Arimaa Forum - Print Page


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

        Arimaa >> General Discussion >> BotBash Challenge of the Month
        
(Message started by: 99of9 on Apr 12^th, 2005, 6:33am)

Title: BotBash Challenge of the Month
Post by 99of9 on Apr 12^th, 2005, 6:33am

When botbashing began someone suggested that one day we might have a challenge of the month. I am now suggesting that we actually start this. And to kick it off, I'm proposing the first challenge:

(Actually it was arimanator's suggestion in response to robinson's game 1362 where he made quite a comeback from some blunders.)

By the end of April, who can beat Bomb2005CC having sacrificed their camel during the first move??

I'm going to have a go now - I don't expect to crack it first try... so hopefully there should be plenty of play in this as we exploit the weaknesses of the best arimaa bot in the universe.

Title: Re: BotBash Challenge of the Month
Post by PMertens on Apr 15^th, 2005, 5:21pm

I just made a succesful try against the fast version ... seems not so hard :-P

Title: Re: BotBash Challenge of the Month
Post by 99of9 on Apr 16^th, 2005, 3:21am

Well I want to at least see people *try*ing against the real slow version. If we won it would be a very convincing demonstration of human superiority - I've never seen anyone even contemplate giving Frtiz a deliberate camel advantage!

Title: Re: BotBash Challenge of the Month
Post by PMertens on Apr 17^th, 2005, 11:48pm

veni vidi vici ;D

Title: Re: BotBash Challenge of the Month
Post by 99of9 on Apr 18^th, 2005, 12:31am

Wow, that's amazing Paul!

and rated too!!!

I can't wait to look through this game.

To me, the fact that you have achieved it shows us even more than a 7-1 victory. This shows just how wide the gap is. In good human vs good human games I'd imagine a free camel is worth about 400-500 ratings points.

Title: Re: BotBash Challenge of the Month
Post by PMertens on Apr 18^th, 2005, 1:08am

you wanna give me a camel free in our next game ? :o

Title: Re: BotBash Challenge of the Month
Post by 99of9 on Apr 18^th, 2005, 1:20am

Only if your baby keeps up the good work :-).

Title: Re: BotBash Challenge of the Month
Post by PMertens on Apr 18^th, 2005, 6:12am

I would not call it great play.

Botbashing is more about knowing your opponents behavior than perfect play, but it certainly helps not to blunder to much :-)

Title: Re: BotBash Challenge of the Month
Post by PMertens on Apr 21^st, 2005, 3:50pm

btw, it could be put on the botbashing page to inspire future generations :-)

Maybe I will try Camel and cat next ;-)

Title: Re: BotBash Challenge of the Month
Post by omar on Apr 27^th, 2005, 3:22pm

I didn't notice those games; does anyone have the game numbers ?

Title: Re: BotBash Challenge of the Month
Post by PMertens on May 7^th, 2005, 11:26am

I really liked toby's idea ...

there is a new month now and since I solved the last challenge first, I feel free to raise the bar a little bit ...
(I just hope for more competition this time ;-) )

I think the camel was relatively easy, since bomb has special feelings for camels ...

How about a horse and a cat this time ?
It does not really matter if played against fast or slow version, since the "how to play this" part is the same with both bots.

Title: Re: BotBash Challenge of the Month
Post by Fritzlein on May 8^th, 2005, 1:30am

I think Omar's idea is to maintain botbashing records only for bots with fixed strength. That's what he created the P1 and P2 series bots. Perhaps one should specifiy one of those bots along with the challenge?

Title: Re: BotBash Challenge of the Month
Post by 99of9 on May 8^th, 2005, 3:15am

In the 99 system a horse and cat should be *easier* :-).

I just hope we don't have to do it rated a-la-Paul. I'll give it a go.

Fritz I agree that not all these results should go on the botbash page since the bots are machine dependent. But I like the idea of having some of these challenges against the current best bot. P2 is quite a handicap. Nevertheless, it's a good idea to have the previous winner choose the next challenge - so you're sure to get to do some nominating soon Fritz :-).

Title: Re: BotBash Challenge of the Month
Post by PMertens on May 14^th, 2005, 12:16pm

ah, the interest in this kind of game seems somewhat limited ...

Even if I am certain it shows the weaknesses of bots far more clearly ...

Title: Re: BotBash Challenge of the Month
Post by 99of9 on May 15^th, 2005, 2:15am

I think in part it's because we both chose such hard challenges. Next time I get the chance I will choose something that nearly everyone can do, but challenge people to do it quickly / innovatively.

Title: Re: BotBash Challenge of the Month
Post by PMertens on May 15^th, 2005, 3:35am

hmmm - I always thought the definition of challenge is äh something that is not easy to be done :-P

Your camel challenge was great, because noone would usually do that and it proved just how clueless bomb really is and how easy to outplay.

I liked the "Kill all rabbits first with no piece crossing the middle line" challenge (did someone actually win that one ?)
Just another fine example of finding a terrible weakness and making a wonderful lesson in defense out of it ...

A "beat bot in x moves challenge" would not give me the same thrill :-/

Title: Re: BotBash Challenge of the Month
Post by PMertens on May 15^th, 2005, 3:53am

no piece of YOUR army (except the phant)

Title: Re: BotBash Challenge of the Month
Post by 99of9 on May 15^th, 2005, 8:45am

Yes, it was a rabbit-pulling exercise, when bomb was too scared to ever go on the attack.

I figured it was possible to play lone elephant forward for the entire game. But often after 5 or 6 rabbits the situation got complicated and bomb would eventually pull one of the human's other major pieces forward of the halfway line.

Title: Re: BotBash Challenge of the Month
Post by Fritzlein on May 22^nd, 2005, 12:00am

Well done, Arimanator! I thought it was a pretty hard challenge; I only tried it once myself before giving up.

I have a much easier challenge to propose, but one that is mind-bending in a good way: start with all your rabbits on the second row and all the stronger pieces on the first row. I've done this against my home version of Bomb, but I suspect it will be slightly harder against a more aggressive bot, so I propose Clueless Blitz as the target bot. Any takers?

Title: Re: BotBash Challenge of the Month
Post by 99of9 on May 22^nd, 2005, 7:25am

on 05/21/05 at 10:04:20, Arimanator wrote:

I did it ;D Not only I killed all the rabbits but also all the pieces without any of my pieces (except the elephant) crossing the middle line ;D QED

Congrats, I'm glad you proved it doable - that will give David a few things to think about over the next year. Your extension is nice too. Does this mean you were left with all your pieces vs a lone elephant? What a humiliation. I'll go take a look right now.

Title: Re: BotBash Challenge of the Month
Post by PMertens on May 24^th, 2005, 12:34am

Game 15122 is a nice example of "I made it I made it ... argh d**n ..."

>:(

Title: Re: BotBash Challenge of the Month
Post by PMertens on May 24^th, 2005, 12:36am

btw I am a little bit disappointed that only 3 players tried toby's last month challenge and only 0 mine for this month :'(

Title: Re: BotBash Challenge of the Month
Post by robinson on May 24^th, 2005, 2:39am

hey paul, Game 14596

Title: Re: BotBash Challenge of the Month
Post by PMertens on May 24^th, 2005, 11:57am

oh - ok .... but you lost ?

Do you think it was harder or easier than a camel ?

Title: Re: BotBash Challenge of the Month
Post by PMertens on May 26^th, 2005, 1:39am

once again I could not beat toby's challenge ...

therefore I got a very special ending :-)

http://arimaa.com/arimaa/gameroom/comments.cgi?gid=15180

Title: Re: BotBash Challenge of the Month
Post by PMertens on May 26^th, 2005, 9:46pm

finally

http://arimaa.com/arimaa/gameroom/comments.cgi?gid=15194

Title: Re: BotBash Challenge of the Month
Post by PMertens on Jun 20^th, 2005, 6:04pm

the bar has been raised:

Bomb Lightning
Camel, dog and 2 cats - go, beat that 8)

Unfortunately it was only victory by repeated position and probably would have been by points if I had the nerve to play til the end (I can last minute push around rabbits quite easily ;-) )

Omar, please add that to the Botbash-page ;D
http://arimaa.com/arimaa/gameroom/replayFlash.cgi?gid=16102&s=w&client=1

I expect someone to beat bomb with a camel+2 dog handicap the old fashioned way soon.
Show some disrespect to those bots.

Title: Re: BotBash Challenge of the Month
Post by 99of9 on Jun 20^th, 2005, 6:42pm

I thought bomb was resistant to repetitions? I wonder if Omar has made that change to the rules already and this is what messed bomb up??

Title: Re: BotBash Challenge of the Month
Post by PMertens on Jun 20^th, 2005, 7:07pm

agreed, I was a little surprised there ... :-/

Title: Re: BotBash Challenge of the Month
Post by RonWeasley on Jun 27^th, 2005, 8:27am

I would be interested in comments by our successful bashers describing the bot weaknesses they are exploiting and what could be done to the bot to address each weakness.

For example, to me it looks like horizon effects. In the E-H attack, flipping minor pieces from e7 to e5 and then pushing them to f4 and then f3 works mainly because the bot can't see three moves ahead. Humans avoid e7 and d6 in this setup. In the MDD odds game (congratulations Arimanator for finding a successful pattern), the bot can free the minor piece from d4 quite easily, but it doesn't. My guess is that it doesn't project the push to e3 and then f3, so it doesn't see the need to defend.

Is the remedy deeper searches or additional terms in the evaluation function? Bot developers would benefit from the commentary as would players.

A good debate for players is whether a bot bashing strategy would be sound against an opponent who understands how the strategy works.

Title: Re: BotBash Challenge of the Month
Post by PMertens on Jun 27^th, 2005, 10:23am

I will only talk about those sacrifice games for the moment

Quote:

A good debate for players is whether a bot bashing strategy would be sound against an opponent who understands how the strategy works.

The strategy used for M - MDDx cannot work against humans.
In a way it is based on a horizon effect, but I doubt that 4-8 ply would help.

In theses games we can see several "stable" positions, where the bot cannot see any significant change in the near future.
From those stable positions the human can try to slowly manouver a weaker piece (far easiest is a rabbit of course) into a position where it cannot be saved anymore without doing a "bad" move.

As I said ... you could either solve this with more ply but I think a better eval would be far more useful.

There are several examples where bomb simply makes the wrong move because it does not see how much worse it will be on the long run.

Bomb might be a killer in trap control and goal search but it has no plan how to proceed when its in one of those stable positions - as long as it does not know how to attack we will always be able to do that.

But as I said - thats not worth a penny against a mediocre human (or an agressive bot).

Title: Re: BotBash Challenge of the Month
Post by PMertens on Jun 27^th, 2005, 11:24am

I think a P3/P4 version of bomb might give us more insight - would that be possible, omar ?

Title: Re: BotBash Challenge of the Month
Post by 99of9 on Jun 27^th, 2005, 8:34pm

Arimanator says:

Quote:

I hear some people say that Bombblitz is not consistent or sometimes plays strange moves.

Apparently those "strange moves" didn't seem to bother them that much when they were well ahead of botbashing. But, when you can't have something it seems that the next best thing is to spoil it to others.

These posts earlier in this thread are relevent:

on 05/08/05 at 01:30:23, Fritzlein wrote:

Then I replied:

Quote:

Fritz I agree that not all these results should go on the botbash page since the bots are machine dependent. But I like the idea of having some of these challenges against the current best bot. P2 is quite a handicap.

I'd just like to reiterate my point of view: Achieving a "challenge" against a very good, but inconsistent bot, is a very good achievement. By saying that a bot is inconsistent - I am not trying to denigrate it. Well done to all who have beaten bomb with more than a camel handicap - that was my original challenge, because at the time I thought it was achieveable, but near the limit. (I agree with your opinion Arimanator that MDD may not be possible for bomb playing at the full time control - but clearly M was not the limit!)

However as Fritz pointed out a long time ago, repeatability and consistency are important when tables like the botbashing table are recorded. No-one wants the possibility open that the bot they have the record for bashing was "just having a bad (high load) day". That is why I asked Omar to remove some of the earliest bot_arimaazilla records, which were obtained when arimaazilla was broken.

I agree with you Arimanator that all botbashing records are exploiting "strange/poor moves". That's the whole point of botbashing afterall! We prove that we are better than the bots, because they make stupid moves. BUT... sometimes strange moves are so strange that they are actually a clear demonstration that something is wrong (either a bug or a very high load/slow computer/too-fast time control). If it's a bug, that's fair game, because it's part of the bot. But if it's the load, then it might be better to do the demonstration another day. The problem is, we as users just can't tell! So just to be on the safe side, I personally would choose not to list load-dependent records. [Although I am actually not against making a different list or putting a separator in between load-dependent and load-independent bot bashes.]

I'd hazard a guess that in 15 years time, computers will be way faster, and it will not be possible to beat this version of bomb_blitz with MDD, no matter what moves you play, so players from that era could rightly complain that it's unfair to fight a record that is physically impossible to match.

Now finally, and importantly, I do not think that there were major spikes in load during any of the recent spectacular wins. I am not saying your achievements are poor at all. I myself tried MDDC and failed utterly a number of times.

Title: Re: BotBash Challenge of the Month
Post by PMertens on Jun 27^th, 2005, 8:53pm

Quote:

I concur 100%

Quote:

Now finally, and importantly, I do not think that there were major spikes in load during any of the recent spectacular wins. I am not saying your achievements are poor at all. I myself tried MDDC and failed utterly a number of times.

I concur again

Ah hell, I concur with the whole post :-)

what ply does the 15s version usually reach ?

Title: Re: BotBash Challenge of the Month
Post by 99of9 on Jun 27^th, 2005, 8:57pm

on 06/27/05 at 11:55:30, Arimanator wrote:

In fact it it were a good chess game that's what it would do. Which is probably a testament to the distance still separating the best Arimaa bots and their chess counterparts. I mean MDD is worth about a queen in chess. Even Kasparov wouldn't try a strong chess game with a queen handicap, he'd be dead before he hit the ground. ;D

I personally think that MDD is worth about QBB. Your achievement is bigger than you think! I venture to say that "Even Fritz wouldn't try a strong arimaa opponent with a camel handicap." (at least... he's never offered that to me ;-))

Quote:

One principle in chess that would be good to follow here is when you have a material advantage, force the exchanges of pieces so that the fewer there are pieces left the greater your advantage.

Apparently David initially implemented this in bomb itself, but found, as Jeff suggests, that sometimes the board became so sparse on both sides, that an enemy rabbit was able to slip through.

My point of view in human games (if you are ahead on material) is:
1) You are more powerful, therefore you should be able to win further advantage, not just an equal exchange.
2) If an equal exchange is offered, accept it if it gives you some other positional or tempo advantage, but don't accept it otherwise (unless of course choosing not to accept costs you position or tempo).
3) This is sometimes modified by an evaluation of who's rabbits are further toward the goal, or who has a greater quantity of pieces. Both of these obviously favour exchanges.

Title: Re: BotBash Challenge of the Month
Post by PMertens on Jun 27^th, 2005, 11:47pm

MDDR in Game 16438 ...

Title: Re: BotBash Challenge of the Month
Post by 99of9 on Jun 29^th, 2005, 3:40am

on 06/27/05 at 13:48:04, Arimanator wrote:

I think MDDR is the limit of the feasible with Bombblitz and is probably beyond feasible for slower bombs.

... Horses ...

As for the second cat I think it's crucial, if Bomb sees that there is a horde of rabbits on one side and only one Horse to defend it, sooner or later he'll be rabbit pulling like crazy, no way you can last a 100 moves without that happening.

As you are so fond of sayings, I think this one is apt "Theres more than one way to skin a cat.". (although admittedly that was P2 rather than blitz)

Quote:

In other words, next weekend, I'll go for it. ;)

Excellent - I think we will find that the handicap keeps rising for a while yet, my initial camel estimate was way out.

Title: Re: BotBash Challenge of the Month
Post by 99of9 on Jun 29^th, 2005, 6:10am

on 06/28/05 at 01:09:51, Arimanator wrote:

As for 99 I would love to see him prove me wrong by making the (in my view) miraculous MDDC challenge (By goal of course not repeat or time) but if he did I would suspect some divine intervention here. ;D

Really?

Quote:

Are you telling me you honestly believe that this kind of game proves anything?

What was you plan here exactly? I'll let myself tear apart and sooner or later there'll be this huge gap in his defense and I'll just punch a rabbit through, pleaaaaase !!!

et tu brute? lol I couldn't resist that one. Grin

Now maybe you could win that way without your elephant?

I suggest you try that next time.

As for my next attempt, forget it!

I think that any future attempt is by advance ridiculed by this stroke of luck turned into a feat by your reciprocal congratulatory fest.

Amusing.

Here's some brief answers to your questions:
Yes.
I rarely have a plan in a game of Arimaa.
Thanks for the suggestion about the elephant, but I don't think I'm good enough yet :-).

Title: Re: BotBash Challenge of the Month
Post by 99of9 on Jun 29^th, 2005, 6:54am

on 06/29/05 at 06:34:56, Arimanator wrote:

I think you should go into politics, you obviously have a knack for rhetoric.

Thanks, I'll keep that in mind.

Quote:

You had an opportunity that you exploited (which has some merit in it) but a few moves before you were obvioulsy losing, you could at least have the honesty to admit that.

Yes, I admit that. Here's some more rhetoric for you: I fully admit that I was obviously losing even after move 1 of the game! The first time I had a winning position was move 39w. :-)

Quote:

So using this game as opposed to my statement as some kind of scientific proof of it's falsity is at least manipulative and certainly not the kind of reaction I would have expected from someone like you.

I disagree. I consider that this kind of game is quite common in botbashing (see most of the other records on the botbash page). Therefore I presumed you included this kind of win in your analysis of what you thought was possible.

Title: Re: BotBash Challenge of the Month
Post by omar on Jul 2^nd, 2005, 12:51am

on 06/27/05 at 11:24:44, PMertens wrote:

I think a P3/P4 version of bomb might give us more insight - would that be possible, omar ?

I think P4 could only be done at the postal level. I'll try setting up Bomb at P3. But I think it will take so long to move that people won't want to play it.

Title: Re: BotBash Challenge of the Month
Post by PMertens on Jul 2^nd, 2005, 1:38am

P4 postal would certainly would not be a bot for daily gaming, but if we can play that one with M+ hanidcap than it answers the question about ply vs eval (at least in my humble opinion)

I doubt that anyone will code a P4+ bot with under 2 mins movetime anytime soon, so that could be considered the upper limit of bot ply for the near future.

It is just far to easy to trick a bot only P2 or less (in case of Blitz)

Title: Re: BotBash Challenge of the Month
Post by Fritzlein on Aug 4^th, 2005, 7:10pm

I think handicap immobilization would be interesting, and particularly handicap, no-capture immobilization. Why not throw in all the requirements at once to get an uber-challenge? I mean, it would be cool if someone could give Bomb a camel handicap and then capture most of its pieces and immobilize the few remaining ones, but even cooler if an inferior force could immobilize a superior force while it was still superior. But as usual, I'm just throwing out an idea that I probably won't attempt for myself, so anyone else is free to take it or leave it.

Title: Re: BotBash Challenge of the Month
Post by PMertens on Aug 5^th, 2005, 9:44pm

I believe Game #17624 meets the requirements ?

Title: Re: BotBash Challenge of the Month
Post by Fritzlein on Aug 5^th, 2005, 10:04pm

Indeed. It still blows my mind that this is possible at all.

Title: Re: BotBash Challenge of the Month
Post by Fritzlein on Aug 6^th, 2005, 9:32am

on 08/06/05 at 08:31:01, Arimanator wrote:

Game 17647 DOG handicap, I won in 40 moves.

Well anyway that's the record for handicap immobilization with capture.

Title: Re: BotBash Challenge of the Month
Post by PMertens on Aug 6^th, 2005, 6:51pm

dog+cat in 49 with silver ... (17666)

bomb was finished around move 20, but it took so long to park all those stupid pieces so I could finally go for his last little breathing space

Eeven more garbage-pieces are possible, but I still have no idea how a horse (or more) might be given ...

Would it be possible to rename bomb into turtle ?
That would describe the exploited behavior quite well ...

Title: Re: BotBash Challenge of the Month
Post by 99of9 on Aug 8^th, 2005, 1:22am

on 08/08/05 at 00:51:14, Arimanator wrote:

Friendly competition on my end (and may I presume on the other end too?) , I swear Omar.

Good to see the competition is friendly now. Maybe one day if it stays that way and you both agree you can actually start competing directly again.

Title: Re: BotBash Challenge of the Month
Post by PMertens on Aug 9^th, 2005, 12:01am

horse and rabbit handicap immobilization with no pieces killed in 32 moves with silver ...

At the moment I assume that 2 pieces might be maximum because anything more gives lots of trouble in covering the ground.

I am not sure wether or not horse and cat/dog is possible ...

How did we define 2 dogs compared to horse and rabbit ?
I would assume that horse and rabbit is worth more, but it should be defined somehow.

Title: Re: BotBash Challenge of the Month
Post by 99of9 on Aug 9^th, 2005, 12:20am

on 08/09/05 at 00:01:48, PMertens wrote:

How did we define 2 dogs compared to horse and rabbit ?
I would assume that horse and rabbit is worth more, but it should be defined somehow.

The 99-system puts them at exactly equal! :-)

This kind of issue is going to be difficult to sort out for certain. (Not just in this case - there could be many like it.) Maybe we should have a rule that one handicap record only overrides another if it is strictly superior (eg HD>DD), and otherwise both records are put up?

Title: Re: BotBash Challenge of the Month
Post by PMertens on Aug 9^th, 2005, 12:26am

the problem is that dog and cat could not be beaten by horse - but it is obvious, that a horse is much harder to do (not only in this special challenge)

Horse and rabbit is not even close to a camel on this challenge - even though a camel is no problem with the normal goal.

We might consider a case dependent evaluation since the 99-system obviously does not take special handicaps into account.

Title: Re: BotBash Challenge of the Month
Post by 99of9 on Aug 9^th, 2005, 2:43am

Don't worry, I'm not saying we should use the 99-system for immobilisation. What I'm saying is that unless we all agree on a very detailed metric, the best we can do is use formal dominance only (and therefore multiple records at a time). For example if one person gets MC, and the other gets HDDRR ... all we can do is list both, right? Or do you think it's self-evidently obvious which of these is harder?

I understand that going with this would mean that HR would be listed alongside M... but I unless someone has a very detailed proposal (that we agree on), I don't think we can get around this.

Title: Re: BotBash Challenge of the Month
Post by PMertens on Aug 9^th, 2005, 3:11am

Actually the problem is not immobilization, but the "no-capture"

If I am allowed to immobilize whenever I want I can give a far greater handicap.

Anyway: for the moment I would accept tobys proposal unless both participants agree on what was harder under the given circumstances.

In case of the current challenge we might agree that a horse is far more than 2 dogs but 2 dogs and a (rabbit/cat) might be interesting though ... and maybe even harder than horse + (rabbit/cat/dog)
given the problem to cover your ground

With silver anything beyond 2 piece handicap becomes even more challenging since you are one move behind and its far harder to get bomb to turtle without triggering an phant-attack ...

Title: Re: BotBash Challenge of the Month
Post by Fritzlein on Aug 9^th, 2005, 3:41pm

on 08/09/05 at 06:43:32, Arimanator wrote:

How do you reconcile the following order ( rough estimation) :

R<C<D<R+R<C+R<D+R<3R<2C<2D(some may disagree here but I believe it's that order)<C+2R<D+2R<H<H+R<2C+R<2D+R (then it becomes a little too controversial)

with anything linear?

You are quite right, that can't be reconciled with anything linear. The proof is that CC < H while CCR > HR, i.e. adding a rabbit to both sides of an inequality reverses the inequality.

If you would only increase the value of H+R, changing it to
R<C<D<R+R<C+R<D+R<3R<2C<2D<C+2R<D+2R<H<2C+R<2D+R<H+R
then you could satisfy it with
R=1
C=1.6
D=1.7
H=3.8

(Well, that's what you get for asking a math question. :-))

Title: Re: BotBash Challenge of the Month
Post by PMertens on Aug 9^th, 2005, 4:00pm

could you include the difference between silver and gold into that formula ?

for silver a higher number of sacrifices is more troubling than higher pieces in comparison with gold - due to the fact that bomb might be triggered differently.

CCR with gold might indeed be easier than HR while I bet with silver its the other way round.

Due to the complexity the problem I would still prefer toby's solution until we have statistical proof ;-)

Title: Re: BotBash Challenge of the Month
Post by 99of9 on Aug 10^th, 2005, 2:56am

[Edit: the links and dates from my quotes are wrong because I haven't figured out how to quote more than one person in the same message.]

on 08/09/05 at 15:01:41, 99of9 wrote:

Maybe we should have a rule that one handicap record only overrides another if it is strictly superior (eg HD>DD), and otherwise both records are put up?

on 08/09/05 at 17:43:41, 99of9 wrote:

What I'm saying is that unless we all agree on a very detailed metric, the best we can do is use formal dominance only (and therefore multiple records at a time).

on 08/10/05 at 00:35:41, Arimanator wrote:

I am starting to think that maybe we should list ALL Handicap records that don't satisfy the following condition :

If there is an obvious Handicap that outmatch them : {snip} otherwise as I said before CC and H is debatable as well as HH and M {...}

I for one would say that when in doubt don't commit the injustice of dismissing a record that may be more valuable than the one you list and let the spectator decide which one s/he likes more. ;)

And now thank you for making MY point!

Title: Re: BotBash Challenge of the Month
Post by 99of9 on Aug 10^th, 2005, 3:32am

on 08/10/05 at 03:01:49, Arimanator wrote:

Forgive me for being so cryptic but acting otherwise would violate a word to which I have the weakness of attaching some value.

I hope you can appreciate that starting or ( at least trying to start ) a debate on something someone swore ( to Omar) no to talk about (only days before) is unfair to say the least and I won't say what it is to say the worst.

Well this time it was definitely too cryptic for me. But I do understand that you would otherwise have talked about something that you're not allowed to. So I'm happy not to continue along these lines.

Title: Re: BotBash Challenge of the Month
Post by PMertens on Aug 10^th, 2005, 2:40pm

horse+cat ímmonokill in 28 with silver ... (17792)
horse+dog ímmonokill in 32 with silver ... (17795)

Title: Re: BotBash Challenge of the Month
Post by Fritzlein on Aug 14^th, 2005, 1:24pm

on 08/14/05 at 11:31:51, Arimanator wrote:

I believe there should be a challenge for fastest extermination, and a bonus for fastest extermination without losing a piece.

There is something suprising about the existence of a category for winning without taking a piece, whereas there is no category for winning without losing a piece. Perhaps not losing any pieces is considered a minor restriction, since one doesn't want to lose pieces anyway, whereas winning without taking any pieces can pose more interesting difficulties.

Title: Re: BotBash Challenge of the Month
Post by Fritzlein on Aug 15^th, 2005, 8:29am

Ah, sorry, I didn't quite understand the challenge you were posing with extermination. I though you simply meant a brutal defeat, as opposed to presicely capturing all opposing pieces. :-) That is an unique challenge, as you say, since you must avoid prematurely winning by immobilization, yet I imagine you must nearly immobilize the bot to bring about elephant suicide. Against bot_Loc, which times out when it has no elephant move, it must be extra difficult to avoid winning in some other way.

Title: Re: BotBash Challenge of the Month
Post by Fritzlein on Aug 18^th, 2005, 5:55am

on 08/17/05 at 15:09:00, Arimanator wrote:

[...] maybe I'll give another shot at SILVER top handicap Bomb p2 goal since it's still untapped. ;)

I'm actually curious to see the rest of the material handicap records get filled in. In particular, I wonder whether Bomb's defensive style makes it more vulnerable to large handicaps than Clueless is, or whether Clueless also has huge issues we haven't even exploited yet.

Of course, as they say, beggars can't be choosers. If I were really interested, I would try for the Clueless handicap record myself, wouldn't I? ;-)

Title: Re: BotBash Challenge of the Month
Post by Arimanator on Aug 19^th, 2005, 3:40am

Game 18122 goal against P2 with silver in 17 moves with my very own method I even sacrificed my camel for mometum.

Nothing borrowed here just exploiting Bomb's greediness.

Title: Re: BotBash Challenge of the Month
Post by Arimanator on Aug 20^th, 2005, 3:42am

Win by goal with gold against Bomb P2 in 17 moves game 18184.

I couldn't walk with one leg shorter than the other, could I? ;)

Title: Re: BotBash Challenge of the Month
Post by Arimanator on Aug 22^nd, 2005, 4:17pm

Make that 13 moves with silver against Bomb2005CC game 18292.

Title: Re: BotBash Challenge of the Month
Post by omar on Aug 22^nd, 2005, 9:53pm

I never imagined we could exploit the bots greedyness this much.

Try this out on bot_Arimaanator and see how well it works when the initial setup is different.

Title: Re: BotBash Challenge of the Month
Post by Arimanator on Aug 23^rd, 2005, 4:16am

on 08/22/05 at 21:53:24, omar wrote:

...I never imagined we could exploit the bots greedyness this much...

I must say the trick was to curb my own greediness and not trying to get back the pieces that Bomb took from me.

Title: Re: BotBash Challenge of the Month
Post by 99of9 on Aug 30^th, 2005, 3:02am

I don't know of any quicker. To be honest, I think somebody will eventually manage much quicker since Gnobby is fully susceptible to horizon effects.

I agree with you that Immonokill is a whole different kettle of fish though.

Title: Re: BotBash Challenge of the Month
Post by 99of9 on Aug 30^th, 2005, 6:53am

on 08/30/05 at 04:42:52, Arimanator wrote:

Always nice to meet a fan

I'm not saying it won't be you, I'm just saying that Gnobby is susceptible.

Title: Re: BotBash Challenge of the Month
Post by PMertens on Aug 30^th, 2005, 8:47pm

Immonokill @ 18992 in 28 ...

Title: Re: BotBash Challenge of the Month
Post by 99of9 on Aug 30^th, 2005, 9:01pm

on 08/30/05 at 03:02:34, 99of9 wrote:

I agree with you that Immonokill is a whole different kettle of fish though.

Looks like I was wrong about this one. Well done Paul. I'm still surprised by Gnobby's final move actually, because I thought the elephant-blockade code would prevent moves like that. When I get back to coding I'll have to check out that kind of position.

Title: Re: BotBash Challenge of the Month
Post by PMertens on Aug 30^th, 2005, 10:20pm

I should mention he did that move after 2 seconds or so

Title: Re: BotBash Challenge of the Month
Post by Fritzlein on Sep 10^th, 2005, 4:40am

If a non bot-basher is premitted to propose a challenge, I am curious to know what is the toughest bot that can be beaten after giving an elephant handicap. 99of9 beat ShallowBlue with only a horse and two rabbits, and I think he also once gave an elephant handicap to Arimaalon, although I can't find the game.

Are there any other bots than can be beaten after giving an elephant handicap? Arimaazilla? LocP1? GnobotP1?

Under the original 99of9 scoring system, E = MH = MDC, and BombP2 has been beaten at MDDC, but I think nowadays 99of9 would agree with me that in the opening E > MH in actuality. I sort of doubt Bomb can be beaten with an elephant handicap, because previous bashing records exploited a camel shyness that would be irrelevant with one's own elephant gone. Bomb is pretty good at the material game, and pretty good at goal defense, so I'm not sure what weaknesses there are to exploit with no elephant.

Gnobot, might be more feasible, because of its poor goal defense. On the other hand, Gnobot's grasp of which friendly piece to put near which enemy piece is strong, so it might be able to neutralize your whole elephantless army.

Probably Loc and Arimaazilla are the best targets for this challenge. On the other hand, I'm not going to say what is impossible this time, after all the times I've been proven wrong in the past. Since I'm not asking, "What is the largest handicap that can be given to bot X?" but rather, "What is the toughest bot that can be given an elephant handicap?" you bashers probably won't rest until you prove that every bot out there can be beaten without the elephant. :-)

Title: Re: BotBash Challenge of the Month
Post by Arimanator on Sep 10^th, 2005, 6:04am

on 09/10/05 at 04:40:57, Fritzlein wrote:

...Are there any other bots than can be beaten after giving an elephant handicap?...

If you remember Fritz, Game 14706 I won against LOCblitz at move 35 after blundering my elephant at move 22 and I wasn't even trying very hard.

I believe that LOC definitely can be beaten with an Elephantless army and probably in all instances. (with a little work maybe).

Title: Re: BotBash Challenge of the Month
Post by Fritzlein on Sep 10^th, 2005, 6:34am

on 09/10/05 at 06:04:45, Arimanator wrote:

If you remember Fritz, Game 14706 I won against LOCblitz at move 35 after blundering my elephant at move 22 and I wasn't even trying very hard.

Good point; I had forgotten that game. However, you did already have CCRRR compensation for the elephant in that game, which loosened up the goal line for you somewhat. Still, it was an impressive comeback (especially at blitz speed) and it's good evidence that Loc is a bot that one might be able to beat when giving an elephant for nothing.

Title: Re: BotBash Challenge of the Month
Post by Arimanator on Sep 15^th, 2005, 11:23am

I guess Yoddle the turtle is still in charge in lieu of Bot Arimaanator, Which means I can't claim the first immonokill ever against that Bot and in 24 moves to boot (my personal record against any instance of Bomb). ;)

Title: Re: BotBash Challenge of the Month
Post by Arimanator on Sep 18^th, 2005, 5:27am

Got clueless fast in 14 moves using the same mutual massacre technique that functionned against Bomb. Clueless is a bit less gullible than Bomb tho you have to use a greater concentration of forbidden candies to keep its elephant far enough away for the plan to succeed.

Title: Re: BotBash Challenge of the Month
Post by mistre on Apr 8^th, 2008, 11:22am

on 09/10/05 at 04:40:57, Fritzlein wrote:

I am surprised that no one answered the call to this challenge until now. I have beaten ArimaaScoreP1, ArimaaScoreP2, Arimalon, AamiraP1, Loc2005P1, and Loc2006P1 without an E. It had already been proven that Shallowblue can be beat without an E. Loc2006P1 was the most challenging. ;D

Arimazilla I have tried and failed, but I think it could be possible. Gnobby uses his Elephant alot and does not advance his rabbits, so I think it might be impossible to beat even Gnobot2005P1 without an E, but feel free to prove me wrong.

Title: Re: BotBash Challenge of the Month
Post by aaaa on Apr 8^th, 2008, 12:02pm

I haven't seen this discussed before, but how much do people reckon an elephant is worth in competitive play? Is it more or less than a camel and two horses? And how would such a game actually play out?

Title: Re: BotBash Challenge of the Month
Post by Fritzlein on Apr 8^th, 2008, 1:53pm

on 04/08/08 at 12:02:19, aaaa wrote:

It's hard to answer this with a complete lack of evidence. The total of my experience is that I gave an elephant handicap to a kid I had just taught the rules to, and then again to his little brother. I haven't done any no-elephant bot-bashing.

I think your question is at the right level: E=MHH is as good a guess as any until we try a few.

Title: Re: BotBash Challenge of the Month4
Post by Janzert on Apr 8^th, 2008, 6:56pm

This got me wondering what the current material evaluators thought:

Score given with gold elephant and the shown silver material taken. Positive scores favor gold, negative silver.

FAME	mhh = 2.3	mhd = -0.15	mhcr = 0
DAPE	mhh = 3.3	mhdc = 0.74	mhdr = -0.73
DAPE(eo)¹	mhh = 4.9	mc = 0.23	mr = -0.53

So all of them think MHH is too much, and DAPE(eo) would be happy to trade E for MC!

Janzert

¹ DAPE empirically optimized, DAPE with optimal constants as computed from game data by IdahoEv.

Title: Re: BotBash Challenge of the Month
Post by mistre on Apr 8^th, 2008, 7:12pm

This is just a hunch, but I think MHH will prove more valuable than E. I think equality might be in the MHD or MHCR range. It would be fun trying it out!

Title: Re: BotBash Challenge of the Month
Post by lightvector on Apr 8^th, 2008, 7:17pm

For the record, sharp currently thinks that E is worth about MHC ;D

But this will probably change soon as I completely overhaul the evaluation.

Title: Re: BotBash Challenge of the Month
Post by 99of9 on Apr 8^th, 2008, 8:02pm

Someone should play a few E vs MHH handicap games sometime, I think that would be a very interesting match.

Title: Re: BotBash Challenge of the Month
Post by Arimabuff on Apr 8^th, 2008, 11:26pm

on 04/08/08 at 20:02:45, 99of9 wrote:

Someone should play a few E vs MHH handicap games sometime, I think that would be a very interesting match.

You'd have to play against someone of your strength though for them to be probative.

Title: Re: BotBash Challenge of the Month
Post by 99of9 on Apr 9^th, 2008, 12:06am

on 04/08/08 at 23:26:43, Arimabuff wrote:

You'd have to play against someone of your strength though for them to be probative.

Is that a challenge? ;-)