Option Chess

[Bob Gillanders]

Congratulations Paul Bonham, for an article in Chessbase on your chess variant game. :)

http://en.chessbase.com/post/option-...by-paul-bonham

[Mathieu Cloutier]

Re: Option Chess

Sounds painfully convoluted.

First, it does change the game completely. Forget about strategy and long-term planning. You now have to spend all your time calculating double moves.

Second, it also fails at making the computers weaker. It makes the game much more tactical, playing to the strengths of the machine. I'm sure we'd be destroyed by series of double moves.

Expect some very negative feedback on chessbase, but alas, they don't publish it right away.

Also, maybe just a detail, but other variants of chess use double moves and in these, a check is always blocked with the first move.

EDIT: Took some time to look at the problem provided as an example and it is exactly like I thought it would be: double moves after double moves. So it doesn't feel like the double move is an 'option' as the name of the game suggests. Then it's also impossible to calculate, whereas a machine would just kill us in that kind of position (forced mate aside).

[Paul Bonham]

Re: Option Chess

First I want to publicly thank Frederic Friedel of ChessBase for going out on a bit of a limb and posting an idea from a relative unknown such as myself. It's nice to know there are still open minds in the chess world.

And thanks to Dr. Kenneth Regan who made this happen by contacting me through Chesstalk and letting me know Frederic Friedel was interested in my idea and was willing to publish it on his esteemed web site. Dr. Regan has his own very interesting variant idea Tandem Pawn Chess which also has a posting on chessbase.com:

http://en.chessbase.com/post/ken-reg...dem-pawn-chess

His variant is similar to mine in the respect of offering up an option (to uncouple or not to uncouple tandem pawns); the computer engine will have difficulty knowing the value of reserving that option for a later time in the game.

Now, to Mathieu's points:

Sounds painfully convoluted.

First, it does change the game completely. Forget about strategy and long-term planning. You now have to spend all your time calculating double moves.

Second, it also fails at making the computers weaker. It makes the game much more tactical, playing to the strengths of the machine. I'm sure we'd be destroyed by series of double moves.

With any introduction of new rules, it is hard to determine what long-term strategy and planning will look like. So that's a criticism I'm ok with, because I believe time will tell. If enough games get played, stategic concepts will emerge.

For example, I think a lot of strategic ideas in Option Chess will center around deliberately allowing various positional or even material deficiencies with the compensation of having extra options for the endgame. In standard chess, there is the idea of deliberately losing an exchange (trading Rook for Knight, or Rook for Bishop) for compensation of maybe a passed pawn or maybe better harmony between remaining pieces which can then coordinate in attack. Option Chess will still allow those kinds of strategies, while adding more having to do with remaining options.

A passed pawn becomes much more dangerous since with any remaining options, it can promote from the 6th rank in addition to the 7th rank. I would expect passed pawns to become a much bigger part of Option Chess strategy. There could be other things as well. The ability to make 2 Knight moves (with neither being a capture, but the second could be a check) gives the Knight more power in the endgame, whereas a Bishop being able to move twice is easier to defend against because of it's restriction to being only on a single color of squares. So this may change assessment of single Knight plus pawns versus single Bishop plus pawns in the endgame (but having the Bishop pair would still be an endgame advantage, I expect).

So just there, I came up with a few possible strategic ideas for Option Chess, all centering around letting the opponent use up options faster than you do and then counter in the endgame, where options become stronger. Other concepts may come out in time. Simply saying "forget about strategy" is naive.

Now consider this: humans can easily fathom these concepts and play deliberately to steer the game into them. But an Option Chess computer engine will have to spend much more time on each ply searching through all the possible double moves (and single moves). The result is that the effective search depth in a given time is reduced dramatically for the engine, so it may never even examine far enough ahead to see that if it plays a line where it sacrifices an exchange it's going to get a passed pawn on the 6th rank that can't be stopped from promoting. In other words, humans can (I believe) take advantage of a much greater strategical advantage against computers in Option Chess than is possible in standard chess.

Of course, the human must also calculate double moves and this will require even more advanced tactical abilities. But hasn't everyone been praising Magnus Carlsen for his ability to nullify his opponent's tactics, playing much like a computer? If everyone in standard chess plays like that (even if not to the strength of Carlsen), chess may become a dry wasteland of positional nuances, eventually bereft of tactics whatsoever. In fact, I think this may be part of the reason why ChessBase's Frederic Friedel is so interested in seeing some kind of anti-computer variant come to the fore (but I don't pretend to speak for him on that).

If you've been reading Wayne Komer's posts on Chesstalk about games from recent super-GM events, you'll note that the most exciting moments in the most exciting games center around... tactics. Nakamura up against Carlsen by almost a Queen (in engine evaluation, not in material) and loses with one move by missing the right tactic.

I'll ask the question: Can anyone name the most brilliant game played in the 21st Century between two top computer engines? I very much doubt it, because there probably isn't such a thing. What Nakamura did just isn't possible in computerland. Computer vs. computer games are almost exclusively dry, dull affairs often going well beyond 100 moves... which incidentally, even human vs. human games are getting more and more into that length. Is that the future you want for chess? More importantly, is that the future SPONSORS want for chess?

But to the charge that Option Chess plays into the strength of computer engines: that will happen, humans will lose to engines on tactics. But circa 1990, top players could still defeat the best computer engines by using strategy, and I believe Option Chess allows a return to those times. As stated in the article, the explosion in growth of the search tree is so fundamentally huge that the engines will spend all their time on tactics and never be able to see far enough to really use strategy, barring some very radical change in computing technology.

Regardless of who may be right or wrong on this question, it definitely should rekindle interest in human versus computer chess... if the chess is Option Chess or Dr. Regan's Tandem Pawn Chess. Will it be tactics or strategy that decides these matches?

Also, maybe just a detail, but other variants of chess use double moves and in these, a check is always blocked with the first move.

EDIT: Took some time to look at the problem provided as an example and it is exactly like I thought it would be: double moves after double moves. So it doesn't feel like the double move is an 'option' as the name of the game suggests. Then it's also impossible to calculate, whereas a machine would just kill us in that kind of position (forced mate aside).

So am I supposed to apologize for thinking out of the box? For not doing what all the other double move variants are doing? For going too far exercising the spirit of the word "variant"?

Restricting defense against checks to that provision would seem to allow more of the tactics you argue against. I thought it best to allow increased defense at the same time as allowing increased offense. But maybe time will prove this wrong and after extensive testing, this rule could be changed. Maybe it will make it too hard to achieve checkmate. We shall see.

The problem position was deliberately composed to show the nuances of double moves, including the special en passant provision that applies to a single piece being moved twice. So yes, every move there was a double move. How often would that happen in real games? Given the strategic importance of having double move option superiority in the endgame, I think that's an open question. And imagine the drama of seeing a player launch into a series of double moves: will it pan out? Can the opponent respond with a few single moves and still survive to carry that advantage into the endgame? It's a go-for-broke tactic that will be something akin to a hockey team pulling their goalie for an extra skater... but in the middle of the game, not at the end.

I think it can safely be said these options offer increased excitement and drama into chess. And at the same time, I do believe they will weaken the engines against strategy-minded humans.

I'm willing to put my money where my mouth is, regardless of armchair critics like Mathieu. You'll be seeing what I mean by that later this year.

[Mathieu Cloutier]

Re: Option Chess

So you agree that it makes the game more tactical... Yet you still seem to think that it will require some fundamental changes in computer technology for the engines to perform at this game? Sorry, but I don't follow you there.

The problem is that you are pursuing two very different goals. First, you don't seem to like chess as it is (we get it), so you want to change it to have less draws, more 'exciting' games. Then, you also want to give humans a better chance against computers, which requires a 'slower', less tactical game ('go' is a good example). Your variant certainly achieves the first objective, but not the second.

[Lee Hendon]

Re: Option Chess

Couldn't the computer engines just use the null move algorithms for the double moves
Correct me if I am wrong but don't engines already use this feature.

[Mathieu Cloutier]

Re: Option Chess

Exactly, current engines would be perfectly OK with 'option' chess. If you want to be fancy, you can add a little value for each remaining double move option, so the computer would not use it for the sake of using it.

Yes, it would be slightly harder to calculate for the machine. Still, handling the double moves would be a lot harder for human players, especially as the rules for double moves are not the same as for single moves. For a machine, it's just a few lines of code to run, for us it's like having two different sets of rules/patterns to handle.

It's completely preposterous to state that this game would require fundamental changes in computer technology in order for the machines to beat us. Current engines would destroy us even more that they are doing now.

[Kerry Liles]

Re: Option Chess

It's completely preposterous to state that this game would require fundamental changes in computer technology in order for the machines to beat us. Current engines would destroy us even more that they are doing now.

I am not so sure that many of the existing programs could adapt so easily. There is a lot of code that currently is structured to take advantage of some short circuit analysis that presumes the current set of rules. I do believe such modifications *could* be made - I am not so easily convinced with a 'wave of the hand' ... In any case, the double move option(s) change the characteristics of the game so much that I doubt most chess players would be interested, but we can wait to see what the reaction will be to the chessbase article.

As a side note, I have observed that the rules of baseball have not changed all that much in 100 years and every single change that is proposed or made is very very carefully scrutinized and, as a result, most changes have been rejected. Unfortunately, especially in the reign of Bettman, the NHL has not fared as well and now has a bunch of ridiculous rules that seem to have been invented by monkeys or lawyers (perhaps the same thing; apologies to the monkeys). There are always clarifications or tweaks based on specific situations (I am thinking again of baseball) but those are usually driven by necessity.

Contrast that with FIDE which appears to be more like the NHL - change for change sake and lawyer-like definition of fundamental concepts like touching a piece or determination of 'intent' etc. Sad that common sense no longer plays any part, but I guess that is progress.

[Louis Morin]

Re : Option Chess

Actually, this new game is very intriguing. My first try, should I play White, would be to play as sharp as possible for the first 8 moves. Forget about existing openings. What we need here is simply develop, open lines, play very sharp, even sacrificing one or two pieces to bring the Black King in the open.

From move 9 to move 20, I would as White use all my double moves in a row, trying at least to force my opponent to do the same in order to survive... if he can. Actually, I believe that most games should be over by move 20 or even before, which means no positional chess and no endgame. The main challenge for Black. I think, would be to survive until move 21, when double moves cease to be an option until move 45.

I really believe that this game could be very exciting for humans. Lots of tactics and attacking play. The only problem is that computers will be so good at it. Using double moves every time they can, they would wipe out any human opposition long before move 20.

But of course, these are only beginner thoughts. Only experience will show how the game fares in practice.

[Paul Bonham]

Re: Re : Option Chess

Actually, this new game is very intriguing. My first try, should I play White, would be to play as sharp as possible for the first 8 moves. Forget about existing openings. What we need here is simply develop, open lines, play very sharp, even sacrificing one or two pieces to bring the Black King in the open.

From move 9 to move 20, I would as White use all my double moves in a row, trying at least to force my opponent to do the same in order to survive... if he can. Actually, I believe that most games should be over by move 20 or even before, which means no positional chess and no endgame. The main challenge for Black. I think, would be to survive until move 21, when double moves cease to be an option until move 45.

I really believe that this game could be very exciting for humans. Lots of tactics and attacking play. The only problem is that computers will be so good at it. Using double moves every time they can, they would wipe out any human opposition long before move 20.

But of course, these are only beginner thoughts. Only experience will show how the game fares in practice.

If you are correct, Louis, then Steinitz if he were alive today would LOVE this game for the chance to play against the strategy you describe. I believe he was the player who at least when he had Black pieces would simply create a fortress-like position and dare his opponents to break through.

I'll reply again to your post later tonight when I have more time. Cheers.

[Lee Hendon]

Re: Re : Option Chess

Think to make the game fair Black should have first choice of the double move at move x
since white can go 2 moves ahead starting at move x if black has first option of the double move then he will only be a move ahead and then white will have a chance to go back to normal tempo of the game kind of like a tiebreak in tennis

[Paul Bonham]

Re: Option Chess

So you agree that it makes the game more tactical... Yet you still seem to think that it will require some fundamental changes in computer technology for the engines to perform at this game? Sorry, but I don't follow you there.

I have an idea! Let's do some math! Oh, come on, it's fun, and you're a whiz, right? If you read through this, maybe you'll catch me on some little mistake or two, and you can have an 'A-Ha!" moment.

So here we go:

Just to gain an idea how much double moves increase the search tree, let's assume both players could use double moves right from the game's start. In the standard chess start position, White has 20 move possibilities, and Black has 20 possible responses. That's 400 possibilities for the first 2 plies of the game.

Now let's add in double-moves on top of the single moves. We still have the 20 single-move possibilities for White. There's 4 Knight moves and 16 pawn moves. If the first move is Nc3, that removes 2 pawn moves (the c-pawn, now blocked by the Knight) from the second move of the double move. But it adds 4 new Knight moves: Na4, Nb5, Nd5, Ne4 (we can't count Nb1 because that is disallowed, you can't move a piece to a new square and then move it back to its original square as a double move).

On top of those 4 Knight moves there are the two moves of the other Knight plus the other 14 pawn moves. So just for that one Nc3 move, you now have 20 possible second moves. Ditto if the first move is Nf6. Wow, we just tripled White's possible opening moves and we've only begun. If White's first move is Na3, the Knight has 2 moves from there (Nb5, Nc4, again can't count Nb1), then there's the other Knight's 2 moves, and 14 other pawn moves. That's 18 new moves, and ditto if the first move was Nh3.

So if the first move of a double move for White was any Knight move, giving the possibility of a double move increased White's total opening move possibilities from just 20 in standard chess to now 96.

And now we consider just pawn double moves, with no moving of either Knight (because moves like 1.d4,Nf3 would be identical to 1.Nf3,d4). There are 16 possible first pawn moves. If the first pawn move is only a single square, there are now 15 possible second pawn moves. But one of those - moving the same pawn another single square - is identical to the single move of that pawn of 2 squares. In other words, 1.c3,c4 is identical to 1.c4 (except the former used up a double move option). So that gives 14 other pawn moves that aren't equivalent to any other opening so far. So there are 8 pawn single-square moves each giving 14 pawn second moves, meaning 8 x 14 = 112 new openings. Well, actually half that, because 1.a3,h3 is the same as 1.h3,a3. So 112 / 2 = 56 new pawn openings if the first pawn is a single square moves.

We add 56 to the 96 when either Knight is involved and we get 152 possibilities, all unique. Now, what if the first move of a double-move is a pawn move of 2 squares? To this we can only add another move of that pawn, i.e. 1.d4,d5 or any other pawn double move, ie. 1.d4,e4. If the first pawn moves is 1.a4, that leaves b4, c4, d4, e4, f4, g4, h4 remaining. So for a4, there are 7 other 2-square pawn moves plus the 1.a4,a5 possibility, making 8 possibilities. If you start 1.b4, you still have c4, d4, e4, f4, g4, h4 remaining and unique (you can't count a4, because 1.a4,b4 is the same as 1.b4,a4). Plus the 1.b4,b5 possibility, that makes 7 possibilities.

So progressing along the ranks, we get 8 plus 7 plus 6 plus 5 plus 4 plus 3 plus 2 plus 1 = 36 new openings, all unique.

( I have to write out the word 'plus' because the plus sign no longer displays in Chesstalk ).

We add 36 to the 152, and voila, we have gone from 20 opening moves for White to 188 opening moves for White. And we're still not done, because some first pawn moves open up moves for the Bishops, Rooks, Queen and King!

Each of the moves 1.a3 or 1.h3 opens up one Rook second moves. Each of the moves 1.a4 or 1.h4 adds two Rook second moves. That's 6 new unique openings. Running total: 188 plus 6 = 194.

Each of the moves 1.b3 or 1.b4 opens up two possible Bishop moves: Bb2 or Ba3. Ditto for 1.g3 or 1.g4. That's 8 more unique openings. Running total now 194 plus 8 = 202.

Each of the moves 1.c3 or 1.c4 opens up 3 new Queen moves (Qc2, Qb3, Qa4). That's 6 new unique openings. Running total: 202 plus 6 = 208.

The move 1.d3 opens up 5 new Bishop moves (Bd2, Be3, Bf4, Bg5, Bh6). We'll discount Bh6 since that would be a giveaway of the Bishop with no compensation. So 4 new Bishop moves, plus 1 new Queen move (Qd2) plus 1 new King move (Kd2). That's 6 new moves. The move 1.d4 opens up all those same moves plus Qd3, for 7 new moves. That's 13 new openings. Running total: 208 plus 13 = 221.

The move 1.e3 opens up 5 new Bishop moves (Be2, Bd3, Bc4, Bb5, Ba6). Again, chop off Ba6 for being a giveaway with no compensation, so 4 Bishop moves. Then there's 4 Queen moves (Qe2, Qf3, Qg4, Qh5) and 1 King move (Ke2). That's 9 second moves giving unique opening double moves. For 1.d4, we have again all those 9 again. That's 18 new openings. Running total: 221 plus 18 = 239.

Finally there's 1.f3 and 1.f4. Each of these offers only one additional move, Kf2. So that's 2 new openings.

Final total: 239 plus 2 = 241 unique openings for just White's 1st turn!

We went from 20 unique White openings to 241 unique White openings. Black will also have 241 unique replies.

Total possible unique openings of just White's and Black's first moves: 241 x 241 = 58,081.

In just 2 plies we've gone from 400 unique openings to 58,081. This is the explosively colossal growth in the search tree I mentioned in my article.

In standard chess the average number of moves available to the player to move is around 30.

In Option Chess, we don't know yet, but taking this opening example, let's be very conservative and imagine 200 possible single- or double-moves available to the player to move. Well, let's be REALLY conservative and chop it down to 120 -- not every move in a middlegame position will open up so many new possible moves for all the major pieces, and in the middlegame, some first moves will be disallowed by either being a capture or a check. So this new value of 120 moves available on average per ply is half of the 241 that we just calculated from the opening position. That should be conservative enough.

So from 30 possibilities per ply to 120: that's a 4-fold increase, being very conservative.

You can see that beyond about 6 plies, the size of the search tree is going to cause problems for even today's best chess engines, regardless of null move or alpha beta pruning.

30 to the power 6: 729 million nodes in the search tree

120 to the power 6: 2.986 trillion nodes in the search tree

Ratio: 4,096 to 1.

To find that mate in 3, the engine has to consider about 4,100 times as many positions as it would in standard chess. So even if your vaunted engine does all its null moves and alpha-beta and bitboard magic and finds the standard chess Black-mates-in-3 in 10 seconds... in Option Chess that same search with the same level of pruning and other magic will take on the order of 40,000 seconds. That's about half a day.

If we're talking about a Black-mates-in-4, that's 8 plies, and the ratio becomes 65,000 to 1. For that mate in 4 search, you'll be waiting weeks.

The viable search horizon of the engine has been reduced in Option Chess to somewhere between about 4 to 6 plies depending on time control. That's going to introduce a hell of a horizon effect, not to mention eliminate any chance the engine will get a sniff of long term strategy.

[Paul Bonham]

Re: Re : Option Chess

Think to make the game fair Black should have first choice of the double move at move x
since white can go 2 moves ahead starting at move x if black has first option of the double move then he will only be a move ahead and then white will have a chance to go back to normal tempo of the game kind of like a tiebreak in tennis

That is not a bad idea. So White's 9th move would have to remain a single move, and anything after that is open to the use of the options. Very good.

On your other post, that was a good question about the null move, although the non-programmers will not know what that means. So for them: In basic terms, the program forfeits a move for the side to move (and what Lee is saying, it could forfeit the double move option) and then does a search with now the other side to move. This is used in combination with alpha-beta to reduce the size of the search tree (moves and counter-moves) by pruning off whole branches.

But the null move algorithm still does require a search to be done, it's just a search to a slightly shallower depth. And it's not always going to find a cut-off and thus be able to prune the branch. It is also known for occassionally allowing tactical blunders, especially in simpler positions. So its use is somewhat risky.

But to really get an idea what a computer chess engine author would be up against in adjusting for this variant, see my reply to Mathieu.

[Paul Bonham]

Re: Re : Option Chess

Louis, would you be up for testing out your theorem? We could play 2 games by email, one where you are White and one where I am White. As much time as we need between moves. Of course, there are no Option Chess engines, so we needn't worry about cheating. Also, we could try Lee Hendon's recommendation in which White's 9th move must also be a single move, so Black is the first player to get to play an option.

My email address is given on the ChessBase article. Cheers.

[Mathieu Cloutier]

Re: Option Chess

[...]
To find that mate in 3, the engine has to consider about 4,100 times as many positions as it would in standard chess. So even if your vaunted engine does all its null moves and alpha-beta and bitboard magic and finds the standard chess Black-mates-in-3 in 10 seconds... in Option Chess that same search with the same level of pruning and other magic will take on the order of 40,000 seconds. That's about half a day.

[...]
The viable search horizon of the engine has been reduced in Option Chess to somewhere between about 4 to 6 plies depending on time control. That's going to introduce a hell of a horizon effect, not to mention eliminate any chance the engine will get a sniff of long term strategy.

Paul is at it again, it seems! On my relatively old and not so powerful computer, an engine finds a mate in three in less than a thenth of a second. So your numbers are probably off by two orders of magnitude, if not more. What's the point of talking about this if you come up with completely ridiculous estimates for calculation times? Current processors calculate much faster than you seem to assume.

Anyway, whatever the numbers are, it's not that we don't agree with you there. The tree of variations is wider and calculation depth will be down a little bit for the computer (it will be much more than 4 to 6 plies, though). No need to write two novels to explain that, we get it.

But that's not a fundamental change, not even close. It just makes the game more a matter of crunching variations than planning. And computers are better at crunching variations. Humans will be totally incapable of accurate calculation at this game. And forget about planning. You'll be mated or clearly down on material long before that.

[Egidijus Zeromskis]

Re: Option Chess

Give us more popcorn.