1. Start with all the pawns on the second rank as usual, but take turns placing one piece (blindly selected from a hat) on the first rank squares a1 to h1. The two sides will have different setups. Now you have your initial setup and are ready to play.

2. Play your first 10 moves attempting to create glaring weaknesses on your own side (because you will switch sides after the 10th move), but keep the following in mind:

- You cannot move any piece twice.
- You cannot castle.
- You cannot capture.
- You must keep all your pieces on your half of the board. - You cannot place yourself or your opponent in check.

3. Switch sides, flip a coin to decide who moves first (yes, black can move first).

4. Play normally from there. Enjoy.

I call it Random Switch Chess.

(1) the rating advantage of pawn and move is less than 300 points and probably less than 200;

(2) the rating advantage of a knight is is less than 500 and probably less than 400;

(3) the rating advantage of a rook is probably less than 600.

Taking pawn and move as an example, the rating differences were 200-399, and the odds-givers won, whereas if the rating advantage of pawn and move were really, say 300 (the middle of the range), we'd expect about a 50% score in these games. Since the score was actually (almost) 100%, the rating advantage is probably not substantially above 200 or we'd have expected the odds-receivers to score something at least.

In my Edo rating system, I've been using (in the most recent version) a rating advantage of 174 for pawn and move; 349 for the knight and 500 for the rook. These are at least consistent with the upper limits suggested by the Marshall Chess Club experiment, though they must of course still be considered very tentative.

In the Kibbitzer's Corner for games of Robert James Fischer, <SBC> mentioned (on July 4, 2006) a match that Kasparov won 2.5-1.5 against Terry Chapman, Kasparov giving odds of 2 pawns (and extra time). Since Kasparov had a FIDE rating at the time of 2827 and Chapman seems to have been rated 2156 (though that might have been a few years later) the rating difference we can take to be about 671. A score of 2.5/4 normally amounts to about an 89 point rating difference, so Kasparov was playing at about 2156+89=2245 compared to Chapman's 2156. This suggests a rating disadvantage of 2827-2245=582 when giving odds of 2 pawns, which is much more than we'd expect from the Marshall Chess Club evidence that even knight odds make less than a 400 point difference. There are arguments in the same discussion that Kasparov could have done better, but the fact is he didn't. In any case 4 games is limited evidence, carrying less weight than the approximately 40 games of the Marshall Chess Club experiment.

Checkers Chess
http://en.wikipedia.org/wiki/Checke...

Checkless Chess
http://en.wikipedia.org/wiki/Checkl...

Knight Relay Chess
http://en.wikipedia.org/wiki/Knight...

http://www.chessvariants.org/diffob...

RULES:

Same as orthodox with these changes:

1) On b1 and g1 there will be a White King and on b8 and g8 there will be a Black King as regular Kings without the King's characteristics. They cannot castle, may be attacked, the attack does not have to be negated and they may be captured as normal pieces.

2) On e1 there will be a White Knight and on e8 there will be a Black Knight as regular Knights with the characteristics of a King, although they move as Knights. They may castle where orthodox rules apply, an attack on them must be negated and they may not be captured.

3) Pawns cannot be promoted into Knights and may promote into Kings.

4) As in orthodox chess, an attacked Knight is in checkmate when the attack cannot be negated and the checkmated players loses, and if a player cannot move when he is not in check will result in stalemate with a draw.

This variant can be immensely complex as one, the King's value cannot be easily disputed, two, Knights are easily checkmated and three, in the beginning it might be completely counterintuitive to move with Kings and let the Knight stay put.