Option Chess

[Paul Bonham]

Re: Option Chess

Maybe I have this wrong, but here goes.

1) As far as I can tell, to a computer, a "double move" is simply 2 plies instead of one ply in the venacular. So the addition of the double move should only change the number of plys.
2) By the same token, humans have to also deal with the same increase in the number of plies in their calculations. Instead of thinking "I do this, he does this, I do this", we also have to think "I do this, then this, then he does this" or "I do this, then he does this and this."
3) Computers have not gotten better a playing chess due to raw computation. They have gotten better at (1) When to stop calculating and simply evaluating a position, and (2) How to evaluate a position. For instance with Deep Blue, I recall the head programmer describing how his program got so much better after he added a routine for evaluating the strength of rooks on half-open files.

So as far as I can tell, the main thing this changes is that the evaluation functions will have to radically revised, which well take time. The same is true for humans as well (the pin example above is just one example).

Garland, your point 1 is somewhat correct, but it has to be noted that changing the number of plies has an exponential effect, not a linear effect. I exemplified this in my post to Mathieu where I imagined one could make a double move right from the start position. That made White's possible unique double moves (and thus White's first turn possibilities) jump to 241, from just 20 in the single-move standard chess. Then Black has 241 possible unique replies, making 58,081 unique positions possible after just 1 turn per player.

So since we are effectively increasing plies on EVERY PLY of the game, we are VERY DRASTICALLY cutting down the possible search depth of today's engines should they be revised to play Option Chess. I don't believe even the fastest engine would be able to go beyond a full-breadth search to 8 plies deep in a reasonable time control, whereas they can do 25 plies and beyond in standard chess. This drop in search depth will correspond to an ELO drop of many hundreds of points, IMO. Exactly how much, only time would tell, and that only if engines are changed to play Option Chess.

Right now, engine authors have no incentive to make such a change... but that incentive could appear within several months.

I think your point (3) needs some further substantiation. For example, if Deep Blue was changed to correctly evaluate the strength of rooks on half-open files, that means Deep Blue started playing more moves leading to those kinds of positions. But at what cost? Did those moves sacrifice King safety? Did they encourage creation of doubled pawns? Perhaps it could be said Deep Blue got "stronger but riskier" in its play? Not every goal in chess comes without a price. Zugzwang shows us that the price of any move can sometimes be fatal.

Many years ago now, I tried writing a chess program with the goal of forgetting about speed and depth of search, and instead focusing almost solely on the eval() function. My reasoning went that if the engine was spending 90% of its time in eval(), that was good! I was thinking I could prove what you wrote above, that evaluation is more important. But no matter how much more detail I put into eval(), the program got crushed again and again against faster, deeper searching engines. So I've learned this lesson through practice.

You could argue that maybe I put the wrong stuff into eval(). But everything I put in there was a reflection of what is taught in chess circles to be good. I probably even put something about creating half open files for the rooks. I learned the hard way that you have to consider WHEN is it good! And that requires (for computer engines) better and better search depth.

So what's more important than Deep Blue creating half-open files for its rooks is the question of WHEN does it do this? For example, if it thought those half-open files were the most important thing in chess, it would start every game by advancing the a- and h-pawns until they were either captured or could make a capture. Voila, half open file for the rook! Meanwhile, Joe the Plumber mates Deep Blue on f2 or f7!

The 'right' answer to the question of when to create some positional advantage can only come by deeper and deeper search of the tree. The reason for that is summed up in one word: tactics.

[Paul Bonham]

Re: Option Chess

from another ChessBase article ...



He goes on.



FWIW - Computer-resistant chess variants

There's something I don't get about this. Fritz Juhnke writes about Option Chess: "Yes, the decision of when to use one's options would be strategic and thus favor humans, but execution of the options would make even more drastic positional changes with each move than chess already has, thus favoring computers."

Yet this is from someone who plays a game that (if I'm reading it right, and I had never heard of this game before) allows up to 4 piece moves per player per turn. Wouldn't that allow "drastic positional changes" even more so than Option Chess? I'll have to ask him that question someday.

It would appear that Arimaa is a very slow, methodical game virtually free of tactics. I'd have to actually play the game to appreciate how this is done. But my guess is that being so free of tactics is the key to its computer resistance, and would also make the game rather dull and dreary to play if one likes complications. It might even be something akin to playing Ken Regan's Tandem Pawn chess... with no minor pieces. Just 8 tandem pawns that can eventually become 16 individual pawns and a King... per side. That would have some tactics... but not much.

I will sometime in the future when I'm not so busy explore this game. But I'm not worried about Fritz Juhnke's opinions, because one thing that is apparent about Option Chess is that being tactics-rich, it will be FUN to play.

Also I should add that Frederic Friedel of ChessBase is searching for a computer resistant game that is as close to standard chess as possible. That is a tough combination of requirements. He wants today's chess players to immediately recognize the game, i.e. no fairy pieces that can (for example) tunnel from their current board square to any empty board square. The pieces should look and act as much like standard chess pieces as possible. Yet somehow, computers should suddenly have trouble with it.

I think both Ken Regan and I came about as close as possible to this double requirement. Arimaa is another animal altogether, and may not have much to teach chess players because of its almost tactics-free play. Chess is predominantly tactics, and strategy to a much lesser degree. Arimaa almost sounds like anti-chess.

[Mathieu Cloutier]

Re: Option Chess

It would appear that Arimaa is a very slow, methodical game virtually free of tactics. I'd have to actually play the game to appreciate how this is done. But my guess is that being so free of tactics is the key to its computer resistance [...]

I will sometime in the future when I'm not so busy explore this game. But I'm not worried about Fritz Juhnke's opinions, because one thing that is apparent about Option Chess is that being tactics-rich, it will be FUN to play.

[...]

You don't seem to be aware of the contradiction there...

Want a version of chess that looks like the current game, is harder for computers and still OK for humans? I have a solution for you. Ready?

Just give each player an 'option' of playing a pawn backwards a couple of times during the game. We'd have to figure out the ideal number of options, because having 10 or more backwards moves would slow down the game too much and just one backward move would revert back to normal chess too easily. Maybe something around 4 would be a good number.

Humans would still be able to calculate and use their strategic thinking. New strategic concepts and openings would be introduced. Endgames could be a completetly different business. And humans would pick up on these concepts much faster than current engines. The game would inevitably be slower, but that would be to our advantage, at least for a time.

I still think engines would beat us with some tweaking, but that would make their job harder.

There, you can't say I'm always negative and trolling!

Cheers,

Mathieu

[Kevin Pacey]

Re: Option Chess

from another ChessBase article ...



He goes on.



FWIW - Computer-resistant chess variants

There's something I don't get about this. Fritz Juhnke writes about Option Chess: "Yes, the decision of when to use one's options would be strategic and thus favor humans, but execution of the options would make even more drastic positional changes with each move than chess already has, thus favoring computers."

Yet this is from someone who plays a game that (if I'm reading it right, and I had never heard of this game before) allows up to 4 piece moves per player per turn. Wouldn't that allow "drastic positional changes" even more so than Option Chess? I'll have to ask him that question someday.

Indeed, there are up to 4 piece moves per player per turn. I played over some of the moves of the sample game that was provided in the wikipedia entry for Arimaa linked to from the article (i.e. in the link provided by Nigel):

http://arimaa.com/arimaa/games/jsSho...i?gid=5641&s=w

[edit: here's the wikipedia entry for Arimaa, for those not knowing the rules to this game:

http://en.wikipedia.org/wiki/Arimaa

]

Within the same article (i.e. in the link provided by Nigel):

Omar Syed's first rule set when attempting to design Arimaa was essentially four-move chess, which massively increased the branching factor, but also brought about drastic board changes on every move. He quickly abandoned this line of investigation, concluding that four-move chess would increase computer dominance of humans, not decrease it.

I believe Fritz Juhnke's view might be that 4 moves in a row by chess pieces could often be by very long range pieces, which would not only help increase the branching factor, but 4 moves in a row by such chess pieces could also increase "drastic positional changes" much worse than anything comparable that Arimaa does (positionally), since in Arimaa all of the pieces only move themselves one square at a time orthogonally (i.e. at right angles), as one move of a maximum of four, per one player's turn. Similarly, in Option Chess, 2 moves in a row might be made, but again these could often be made by very long range chess pieces. [edit: Fwiw, I'm not entirely convinced by this argument, if this is indeed more or less why Fritz Juhnke thinks Option Chess isn't so computer-resistant, as compared to Arimaa, at least; for one thing, in Arimaa enemy pieces might be pushed, pulled, 'frozen' or even captured, at times, as part of a single move (out of 4 max. on a player's turn), which adds to the positional changes made by a player's turn]

It would appear that Arimaa is a very slow, methodical game virtually free of tactics. I'd have to actually play the game to appreciate how this is done. But my guess is that being so free of tactics is the key to its computer resistance, and would also make the game rather dull and dreary to play if one likes complications. It might even be something akin to playing Ken Regan's Tandem Pawn chess... with no minor pieces. Just 8 tandem pawns that can eventually become 16 individual pawns and a King... per side. That would have some tactics... but not much.

I will sometime in the future when I'm not so busy explore this game. But I'm not worried about Fritz Juhnke's opinions, because one thing that is apparent about Option Chess is that being tactics-rich, it will be FUN to play.

Also I should add that Frederic Friedel of ChessBase is searching for a computer resistant game that is as close to standard chess as possible. That is a tough combination of requirements. He wants today's chess players to immediately recognize the game, i.e. no fairy pieces that can (for example) tunnel from their current board square to any empty board square. The pieces should look and act as much like standard chess pieces as possible. Yet somehow, computers should suddenly have trouble with it.

I think both Ken Regan and I came about as close as possible to this double requirement. Arimaa is another animal altogether, and may not have much to teach chess players because of its almost tactics-free play. Chess is predominantly tactics, and strategy to a much lesser degree. Arimaa almost sounds like anti-chess.

I came up with something similar to Frederic Friedel's criteria independently (indeed almost any chessplayer might have thought of them, or even wished for them to be actualized). However I was willing to experiment with fairy chess pieces or board dimensions at least to some degree (from what little I've read, I can't see why Frederic Friedel would rule such out). Also, I did want just a 2 player game, of course. An extra criteria I had was that any variant I came up with could look nice on someone's coffee table as a decorative set. I felt this could be important in marketing and popularizing such a variant, and when you have a decorative set, a guest may even be inspired to try playing.

This led to another criteria for me, that there be a standard start position for any desirable variant. An empty board for a variant wouldn't look so nice on someone's coffee table. Hence, no Chess960 style random setups and no Arimaa either, since their setups at the start of a game can vary because of their rules. This would also have the side effect that opening theory for such a desirable variant would accumulate over time, like for standard chess [edit: opening study of the variant could be rewarded, and opening books (etc.) for the variant could be authored and sold - not necessarily bad things]. Therefore I wanted a complex variant that would be much harder to virtually exhaust opening-theory-wise than standard chess looks like it may be. My efforts led a 10x10 variant of my own creation that I showed in another chess variants thread, and on my chesstalk blog. Unfortunately, Fritz Juhnke's criticism of S-Chess (i.e. introducing a few more pieces that are Queen-like favours a machine more than a human) is something I can agree with, so my 10x10 variant attempt now clearly looks wasted to me, at least as far as its computer-resistance goes.

My latest thought is an old one of mine, namely that a fairly chesslike variant that could well be quite computer resistant (and could well be accepted, as far as growing its popularity goes) could be Double Chess. It almost fully satisfies me, except for the fact that it is ideally a 4 player game (though 2 [or even 3] people could play instead), the clocks can often be a huge factor, some rules have variants depending on local tastes, and it's also not suitable for coffee tables. Double Chess might address Fritz Juhnke's concerns and advice regarding computer-resistance to a surprising degree, IMHO, and it retains the element of having lots of not-too-overwhelming-to-calculate tactics, which is enjoyable as far as a lot of chess players are concerned.

[edit: from what little thought I've given it, Crazyhouse (a Shogi-like chess variant, i.e. it's played with 'drops' like in Double Chess) might be less computer-resistant, and perhaps less fun, than Double Chess] [edit: it's possible opening books could be written for Double Chess, if its rules are standardized globally. However, opening theory in Double Chess might need to include an indicator at any given move, as to what one should do if one or both players on a team are clearly ahead or behind on their respective clocks - clock times can be vital if deciding whether to wait for a piece to be passed between partners, rather than make a move(!)]

With regard to Tandem Pawn Chess, I think it would have a significantly different feel to it than standard chess, without having the proven following of Double Chess, which seems somehow a bit more like standard chess to me. Ditto with regard to Option Chess. Both might have a problem satisfying my 'coffee table' criteria too, as far as aesthetics go, but if they could be made to look alright on a coffee table, they would have that advantage over Double Chess.

Fritz Juhnke's liking of Arimaa, in regard to computer resistance, must be taken with a grain of salt, as he has an interest in promoting that game, as a former world champion and author on the game. It's possible that as soon as Arimaa strategies are well known and understood, computers will dominate that game, like in standard chess (which took more time for them to conquer than has been available for them to master Arimaa similarly). If more time goes by and this has not happened, more people might become interested in Arimaa, however. As far as Arimaa being virtually tactics-free goes, I don't know (and doubt) if this can be said, even without my knowing anything about how to play it well. If there are indeed significant tactics, and if they are pleasing to those who play, such tactics could be rather alien for players of chesslike games, which is almost (though not quite) akin to what you've written about Arimaa, Paul.

[Felix Nemenyi]

Re: Option Chess

Option chess. Who play! Not I! Who play! Nobody! Alone, are you now, Mister Bonham. I play my good ol' chess.

[Kevin Pacey]

Re: Option Chess

From the rules of Option Chess:

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

7.For OTB play, neither player may be more than a few steps away from the board on their opponent's time, in case the opponent wishes to execute a double move. Players may only leave the game area on their own time (this may require some lengthening of time controls for Option Chess).

It may be a minor point, but I like being able to leave the board right after I move, if I wish, in a regular chess tournament game (granted, one is also not supposed to leave the board on one's own turn). With the rule #7 of Option Chess, people may more often (than in standard chess) find themselves waiting quite a while (i.e. for the opponent to move), rather than slightly hurrying to make a move of their own, if they need to take a washroom break.

(Btw, I edited my previous post somewhat since I posted it, for those who haven't noticed)

[Mathieu Cloutier]

Re: Option Chess

From the rules of Option Chess:

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

7.For OTB play, neither player may be more than a few steps away from the board on their opponent's time, in case the opponent wishes to execute a double move. Players may only leave the game area on their own time (this may require some lengthening of time controls for Option Chess).

It may be a minor point, but I like being able to leave the board right after I move, if I wish, in a regular chess tournament game [...]

Well, that 'rule' is simply unecessary, IMO. What's the difference with normal chess, where a player could play a move, press on his clock and change his move afterwards if the opponent is not there? The rules of chess already state that your move is completed when you press the clock.

And as you wrote, this 'rule' induce a much more important problem. You just don't want players going to the washroom on their time.

[Kevin Pacey]

Re: Option Chess

http://www.chessvariants.org/

I noticed on the 'The Chess Variant Pages' website that they prefer that any variants submitted to them include tips and sample game(s) played with the variant being submitted. This may have been a very good policy for Chessbase's website to follow as well, in regard to the various (hypothetically/supposedly) computer-resistant variants that have recently been submitted to them in a number of articles (including those on Option Chess and Tandem Pawn Chess) refered to in the link given by Nigel H. above.

[Louis Morin]

Re : Re: Option Chess

I think there is one rule that should be explained further in Option Chess: the 3-time repetition.

In chess this is a draw because the position can be repeated infinitely, but in Option Chess, what if the repetition involves double moves? Since there is only a finite supply of double moves, the position cannot be repeated infinitely, so it should not be a draw anymore?

[Kevin Pacey]

Re: Re : Re: Option Chess

I think there is one rule that should be explained further in Option Chess: the 3-time repetition.

In chess this is a draw because the position can be repeated infinitely, but in Option Chess, what if the repetition involves double moves? Since there is only a finite supply of double moves, the position cannot be repeated infinitely, so it should not be a draw anymore?

To try to keep this variant's rules no more complex than they already are, personally I'd say (or guess) that if the players repeat the position three times in Option Chess, with double-moves involved, the ruling ought to be that they are tacitly (or perhaps unwittingly) agreeing to a draw.

[edit: note that in standard chess, having a draw declared after three-fold repetition is at least partly arbitrary - why not after, say, four repetitions? what if it could be shown that either player could easily afford to deviate, rather than repeat forever?]

[Kevin Pacey]

Re: Re : Re: Option Chess

I think there is one rule that should be explained further in Option Chess: the 3-time repetition.

In chess this is a draw because the position can be repeated infinitely, but in Option Chess, what if the repetition involves double moves? Since there is only a finite supply of double moves, the position cannot be repeated infinitely, so it should not be a draw anymore?

Oops, I see your point more clearly now, Louis, which I didn't in my previous post. Namely, one player could be forcing the play in an effort to repeat until claiming a draw by three-fold repetition, against his opponent's wishes, using double-moves possibly. Good luck to Paul in solving this issue smoothly.

[Simon Valiquette]

Re : Option Chess

Louis is right and it would not be a draw because the possibilities won't be the same for the 3 repetitions (since options are limited). Supporting it is the fact that in regular chess, even if you get 3 times the exact same position, if in the 1st repetition a prise en passant or a castling was possible, it is not really a 3rd repetition since the possibilities were not the same.

A player with 5 options left against another with 4 options left could certainly wish to put his opponent in zugzwang to be the only one with an option left, and for some endings it could be the only way to force a win, so it shouldn't count as a draw.

Now for the computer-resistant claim, I guess it is time that someone with enough computer-chess knowledge steps in. First, let's put in some definitions. A turn is when both player have played, a ply is a move made by a single player (often called an half-move), and an option includes 2 ply. So if 2 turns are played and if both players use options at each turn, 8 ply would have been played after 2 turns, and in any case a computer would need to calculate 9-10 ply (instead of 5) to calculate white's 3rd move. So far, Option Chess seems to be doing fine, if it was not for a major optimization trick that is usually not as powerful with regular chess.

I also note that it is not that useful to count the number of turns, as it is probably more relevant to compare the number of ply calculated by the computer to the number of really played half-moves to really appreciate the value of Option chess, as calculating in terms of turns is effectively less relevant for both humans and computers (it is useless to say that you played 4 turns without knowing that one player used 3 options).

If a very strong computer is able to calculate 40 ply (so 20 moves in regular chess), it should still be able to calculate close to 40 ply with Option chess regardless of the number of turns. Now the claim seems to be that this same very powerful chess engine won't be able to calculate much more than 10 turns in Option chess even with proper optimization, since it won't know for sure when each option will be used. I believe that such a computer would still be able to calculate at least 15 turns ahead (so about 38 ply, a lot more than most humans) much of the time, but before let's talk about the tablebases.

Tablebases: begone!

First, the article said «Tablebases: begone!». Perhaps that quick claim was made without too much tough, but it is not accurate. First, tablebases could still be useful if the computer can calculate until the point where one of the player have run out of option. Secondly, the tablebases could be recomputed to include the possibility of each players having between 0 and 7 options each.

At first sight, it seems that the table would need to be 64 times bigger (128 results per table instead of only 2), and be at least 64 times slower to compute, but it is not the case. For example, if you have already calculated the table up to 1-1 options left and start to calculate with 2-1 options, if you see a position where 1-1 would already be winning, you don't have to calculate anything more even without using your option, and you already know that 3-1 up to 7-1 (and higher) is winning. There is plenty of more complex optimization tricks available, so rebuilding the tablebases will be much faster to compute than it may seems. Also, it would be possible to compress it a lot more than regular tablebase, so my guess is that it would be no more than 8-10 times bigger than regular tablebases, perhaps much smaller than that. Please note that it would be useless to specify in the tablebase if the next move require using an option: even a cellphone should be able to check all possible positions and find in the tablebases the best move in less than 1 millisecond.

Those tablebases could then be completed by concentrating on the more common endings and while it would not be as complete as the normal tablebase, it would still be much bigger than what a human can understand, and some optimization could be used to help the computer to start using those tablebases even if there is more material left. Anyway, the computer just need to calculate up to a point when he can reach a position where he already know the theoretical result.

In theory, the human would still have an advantage in endings if there is still way too much material for the computer to get help from the tablebases, but humans would first need to learn the new endings rules from those same tablebases (since computers will master them way before humans), but even tablebases will only show the shortest win, not the easiest one. Also, the effective search tree may not be as large as some may think (see next section).

Exponentially reducing the search tree

There was the claim that the search tree would expands exponentially, while forgetting that the potential positions would still be quite limited. It means that after enough turns, there will be an exponentially growing number of ways to reach the exact same position with the same player to move. The optimization trick here would be to encode the position so as to be able to put it directly in a hash table so that when the same position is reached by another search thread, it will realize it instantly and won't bother calculating it again (this technique is already used by chess engines, but the boost in performance with Option chess would be huge compared to regular chess). If the same position is reached, but not with the same number of options, a penalty (or bonus) could probably be calculated depending of how many options each player have left. It won't be as precise, but it could be good enough most of the time and much faster than recalculating everything. There could even be a function to evaluate if an approximation is likely to be good enough, or if it is safer to branch. In any case, the computer will still see further moves than most humans.

A special attention would have to be made for when a piece could be taken en passant, but we could simply check how this is already managed for a pawn, and extend it to every pieces. Note that a bishop have 6 ways to move from a1 to h8 in 2 ply, and a Queen potentially 8 ways to reach it, all with different prise en passant squares. Many variations could easily be optimized away (like when nobody can take it en passant), but it would certainly waste quite a bit of CPU time. Open positions with Queens, Rook and a minor piece on both side could be very difficult for computers if many options are left and the pieces have the potential to be taken en passant on many squares. There might be some way to optimize many cases, but humans will probably have an easier time at ignoring useless/equivalent variations.

Of course, it will be necessary to also tune the engine with adjusted values for each pieces, advanced pawns, open column, et cætera and introduce new and adapted optimizations, but I wanted to keep the discussion on the main issue: the search tree.

The issue with Option chess and many other variations is that the number of possible positions is limited, and I heavily take advantage of it to reduce the search tree as much as possible. Many other chess variants are vulnerable to this kind of "attack", which makes it very difficult to develop new rules that would make the game more difficult for computers than for humans without adding squares or deeply change the game.

In favor of Option chess

I have to recognize that finding a new idea to make the game harder for the computers than for humans is very difficult. Now my optimization trick have a very big weakness: it only really kicks-in after enough moves have been calculated to drastically shrink the search tree. It means that Option chess may do the job for blitz games. Or maybe not: I didn't have yet put enough tough on it to find out how strong the chess engine would be in blitz games.

The real main advantage of humans is the flexibility to quickly learn new games or rules, something that computers are seldom good at. So if we changed the rules every 5 years, chess engines would probably not have enough time to catch up, but too few humans would be willing to learn new rules every few years

Simon

[Paul Bonham]

Re: Re : Re: Option Chess

Oops, I see your point more clearly now, Louis, which I didn't in my previous post. Namely, one player could be forcing the play in an effort to repeat until claiming a draw by three-fold repetition, against his opponent's wishes, using double-moves possibly. Good luck to Paul in solving this issue smoothly.

Kevin, I'm not clear on what you think the issues are here. However, since options remaining are an intrinsic characteristic of an Option Chess position, a position will not be regarded as a repeat if the remaining options are different for either player. Therefore for a 3-fold repetition draw to be claimed, each repeat of the position must have the exact same number of options remaining for each player.

Also, Louis asked me a question about castling: is the castling move regarded in Option chess as a King move or a Rook move? Castling as part of a double move might allow either the King or the Rook to be moved in the SECOND move of the double move, therefore could either of them capture material?

The answer I gave Louis for this excellent question is that castling in Option Chess is both a King and a Rook move, therefore if EITHER piece is moved again as the second move of a double ove, that piece (King or castled Rook) cannot capture material on that second move. But the Rook could give check.

[Paul Bonham]

Re: Re : Option Chess

There was the claim that the search tree would expands exponentially, while forgetting that the potential positions would still be quite limited. It means that after enough turns, there will be an exponentially growing number of ways to reach the exact same position with the same player to move. The optimization trick here would be to encode the position so as to be able to put it directly in a hash table so that when the same position is reached by another search thread, it will realize it instantly and won't bother calculating it again (this technique is already used by chess engines, but the boost in performance with Option chess would be huge compared to regular chess). If the same position is reached, but not with the same number of options, a penalty (or bonus) could probably be calculated depending of how many options each player have left. It won't be as precise, but it could be good enough most of the time and much faster than recalculating everything. There could even be a function to evaluate if an approximation is likely to be good enough, or if it is safer to branch. In any case, the computer will still see further moves than most humans.

A special attention would have to be made for when a piece could be taken en passant, but we could simply check how this is already managed for a pawn, and extend it to every pieces. Note that a bishop have 6 ways to move from a1 to h8 in 2 ply, and a Queen potentially 8 ways to reach it, all with different prise en passant squares. Many variations could easily be optimized away (like when nobody can take it en passant), but it would certainly waste quite a bit of CPU time. Open positions with Queens, Rook and a minor piece on both side could be very difficult for computers if many options are left and the pieces have the potential to be taken en passant on many squares. There might be some way to optimize many cases, but humans will probably have an easier time at ignoring useless/equivalent variations.

Hi Simon,

If you are an author of a standard chess engine, I encourage you to write a chess engine to play Option Chess. You bring up some interesting points and it would be great to have them actually documented with a real engine.

The first paragraph above that I've quoted from you is the most interesting one: you are proposing that Option Chess will produce many more duplicate positions in the engine search than standard chess. I don't follow your logic on this, but this would be the most interesting one to have either proven or disproven.

In the second paragraph, I should note that one would not use a double move to move a Bishop from a1 to h8, because that means the same move could be done in a single move. Maybe it's a typo and you meant to type different squares, I'm not sure.

Anyway, again, I hope that the publication of the Option chess article will encourage engine authors to attempt to create Option Chess engines. Right now there may not be much incentive and I would like to offer financial incentive, but that will have to wait for at least a few months. Still, you could start work on it right now.

Cheers,
- Paul

[Simon Valiquette]

Re: Option Chess

The answer I gave Louis for this excellent question is that castling in Option Chess is both a King and a Rook move, therefore if EITHER piece is moved again as the second move of a double ove, that piece (King or castled Rook) cannot capture material on that second move. But the Rook could give check.

That just gave me a new idea. If it is my turn to move, and that I place a green token close to the clock to show that I plan to do a double move, this decision is irreversible once I have released the token, right? So let say that in the following position, Black just played Be3+ déc.



Of course, I put my tocken close to the clock and play Ng1 stalemate! Think about it: I must play my second move since this decision is irreversible, but there is no legal move left while I am not in check anymore!

By the way, I am fine with this. But the following is more tricky.



Again, I use my token and then I play Ka1 Draw! It is a stalemate because a piece that moved twice cannot make a capture, so there is no legal move left! There is plenty of other stalemates ideas based on the use of an option. You may well decide to leave it as it, but perhaps you should at least clarify if it is a valid stalemate idea.


Simon