Tuesday, March 20, 2012

Chess960: The more the better? part 2

Lately I've been experimenting with "Asymmetric Chess960" which is also called "Chess960 squared" (see comment below) or "Double Chess960". It is where the start position is a unique SP for each color.

One of almost one million possible Asymmetric Chess960 starts

Now you might think that this is crazy to do. But after testing quite a few positions against a computer, the rough start position evaluation range that the computer gives is kinda unexpected:

Approximate evaluation range according to Rybka 4 960 on average across many of these asymmetric positions:
White max eval roughly +0.45 at best
White min eval roughly  -0.25 at worst

Now that last result is the amazing part. In asymmetric Chess960 you get some start positions where black has an advantage at the start! It produces the bizarre situation where white is sitting there thinking about playing the first move, and they are already in a quandry. This could be because white's queen is not so well placed relative to blacks for example.

So what about the idea of Basque asymmetric chess to make asymmetric chess completely fair for both players?

Taking a leap from Mark Weeks blog on simultaneous chess, I thought I'd give a special case of  asymmetric Chess960 games a go on a computer. As Mark says, the advantage of the Basque system is that if players play the same start as both black and white simultaneously, then any imbalance from white starting first is removed (so long as you score the result properly).

A Simultaneous Asymmetric-Chess960 game using Winboard

As you can see, the same  asymmetric Chess960 position is played as both white and black simultaneously. I tried this out and it turns out to be a lot of fun and is by no means overwhelming so long as you make the time control long enough! It actually refreshes single board play, because the multitasking on two boards really teaches you how to organize the way that you think, which is useful for single board play.

Here is the only Chess software that I know of on the planet right now, that is capable of such a feat:
  1. Winboard 4.60 and Aquarium 3.11 are the only GUI's that can cope with Chess960 castling uniquely for each color on the same board, while simultaneously being able to run two boards side by side on the same screen.
  2. Arena 3.0 and Fritz 12 cannot handle it.
  3. The only high quality chess engine that I know of that is commercially available that can handle  asymmetric Chess960 start positions is Rybka 4 960. This is because the author must have written the engine to generate the correct bitboards for this unique start concept. This was extra work for him but much appreciated.
  4. Critter 960 and Fritz 960 do not understand  asymmetric starts
So, on a rainy Sunday afternoon, fire up two instances of Winboard 4.6 of Aquarium 3.11 with the UCI engine Rybka 4 Chess960 running on each. Paste in a unique asymmetric Chess960 FEN into Winboard and have fun! Just remember to keep an eye out on two chess clocks not just one and try not to get distracted when the computer plays a move on the other board.

So, enjoy a bit of Chess960 any way you like it.

PS) I've since learned that Houdini 2c is also capable of asymmetric Chess960

PSS) I'm in the middle of writing some software that rejects the top end of white's advantage in asymmetric Chess960. This theoretically leaves the start evaluation in the range of +0.25 to -0.25 for many hundreds of thousands of asymmetric Chess960 starts.

PSSS) Having experimented with asymmetric Chess960 some more, it really is enjoyable. This is because both sides have more flexibility and resources, because they are less restrained by the problem of symmetry stifling the opening sequence. This actually tends to make asymmetric chess a bit easier than straight Chess960 I think.

PSSSS) If the top end of white's favourable asymmetric Chess960 starts are not removed, the main theme that gives white a big advantage, is when white has two of the major pieces on one side of the board and black does not. This allows white to push the pawns out on the side where the major pieces are and gives white a lot of space. However what I am finding is that black has a lot of resources, surprising number of ways to counter and sometimes it is white that is on the defensive. That said, the one million possible asymmetric Chess960 starts probably does contain a few problem starts, and that's why they need to be pruned. But there are not as many as you would think!


  1. I've been playing Chess 960 and I also find Asymmetric Chess 960, which I like to call Chess 960 Squared, to be superior. It's so refreshing to have the board not be symmetric.

  2. Hey Aaron, yeah it unlocks the play and gives players even more creativity than Chess960 and it is surprising how many viable starts there are (tens of thousands).

    It is a bit easier as well because you don't have to calculate the outcome of symmetrical play which often can't be seen for many moves. In chess960, symmetrical play often has to be calculated not because it gives white an advantage, but because players that want to play for a win have to calculate how drawish a symmetrical line is.

    So those issues don't exist in asymmetric chess. If we could find a scientific way to prune out the few really unfair starts, I would love to play asymmetric chess even more!