Comments/Ratings for a Single Item

?t=l worked I got the new version but I have no some trouble with the antivirus after I solve those I see no reason why winboard wouldn't run the new version!

H.G.

It now works pawns and everything, but here is what I've done I hope not to cause you any trouble.

After download in the initial run of the fairy-max5.0b2 avast antivirus started a scan and I interupted it abruptly. I think it has thought that your program did and now it says avast has this program under scrutiny.

Don't worry about me I just abort the antivir and is fine. Fairymax works.

I honestly hope I haven't caused you any trouble.

No trouble at all. I am glad you detected this double-push bug.

I think it would benefit Fairy-Max' strategy when you assign the piece values such that the minors have values just above 256, Rooks 400-500, super-pieces starting just above 768. There is no harm in lowering the Pawn value to around 60. If you set the values much higher Fairy-Max will be to slow in recognizing the end-game started and it should involve its King, and too reluctant to push Pawns in the opening.

I think Fairy-Max could be a very useful tool to determine piece values, despite the problem that it cannot implement the exact promotion rule. Because piece values will not be very dependent on the promotion rule. So you could for instance play from a start position where white has to Griffins, and black two Aancas, and see which side wins more often, to determine if the Aanca or the Griffin is the stronger piece on this board. If, say, the Griffins would win this convincingly, you could remove one of their Pawns from the start position, to see if that is enough to flip the odds.

With your latest promotion rules it becomes kind of hard to decide what should be Fairy-Max' default promotion choice, as it would probably still be better on average to always promote to minor on 8th than to always defer there and take a Rook on 9th. But since the minors are all very close in value in Apothecary 2, it would depend very much on the situation which one would be best. But I would go for the Bishop, as, being a slider, this has the biggest chance to stop an enemy promotion in case of a promotion race.

Actually for now I'm enjoying the fruits of my and especially your work H.G. . Apothecary 1 is quite slick and interesting from the games I see it has an interesting assortment of minor leapers, and the aanca and griffin are quite new on the play field. Apothecary 2 is a bit more stale with not so interesting leapers and usual power pieces. Tomorrow I'll be starting serious work.

H.G. I understand that the power of Aanca and Griffin are virtually unknown, so there is research to be made, but I have to take care of both of my twins so we could see a game with 5 elephants versus 6 zebras or something of sorts. I initially gave more points 3.7 vs 3 to the elephant as the just move enhancement works very well with the rest of the powers. Now I think it's a bit much.

It is probably not a good idea to try too many pieces of the same type (especially weak pieces), because that would allow them to protect each other in an unnatural way. (Unless they are Pawns, of course, as for those this is natural.) So I would not go beyond two Elephants against two Zebras.

Ok, thanks!

About the promotion rule, I'm not that sure that you are correct, waiting 1 turn to get a rook is a big deal, what can the oposite bishop do in 1 turn to turn the tables. A rook can win you the game. Also now you have to pay 1 turn from rook to queen but I don't think this will come into play unless very weird situations like many pawns vs minor piece and the rest of pieces the same, although the rest of the pieces will be used to capture/block pawns!

I left Fairy-Max at 15min+15secs incr. twice, once with each initial position and not that 2 points are statistically relevant but the game seems to last 70-80 moves ( maybe a bit more for apothecary 2 that is the slower game), so not that bad.

It would probably be good enough to test at 40 moves/1 min. There still seems to be a problem running matches starting from set-up positions with 'engine-defined variants', where the engine also sends a position. At least in the WinBoard version I distributed in the Sac-Chess package. I will check if the latest WinBoard also has that problem.

- waiting 1 turn to get a rook is a big deal, what can the oposite bishop do in 1 turn to turn the tables.

The Bishop might be able to attack the square in front of the opponent passer, so that he cannot promote at all. This is only a 50-50 chance, because the Bishop might have the wrong color, but the chance that a Knight, Zebra or Camel (or a Wizard or Champion) can do that in one move seems much smaller. And if there are other, blocked Pawns elsewhere on the board, you might win much easier with a minor against nothing than with a Rook against a minor. Which is what you would get if you delayed promotion to go for the Rook, while in the mean time he would promote to minor.

If I decrease the time that much Fairy-Max runs out of time all the time(! and - signs). I also wanted to see how stronger games go, because all the games I've seen so far were pretty imbalanced to a lesser degree the 2 pairs of games at 15 min. The real tests will be at shorter time spans in order to make more points on the graphics.

Also I have a slow computer. Just a probook Laptop.

KPvsK still a win, I forgot about it, and is very important. It seems my rule devalues promotion rather than giving extra options as it was intended. It could work if you fine tune a variant just for it. It is not the case here. I think postponing for a rook could still work in many cases, I just don't see postponing for a queen happening to soon.

KPK would definitely be a case where you must promote to at least a Rook, and can afford to do it. If you would use Fairy-Max for piece-value testing, it would probably be better to keep Rook as the promotion choice in the entire zone. I don't expect the details of the rules to affect piece values much.

In fact Fairy-Max can be configured to use different promotion piece for white and black. This was needed for asymmetric variants like Spartan Chess. It is a bit tricky, though: white always promotes to piece #7, but if black does not have a piece #7 in its initial setup (as Fairy-Max thinks it should be, so the one given in the 3rd and 4th line of the game definition), but white has one, but black has a piece #9, it uses the latter as promotion piece for black. So you could make piece #7 and piece #8 both Bishops, and have white use #7 and black #8 for the Bishops in the initial setup. Then you can define Rook as #9. Black Pawns would then promote to Rook, and white Pawns to Bishop. You can then play many games to see how much white is handicapped by this. My guess is that it would be almost nothing.

I must say I'm pleasantly surprised by the way Fairy-max handles openings. Being said that I used a generous time counter 15mins+15secs, I've seen pawn sacs and fianchetto bishops..

A weird thing is not until now there were no draws in 4 games of each incarnation. Fairy-max usually avoids repetition draws, which is also cool.

I have completed my general view on the 2 twin games:

There were six games of apothecary 1 and six games of apothecary 2.

Scores:

Apothecary 1: white 3-3 black no draws

Apothecary 2: white 2-4 black no draws

Length:

Apothecary 1: 50-80 moves

Apothecary 2: 70-90 moves

Piece values used by Fairy-max:

Apothecary 1

pawn:85
knight:272
bishop:286
wizard:258
champion:272
rook:510
aanca:612
griffin:748
queen:816

Apothecary 2

pawn:85
knight:231
bishop:286
elephant:251
camel:218
zebra:204
rook:510
archbishop:680
marshal:748
queen:816

The high rook was an accident It was supposed to be 493, but I won't redo this.

Now I'm starting the serious experiments. I'll use values normalized to a pawn of 60 as H.G. suggested, and the rook lowered acordingly.

Here are the result of some preliminary experiments I've done:

bishops in aancas white:1.5
bishops in griffins black:4.5
1 draws (draws enter in the points above)

knights in aancas white:4
knights in griffins black:2
0 draws (draws enter in the points above)

bishops in griffins white:5
bishops in aanca black:1
0 draws (draws enter in the points above)

knights in griffins white:2.5
knights in aancas black:3.5
1 draws (draws enter in the points above)

total:
games 24 from which draws:2
aancas poins:10
griffins points:14

[edit:]

Here I pited an army with 2 aancas against an army with 2 griffins.

the promotion rule was rook at 8 rank.

So the total number of wins for griffins is slightly higher but not convincingly. Now I believe that the difference between them is less that a pawn. I think I have to continue the experiments as things don't seem to go any particular way, yet. If the 41% wins for aancas hold then the distance between aanca and griffin is probably small then a pawn. I'm not sure though. Then I'll do something else maybe take a pawn from the griffins or give a pawn to the aancas.

On second thinking I should find out how much a pawn translates into a win, in both apothecary games!

Indeed, you would have to know how much deletion of a Pawn would change the result. E.g. if the Griffins beat the Aancas by 60% with equal Pawns, but only score 45% with an additional Pawn handicap, you know the Pawn was worth 15%, so that the original superiority of the Griffins was 2/3 of a Pawn (i.e. 33cP per Griffin).

Note that the random error of the average result in N games is 50%/sqrt(N). So in 100 games the error would be 5%. If a Pawn indeed corresponds to 15%, that would mean the random error in the value determination is 33cP. By playing 2 vs 2 that error is divided by 2 in a singlepiecevalue, i.e. you would measure the value to an accuracy of 1/6 of a Pawn. To make the error twice as small you already need 400 games. This is why I suggested to use time controls as fast as possible, If you can do 12 games per hour you can do 100 games overnight. (And if you have a multi-core CPU, you could play several games in paralel.)

In any case it seems that the Griffin - Aanca value difference is not spectacularly large,

The difference between aanca and griffin is  much smaller than I expected indeed!

I use as time control 2 mins for 30 moves as my computer is rather weak but it has 2 procs indeed.

How can I setup experiments over night or while I do other stuff? I have to copy paste the initial position everytime anyway and the game does not just restart, Also if I could restart it how do I get hold of the results?

H.G.

I think fairy max has some clock problems, it runs out of time on ocasions. Is that normal?

I'd like to conclude this discussion with the hope that grandmaster level chess will eventually evolve to include variants like Grand Chess and Omega Chess or why not for the really weird ones my 2 apothecary variants I proposed during this discussion.

Fergus wrote on 2016-09-13: None

...Chess has been finely honed by natural selection to be free of arbitrariness. Every rule and piece in Chess serves a purpose, and none are arbitrary. Since Chess is what won the survival of the fittest among Chess variants, I expect that any variant capable of succeeding Chess would also have to be free of arbitrariness. But most Chess variants differ from Chess through some arbitrary change to it, and they easily get lost in a sea of variants that each differ from Chess in their own arbitrary ways...

In a way I like seeing this opinion, as a chess player (who also has fully recovered from some loss of faith in chess due to computers). However, I'm fairly sure you didn't mean to say that chess has so far been clearly superior in terms of merit to all chess variants. That would be a little disconcerting, even to me (after my venturing into the world of variants), since I concluded that among the dozens of variants I looked at (however briefly), many even in my eyes had compensating merits for anything they might lack compared to chess - some are quite different to chess and are hard to fairly compare. What e.g. Circular Chess lacks in terms of basic mates compared to chess, it makes up for in certain other ways (though initially these ways did not impress me so much), and Circular Chess may well not even be one of the better variants objectively (if objectivity is possible comparing variants).

I'd also wonder a little about whether chess doesn't have any arbitrary aspects to it, too (some might say any game or sport must have some arbitrary rules/kludges). For example, stalemate being a draw could be ruled as a win (or 3/4 of a point) for one side, instead (but that would spoil many fine stalemating combinations/swindles, besides altering current endgame theory). One thing 8x8 chess has going for it is that bishops are very close to knights in value (even for those who quibble), so variants on other board sizes and/or shapes may lack this nice feature. However, in Glinski's Hexagonal Chess, for example, where apparently a knight is worth a pawn more than a bishop, interesting trades of knight for bishop and pawn can frequently occur (I think), rather than bishop for knight as in chess, making the nice equivalence of a B for N in chess something that's not so meaningfully special.

I think what has made chess so popular is that given its rules & 8x8 terrian, it works remarkably well (e.g. in producing many brilliant games between people), and no one seems to know exactly why. Still, remember that Shogi & Chinese Chess are considered "Classic" variants, too.