you expecting when you ask "what are these positions"?
    – Wes
    Apr 23 '14 at 21:53
  • 6
    @Wes I highly doubt that out of 960 possible positions, there would be many dozens of them. One or 2 dozens, may be. I expect to see something like this: here is a position (example). It leads to a forced N moves (because everything else give you a terrible play). After these N moves White is better by 0.7. By the way here are some other bad positions for white. And if you want - here is a study about some other positions. P.S. - I would be highly surprised if such answer will appear, because it requires a lot of search and analysis. But there is always hope inside of me. Apr 23 '14 at 22:14
  • 4
    There is a Boinc project (distributed computing that use idle time of personal computers) that aims to develop a kind of opening theory about chess 960. However, I could not find any result or any recent update about it.
    – WilQu
    May 2 '14 at 12:48
11

I asked myself the same question a year ago.

I found this interesting page: http://computerchess.org.uk/ccrl/404FRC/opening_report_by_white_score.html#table_start

With statistics for every 960 positions based on 144,300 games played by 123 engines.

Apparently the position that gives the biggest advantage to White is RKNQBBNR with a White score of 63.9%.

Note that on the previous page the standard position has a White score of 56.0%. I once read that the true White score in standard chess is 55.5% for GM games. Since these two numbers are pretty close it would seem that the statistics of the previous page are meaningful and pretty accurate.

Also, someone once told me that if we just inverted Knights and Rooks then White would have a big advantage: NRBQKBRN. But the previous page actually says that White only scores 47.4% in that position, so maybe he was wrong.

Anyway, I seriously doubt there exists a position where White has an advantage of +0.50 or more. And every single one of the 960 positions are very probably just draws with best play.

3
  • Cool, this is the best resource I have seen so far. I am still confused with what white score of 63.9% mean. Does this mean that out of all non drawing games white won 63.9%? Or out of all games white won 63.9%? Jul 3 '14 at 1:43
  • 1
    White score = (White's winning percentage) + (Draw percentage / 2)
    – Fate
    Jul 3 '14 at 1:52
  • 1
    @Petrosian Good response, until the last paragraph.How do you know or justify the claims made in the last paragraph? Aug 27 '17 at 23:15
4

I built this analysis over more than 4.5 millions of games played in Chess960 variant: https://github.com/welyab/chess960-win-by-position-setup

The most unbalanced setup is RNBKRBQN where white has 51.44% of win, and black 44.75% of win.

At general, white has a win rate of 49.17%, and black with 47.1%. There are more information on github repository.

1
  • Thank you very much for work +1, but I do not think that this is a proper analysis. Basically games on lichess are very noisy and with huge amount of mistakes/blunders as the games are mostly played with bullet time control (1 min) and by low-quality gamers (~1700 rated). Just as a suggestion for future improvement, you can do the following: for each position run a strong engine to play 100 games. If you assume each game to last 100 moves and 1 sec per move, you will need ~200 days on one machine. Because of straight-forward parallelization this can be reduced significantly May 15 '20 at 6:52
3

This question was put to the Norwegian supercomputer Sesse in 2019, which analyzed each opening position using Stockfish 9 to a depth of 40.

Its conclusion was that there are four initial positions that give white an advantage of 0.5 or better. The maximum imbalance (worst for black) is BBNNRKRQ which has an evaluation of +0.57 in white's favor.

position eval
BBNNRKRQ 0.57
NNRKQRBB 0.52
NNRBKQBR 0.50
QBBRKRNN 0.50

A side note: none of the 960 possible starting positions had an evaluation in black's favor. They're all even or favor white.

See more details in this article on chess.com.
And this report on blogspot.

0
1

While probably all 960 positions are a draw with best play, some positions give White the possibility of entering a position that forces Black to play exact, in order to not get in material disadvantage. In other words, White can make immediate threats that must be countered correctly. You can see this page for the motives you have to look out for:

https://lichess.org/study/UQTwIHXt

These are not the only starting positions with that motive but they are prototypical.

2

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.