Arimaa Forum - Print Page


    
      
        Arimaa Forum
        (http://arimaa.com/arimaa/forum/cgi/YaBB.cgi)
      

        Arimaa >> General Discussion >> Push a rabbit to goal temporarily
        
(Message started by: Cobra on Aug 7^th, 2009, 3:16am)

Title: Push a rabbit to goal temporarily
Post by Cobra on Aug 7^th, 2009, 3:16am

Tonight I was surprised to see a bot push a rabbit of mine to goal and then pull him back, without losing. Consider that you can't move your own piece onto a trap and then move another piece to guard the trap. If you follow that order, you lose the first piece. The action is instant--not taken after the whole move is completed.

So I would expect pushing a rabbit onto goal would also be instant in the same way.

Thoughts?

-Chuck
--
http://chuckesterbrook.com/

Title: Re: Push a rabbit to goal temporarily
Post by Arimabuff on Aug 7^th, 2009, 3:57am

That's just the way it is. Every game has its quirks. Consider the "en passant" and "sub-promotion" in Chess, not to mention the complicated set of rules about castling ( e. g. you can't castle if the king is in check in between).

Title: Re: Push a rabbit to goal temporarily
Post by Cobra on Aug 7^th, 2009, 4:04am

Is it considered an essential element of the game in the way that en passant is in chess? Would Arimaa play differently?

Or is it just an accident / coincidence?

-Chuck
--
http://chuckesterbrook.com/

Title: Re: Push a rabbit to goal temporarily
Post by arimaa_master on Aug 7^th, 2009, 4:18am

on 08/07/09 at 04:04:50, Cobra wrote:

It is defined in the Rules of the game (section: "Special situations"):

"
A player may push or pull the opponent's rabbit into the goal row it is trying to reach. If at the end of the turn the rabbit remains there, the player loses. However if the opponent's rabbit is moved back out of the goal row before the end of the turn, the player does not lose.
"

Title: Re: Push a rabbit to goal temporarily
Post by Arimabuff on Aug 7^th, 2009, 4:27am

on 08/07/09 at 04:04:50, Cobra wrote:

It may not be as strategically instrumental as en passant but for one thing if we changed that rule we'd have to reprogram all the bots (and there are tons of them) and that's just too much hassle for such a rarely used rule. I may have used it a couple of times in thousands of games myself. Omar would only be asking for (big) trouble if he decided to change that rule with no sizeable benefit in return.

Title: Re: Push a rabbit to goal temporarily
Post by Manuel on Aug 7^th, 2009, 4:36am

Actually, I can hardly come up with a situation where it could be advantageous, as the pushing on and off the goal takes all your steps (and your back to where you were) anyway!
It could be a way to pass the rabbit, but that can usually be done differently as well.

Title: Re: Push a rabbit to goal temporarily
Post by aaaa on Aug 7^th, 2009, 4:59am

http://arimaa.com/arimaa/puzzles/show.cgi?p=p24

Title: Re: Push a rabbit to goal temporarily
Post by Fritzlein on Aug 7^th, 2009, 6:50am

The rule isn't essential, but it is intuitively in keeping with other rules that the middle of your turn doesn't count, only the end. Two other examples:
1) The repetition rule does not consider positions that occur between each step. So it is legal to create a position for the third time on the way to a new position, even though it is against the rules to create a third repeat at the end of the turn.
2) Eliminating all opposing rabbits is a win even if you sacrifice your own last rabbit on the same turn. So it could be that within the turn, your last rabbit died first or second or simultaneously. It doesn't matter; if, at the end of your turn, your opponent has no rabbits, you win.

Thus the goal rule is at it is for intuitive consistency. None of these rules change game play much, but it is just easier to remember the mantra: the middle of the turn doesn't matter, only the end result.

Title: Re: Push a rabbit to goal temporarily
Post by jdb on Aug 7^th, 2009, 9:31am

on 08/07/09 at 06:50:53, Fritzlein wrote:

Traps are cleared after every step.

Title: Re: Push a rabbit to goal temporarily
Post by Arimabuff on Aug 7^th, 2009, 11:15am

on 08/07/09 at 09:31:48, jdb wrote:

Traps are cleared after every step.

The traps are similar in principle to the freeze rule. A piece can be unfrozen for only one step and so can it be trapped even for one step, with the notable difference that for the latter it’s irreversible. The game would be radically changed if you could only move a piece when it is unfrozen at the start of the move.

Title: Re: Push a rabbit to goal temporarily
Post by Fritzlein on Aug 7^th, 2009, 11:33am

on 08/07/09 at 09:31:48, jdb wrote:

Traps are cleared after every step.

Sure, traps are cleared after every step, and the position of the pieces changes with every step, and all of this "matters" in an obvious sense. To be clear I should have said that the middle of the turn doesn't matter for victory conditions. The three examples (rabbit temporarily on goal, repetition of position, elimination of all rabbits) all have to do with the victory condition appearing in the middle of a turn. I could add a fourth example: your opponent might be immobilized in the middle of your turn, and that doesn't cause him to lose the game either if you take another step that unfreezes one of his pieces.

I see an intuitive consistency there, but I can understand the dissonance with things that change in the middle of a turn, like being safe on a trap or being frozen. There was a rules question on BoardGameGeek along the same lines, so I guess it is natural to consider that victory status changes with every step too.

Maybe intuitions differ. And maybe it is all an accident of the algorithm. Perhaps Omar originally set up the server to check for victory only between turns, not between steps, and the rules were derived from the implementation rather than vice versa. :)

Title: Re: Push a rabbit to goal temporarily
Post by Cobra on Aug 7^th, 2009, 11:40am

Well this has definitely been an interesting thread. I agree that this does not have a huge affect on the game, and I'm not suggesting any changes. I was just surprised when I saw it. After ~ 200 games, I don't think I had seen that before which shows how rare it is.

I was thinking: "Hey, where's my victory?" ;D

So in summary:

For victory conditions, the middle of the turn does not matter.

Thanks for all the clarifications.

Title: Re: Push a rabbit to goal temporarily
Post by SpeedRazor on Aug 9^th, 2009, 10:35am

I've only just recently gotten my head around how rabbits have backwards powers (and now it makes sense to me), but these two other rules mentioned above seem to maybe not belong in the game.

I understand that rabbits should have the option of being sacrificed in traps, but I can't conceive of a situation where it would be advantageous to sacrifice your last rabbit on the same turn where you win the game by other means: rabbit reaches the victory line, or you eliminate your opponent's last rabbit. Can anyone come up with a scenario where sacrificing your last rabbit is logical? It's like putting your king into checkmate in chess. Intuitively, it seems to me that you must always have the suffecient potential victory condition (mating material) in hand to win: at least one rabbit remaining on the board.

There does seem to me to be a logical consequence to the notion of being able to sacrifice your last rabbit. And that is: that upon your last rabbit being eliminated, your next least valuable piece - the cat - becomes your victory condition, and all that would entail: the cats can no longer move backwards, push/pull anything. If your cats are eliminated, then the dog has the new victory conditions, et cetera... If you are allowed to sacrifice your last rabbit, than this 'ennobling' of your next piece up seems the obvious consequence (plus the penalty of not being able to move backwards).

Personally, I would rather have the option of sacrificing your last rabbit rule removed - unless there really is some useful reason for doing it.

Title: Re: Push a rabbit to goal temporarily
Post by Fritzlein on Aug 9^th, 2009, 4:28pm

on 08/09/09 at 10:35:14, SpeedRazor wrote:

I understand that rabbits should have the option of being sacrificed in traps, but I can't conceive of a situation where it would be advantageous to sacrifice your last rabbit on the same turn where you win the game by other means: rabbit reaches the victory line, or you eliminate your opponent's last rabbit.

In fact, we had to clarify the rule precisely because JDB came up with such a situation. I'm not sure if this is the same one he produced, but consider Ee6 Rf6 ef7 cg6 rd6. Gold can capture the last silver rabbit, but only by sacrificing his own last rabbit. This is why the rule matters, although I don't know of any game in which it has arisen. In fact, I can recall only one game off the top of my head in which both players were even reduced to one rabbit.

Title: Re: Push a rabbit to goal temporarily
Post by SpeedRazor on Aug 9^th, 2009, 6:30pm

Hhhmmmnnnn..... What an position! Still contemplating...

Amazing position! Extra-ordinary ... are you an chess master, Fritzlein? This is so like an chess endgame, but with mutual zugzwang...

My Retrograde Analysis Intuition radar says that this position can't be reached without one (or the other) player overtly missing a win...

Title: Re: Push a rabbit to goal temporarily
Post by SpeedRazor on Aug 9^th, 2009, 8:05pm

Wow. Whoever goes first: loses. (This position needs an rule clarification, maybe). Absolutely fascinating position! But...

Let's say it's Golds move. Her only realistic move - as she's losing the game horribly - is to "push" the d6 rabbit into the c6 trap. But which happens first? The instant the Gold Elephant moves, F6 rabbit goes down the trap (my first opinion). Somebody must win. But whom? Gold is completely busted, materially. Whomever is on the move is in zugzwang - (zoogz•vahng). Can... a victory condition be evaluated between steps of your up-to four-step move? Would the instant Gold moves her Elephant (condemning her last rabbit to the F6 trap, of which she's guarding) happen before the c6 trap drops the last Silver Rabbit: an 2-step process)?

I ... kind of feel like a jerk because I'm putting up problems for this great game. I don't want to be that problem. But if I don't say it now, won't somebody just point out that we glossed over things earlier on, later? In 2112, Chess, Go, Arimaa, will be huge. Let's do the math, now...

[Speedy]

Title: Re: Push a rabbit to goal temporarily
Post by Janzert on Aug 9^th, 2009, 10:37pm

Not sure what clarification you're thinking the rules might need, but they do already cover both players losing all rabbits in one turn. From the rules page (http://arimaa.com/arimaa/learn/rulesIntro.html):

Quote:

If in the rare case both players lose all rabbits on the same move then the player making the move wins the game.

Also as a side note, Bomb actually lost a game (#86642) because it apparently incorrectly implements the rule for allowing an opposing rabbit to temporarily get pulled to the goal line. To be fair though the situation only arose because Gnobot also incorrectly understood the rule at the time. I adapted the situation from that game to the following position which shows as an immediate forced win if a bot has the rule wrong and should be an eventual loss if the rule is understood correctly.

43g
+-----------------+
8| . . d . . . . . |
7| r r r C . . . . |
6| D m R E . x . r |
5| d H . . . . . . |
4| . . . . . . . . |
3| . . x . . x . . |
2| . . . . . . . . |
1| . . . . . . . . |
+-----------------+
a b c d e f g h

Janzert

Title: Re: Push a rabbit to goal temporarily
Post by Cobra on Aug 9^th, 2009, 11:08pm

on 08/09/09 at 20:05:51, SpeedRazor wrote:

Can... a victory condition be evaluated between steps of your up-to four-step move?

No, that was the upshot of this discussion as I mentioned earlier:

on 08/07/09 at 11:40:55, Cobra wrote:

So in summary:

For victory conditions, the middle of the turn does not matter.

You expressed some concern that you might be hurting Arimaa in some way, but in actuality Arimaa's rules are very refined with the results of:

(a) lack of ambiguity about how to play

(b) almost (or is it entirely?) no chance of a draw

What needed refinement was the knowledge base among players such as you and myself. :-)

-Chuck
--
http://chuckesterbrook.com/

Title: Re: Push a rabbit to goal temporarily
Post by Fritzlein on Aug 10^th, 2009, 6:49am

on 08/09/09 at 20:05:51, SpeedRazor wrote:

I ... kind of feel like a jerk because I'm putting up problems for this great game. I don't want to be that problem. But if I don't say it now, won't somebody just point out that we glossed over things earlier on, later? In 2112, Chess, Go, Arimaa, will be huge. Let's do the math, now...

The math is already done. JDB raised the problem, Omar resolved it. Gold captures the last silver rabbit while losing his own last rabbit, and thereby Gold wins the game. There is no ambiguity. Been there, done that.

The example only came up because you were claiming that the rule was unnecessary. Now you know the rule is necessary. Aren't you glad we already thought it through? ;)

on 08/09/09 at 18:30:11, SpeedRazor wrote:

My Retrograde Analysis Intuition radar says that this position can't be reached without one (or the other) player overtly missing a win...

That is probably true of the example that I gave you. However, the bare existence of such a position leads me to suspect that a similar position could arise without either player missing an obvious win. It hasn't happened yet, to my knowledge, but as Janzert pointed out, we all thought that "rabbit temporarily on the goal line" was just a theoretical problem until it popped up in an actual bot game. Maybe it will turn out similarly for "simultaneous last rabbit death".

Title: Re: Push a rabbit to goal temporarily
Post by Fritzlein on Aug 10^th, 2009, 7:02am

on 08/09/09 at 23:08:02, Cobra wrote:

(b) almost (or is it entirely?) no chance of a draw

Whether there is no chance of a draw or not depends on your philosophical outlook. The repetition rule insures that the game is finite, and there are no in-game drawing conditions, so theoretically one player or the other must win. Arimaa is theoretically drawless.

BUT what if there is a win on the board from the repetition rule only (i.e. not from goal, elimination, or immobilization), and what if that win takes a million moves to kick in? For all practical purposes such a position is drawn. Chess recognized the futility of such positions with a fifty-move draw rule.

BUT Omar has imposed a time cutoff rule. If the game doesn't end within a certain time, we stop and count the pieces. Whoever has more pieces wins, and if no captures were ever made, Silver wins. So Arimaa is drawless in practice as well as in theory.

My viewpoint is that we are lucky the time cutoff never comes into play except in games between dumb bots. If human games started to be decided by time cutoff, that would be a serious blow to the viability of Arimaa. It would call into question whether Arimaa is really worthy to stand alongside chess and Go. If Arimaa is drawish by nature, then trying to patch it up with a rule hack isn't going to help anything. In fact, if Arimaa is drawish by nature, we should probably allow drawn results rather than imposing arbitrary victories.

Lucky for us, all the evidence so far is that Arimaa is not drawish by nature. The time cutoff has never come into play in a human vs. human game, at least as far as I am aware. So the drawless nature of Arimaa is an observation from experience, not a mathematical proof from the rules.

Title: Re: Push a rabbit to goal temporarily
Post by SpeedRazor on Aug 10^th, 2009, 2:24pm

Thanks, virtually all of my questions have been answered. And thank you for treating me in moderate tones, as I meant no ill-will.

Two interesting questions in my mind: 1.) The greatest historical Arimaa player - Fritzleine (my humble opinion) - still has questions about Arimaa himself (as it should be): not exactly a question as a point; but still... And 2: I'm still studying Janzert's endgame problem. Who's on move? Probably not necessary to know as the best problems chess problems. I'm looking forward to figuring it out, or not...
Thank you all, [Speedy]

Title: Re: Push a rabbit to goal temporarily
Post by Fritzlein on Aug 10^th, 2009, 4:01pm

Thank you for your questions, SpeedRazor. As for my own doubts about Arimaa, it is a function of being old enough to have been wrong many, many times. We can't know whether Arimaa will stand the test of time until, um, well, until it stands the test of time.

In my youth I got a second-hand book on Teeko by its inventor/promoter John Scarne. He explained six ways to Tuesday how Teeko was the best game ever, and would displace checkers and chess as the pastime of choice. Unfortunately, it turned out that Teeko was a dead draw. However, Scarne didn't turn down the hype meter even by a notch. Instead he changed the rules, called it "Advanced Teeko", and claimed again it was the best game ever. Unfortunately, it turned out that Advanced Teeko was also a dead draw. Scarne sold a ton of Teeko sets due to his tireless promotion, but it turned out the game just wasn't very deep after all.

Teeko was before my time (and you have probably never heard of it despite its temporary success), but I am old enough to have lived through the Pente craze. Listen to how the inventor Gabrel describes it:

Quote:

One year I logged 50,000 miles in a van selling PENTE at craft shows from New Orleans to Denver. Then my persistence paid off. I began to transform the game visually to make it the beautiful game it is today. It caught on in night clubs as an alternative to backgammon, and stores from Neiman Marcus to Macy's couldn't get enough. We were on the verge of making it big on our own. But, in '83 we just ran out of the steam and money needed to push PENTE over the top. PENTE was sold to Parker Brothers. Unfortunately, the game and toy industry is a fickle one, and Trivial Pursuit took center stage at Parker Brothers.

Gabrel thinks the success of his game was all about marketing, and he did market it tirelessly. But he is deluded to think that if only he had had the money to get over the hump, or if only Parker Brothers hadn't been distracted by Trivial Pursuit, then Pente would have become the next chess. What is missing from Gabrel's account is the fact that the time things started going downhill for Pente was just after the 1983 Pente World Championships when the winner Rollie Tesh declared (in essence) that the rules for Pente were broken due to too large a first-player advantage. World Champion Tesh proposed alternate rules, and Pente sales took a nose dive.

It has happened before that great games have turned out to be broken in some way that was not immediately obvious, and it will happen again. Good marketing can create fads, but only good games last forever. If it turns out that Arimaa is flawed in some way, I'm not going to be the devotee in denial, claiming that nothing is wrong and that the rest of the world is blind to the virtues of my favorite game. When I say that Arimaa has revealed no flaws yet, I'm being as objective as I can be, and if Arimaa reveals flaws in the future, I'll do my best to admit that candidly as well.

How many abstract strategy games make the big time? Teeko, Othello, Pente... it's a short list. I am fervently hoping that Arimaa becomes the latest craze on that list. If Arimaa gets played millions of times all over the world, then (and only then) will we know whether it is a truly great game. Until such time there simply won't be enough eyeballs on it to ferret out all possible problems.

Title: Re: Push a rabbit to goal temporarily
Post by Janzert on Aug 10^th, 2009, 5:44pm

on 08/10/09 at 14:24:05, SpeedRazor wrote:

I'm still studying Janzert's endgame problem. Who's on move?

The line above the position diagram gives the move number and side to move. So 43g means the 43rd move of the game with gold to move. See the Game and Position Notation page (http://arimaa.com/arimaa/learn/notation.html) for more details.

Also when I said "immediate forced win" it is not a literal goal in one, I was meaning a bot would find it very quickly. If a rabbit is not allowed even temporarily on the goal line then it is a forced goal in two for gold.

Janzert

Title: Re: Push a rabbit to goal temporarily
Post by omar on Aug 11^th, 2009, 3:47pm

Wow, a lot of posts on this. As Karl pointed out the evaluation for win/loss condition is done between turns and the traps are checked after each step. However, a push/pull is atomic in the sense that even though it is executed with two consecutive steps, you have to think as if the two pieces are moving simultaneously. Thus a stronger piece can suicide into a trap while pulling an opponents piece.