Arimaa Forum - Print Page


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

        Arimaa >> Events >> 2008 Postal Tournament
        
(Message started by: Fritzlein on Apr 13^th, 2008, 8:31am)

Title: 2008 Postal Tournament
Post by Fritzlein on Apr 13^th, 2008, 8:31am

And they're off! I'm so excited. I can't wait to see if my rabbit columns can hold up under deeper analysis, or if they were just a sort of trick that only works in live games.

The number of players is down this year, but the number of games is still high since most of the folks that did sign up are in for a lot of games.

	2005	2006	2007	2008
Players	15	19	20	15
Games	75	95	85	84

This year we will get to see the first ever BvB postal as sharp takes on OpFor. That should be a very intriguing contest, particularly if both developers are still working on their bots as the match progresses.

Title: Re: 2008 Postal Tournament
Post by chessandgo on Apr 13^th, 2008, 10:03am

Good games and a lot of fun to everyone !

Title: Re: 2008 Postal Tournament
Post by nbarriga on Apr 13^th, 2008, 11:00am

Does anyone knows the trend for the ratings of the players in the postal tournament?

It seems that this year it's gonna be far tougher for us low ranking people.

Title: Re: 2008 Postal Tournament
Post by Fritzlein on Apr 13^th, 2008, 1:04pm

on 04/13/08 at 11:00:51, nbarriga wrote:

It seems that this year it's gonna be far tougher for us low ranking people.

Yes, this year is going to be more brutal for the lower-ranked players. If 400 rating points makes a mismatch, then a full quarter of the games will be mismatches.

	2005	2006	2007	2008
Players	15	19	20	15
Games	75	95	85	84
Mismatches	9	5	3	21

This year for some reason there weren't a lot of newer players signing up as there have been in years past, and a few of the old faithful players (99of9, jdb, naveed) didn't return. If there were more players, then the games everyone requested would be used up on pairings closer to their own rating, and the percentage of mismatches would be lower.

I guess we just didn't promote the tournament enough this year. Or maybe the longer format of the World Championship wore some people out, and if we had waited longer to start the Postal Tournament, there would be more game-hungry people around.

Title: Re: 2008 Postal Tournament
Post by Fritzlein on Apr 13^th, 2008, 7:34pm

I'm curious what OpFor's thinking policy is going to be for this tournament. I notice OpFor took nine hours to make its last move against me; it can't be thinking too terribly long on each game if it got back around to me so quickly. Janzert, do you mind sharing your time management strategy? Of course I understand if that's a trade secret.

Title: Re: 2008 Postal Tournament
Post by Janzert on Apr 14^th, 2008, 9:49am

I'm starting out with opfor thinking a minimum of 2 hours and maximum of 4 hours for a move. I may end up deciding I need my computer for other things and cut that back though.

One thing yesterday was that I ended up restarting it a few times as I found a few bugs in the manager and then later realised a search "improvement" I made on saturday introduced a bug in the search. So a few games actually got multiple moves before the bot cycled through the full list of games.

Janzert

Title: Re: 2008 Postal Tournament
Post by Fritzlein on Apr 14^th, 2008, 3:37pm

on 04/14/08 at 09:49:39, Janzert wrote:

[...]a search "improvement" I made on saturday introduced a bug in the search.

It's wonderful to hear that you are continuing active development! ;) Seriously, it's a nice gift for you to enter OpFor in the tournament, and I hope the games give you food for thought and guidance for taking OpFor to the next level.

Title: Re: 2008 Postal Tournament
Post by RonWeasley on Apr 15^th, 2008, 12:07pm

Game 88295:

They're mistre and The_Jeh.
Yes mistre and The_Jeh.
One is a genius,
The other one ... meh.

They came to play the game.
They hope it won't be lame.
They're hist'ry, they're mistre and The_Jeh (jeh jeh jeh)

They're playing really fast.
They're having so much fun.
If the future's like the past,
They'll be the first ones done.

They're mistre and The_Jeh.
Yes mistre and The_Jeh.
They're simpletons, oh yeah.
They make you say, "Bleh!"

To prove they're dumb as logs,
They'll blunder both their dogs.
They're hist'ry, they're mistre and The_Jeh (jeh jeh jeh)
NARF!

Title: Re: 2008 Postal Tournament
Post by Fritzlein on Apr 15^th, 2008, 12:47pm

Sung to the tune of?

Title: Re: 2008 Postal Tournament
Post by RonWeasley on Apr 15^th, 2008, 1:18pm

on 04/15/08 at 12:47:35, Fritzlein wrote:

Sung to the tune of?

http://www.youtube.com/watch?v=iJPFSNu_QNs

Title: Re: 2008 Postal Tournament
Post by mistre on Apr 15^th, 2008, 1:46pm

My game against Arif_Syed has actually caught up with that one, you might need a new song ;D

Title: Re: 2008 Postal Tournament
Post by mistre on Apr 17^th, 2008, 9:12am

Opfor is moving in my game, but the interface always shows it as my turn. Can this be fixed?

Sharp on the other hand, hasn't done anything.

Title: Re: 2008 Postal Tournament
Post by Fritzlein on Apr 17^th, 2008, 10:15am

on 04/17/08 at 09:12:18, mistre wrote:

Opfor is moving in my game, but the interface always shows it as my turn. Can this be fixed?

That's weird, exactly the same thing is happening to me. The e-mails are correctly informing me when it is my turn to move, but the game room shows that it is always my move. The problem only occurs with OpFor, and no other opponent. Might it have something to do with Janzert's customized bot interface?

Quote:

Sharp on the other hand, hasn't done anything.

We need to send lightvector an e-mail.

But everyone else has started, and Arif_Syed vs. The_Jeh is up to move 20! I'm still trying to figure out if chessandgo's move 3 inevitably puts me at disadvantage. I would hate for my new setup with a wing camel as Silver to be so easily refuted.

Title: Re: 2008 Postal Tournament
Post by Janzert on Apr 17^th, 2008, 12:38pm

on 04/17/08 at 10:15:14, Fritzlein wrote:

Hmm, that is wierd. Given that it has only happened with OpFor I would guess that it's being caused by the custom interface as well, but I'm at a loss as to what it could be.

Janzert

p.s. I increased the logging level to record all network traffic, so after the next move I'll take a close look and see if I can spot anything that looks wrong.

p.p.s Does anyone remember if the games updated correctly for zombie last year?

Title: Re: 2008 Postal Tournament
Post by ChrisB on Apr 17^th, 2008, 2:08pm

I sometimes also don't get gameroom updates of the correct move number, but not just from my game with OpFor, but in other games too. For example, currently the gameroom says it's my move #2 in my postal game with nbarriga, but I have already made that move (over 19 hours ago). So the problem may be more global than Janzert's interface for OpFor.

Title: Re: 2008 Postal Tournament
Post by mistre on Apr 21^st, 2008, 11:31am

When can we expect to see the progress/results of games on the postal mixer page to show up?

Title: Re: 2008 Postal Tournament
Post by Fritzlein on Apr 21^st, 2008, 2:17pm

Especially given that a game has finished. The scores are not all zero!

Title: Re: 2008 Postal Tournament
Post by mistre on Apr 21^st, 2008, 4:13pm

Two games are finished actually. Mine vs Arif_Syed is also done.

Title: Re: 2008 Postal Tournament
Post by The_Jeh on Apr 22^nd, 2008, 1:38pm

When the scoreboard starts working, what will the points column indicate?

[edit]

Oh,

A player receives 60 points for winning.
Both players gets 60 points if the game is a draw.
A player loses 60 points if they forfeit, resign or lose on time.
If a player loses any other way they receive points equal to the number of moves in the game up to a limit of 60 points.

Title: Re: 2008 Postal Tournament
Post by omar on Apr 22^nd, 2008, 10:05pm

on 04/21/08 at 11:31:18, mistre wrote:

When can we expect to see the progress/results of games on the postal mixer page to show up?

Wow, didn't realize some games were finishing already. I've setup the cron job now to update the results once an hour.

Title: Re: 2008 Postal Tournament
Post by Fritzlein on Apr 22^nd, 2008, 10:21pm

Thanks, Omar!

Title: Re: 2008 Postal Tournament
Post by mistre on Apr 24^th, 2008, 8:27pm

on 04/15/08 at 12:07:20, RonWeasley wrote:

For anyone that is not watching this epic battle, we are on Move 52 and The_Jeh is going for the upset!

Title: Re: 2008 Postal Tournament
Post by Fritzlein on Apr 25^th, 2008, 10:28pm

on 04/24/08 at 20:27:00, mistre wrote:

For anyone that is not watching this epic battle, we are on Move 52 and The_Jeh is going for the upset!

And The_Jeh scores a second point while my average game is still on move 7. Well done, John!

Title: Re: 2008 Postal Tournament
Post by Janzert on Apr 25^th, 2008, 11:25pm

Has anyone had any contact with lightvector about bot_sharp? I'm still hopeful to see it play.

Janzert

Title: Re: 2008 Postal Tournament
Post by lightvector on Apr 26^th, 2008, 12:24am

Sorry everyone, I've been quite busy over the past month, and in the bits of time available, I have only just gotten sharp into a state for playing.

Anyways, this included a much-needed overhaul to the evaluation. The CC version was laughably crude: mostly just raw material value + piece-square tables + have the strongest piece adjacent to a trap, so I was quite surprised when it actually won some games. The new eval is quite untuned, but probably much better (and hopefully won't as easily give away camels and frames every game, among other things!). Also, I hastily made and tested a few bugfixes - no more missing easy goals or 1 move captures.

So I believe sharp will finally be coming online this weekend.

Incidentally, how do you get a bot to connect to specific games to make a move?

Title: Re: 2008 Postal Tournament
Post by Fritzlein on Apr 26^th, 2008, 10:33am

Lightvector, I'm excited to hear that not only will sharp play in the Postal Mixer after all, but also that you are still actively developing! It's a bit scary to think that you took second place with a bot you consider brain-dead. I'm eager to see how the improved evaluation plays out in the tournament. Thanks for keeping at it.

Title: Re: 2008 Postal Tournament
Post by Janzert on Apr 26^th, 2008, 11:52pm

Great to hear that it will be playing, and that you're still actively working on it.

on 04/26/08 at 00:24:35, lightvector wrote:

Incidentally, how do you get a bot to connect to specific games to make a move?

You can call the bot interface with 'bot move <game number or opponent name>' to play one move in a specific game. Otherwise take a look at the botman script to have it basically do the same for all the postal games your bot has.

Janzert

Title: Re: 2008 Postal Tournament
Post by Fritzlein on May 5^th, 2008, 5:25am

The move counter has gone berserk, at least for the Gold side of all of my postal games. My game with OpFor is showing that it is move 191. (I guess the formula is the real number of moves times ten plus one...) Also I am apparently not getting e-mail notifications on all of my games. Everyone please be sure to check the game room instead of relying on mail.

Title: Re: 2008 Postal Tournament
Post by omar on May 5^th, 2008, 7:53am

The counter has been fixed. Instead of adding the number 1 it started concatenating the string '1'.

Title: Re: 2008 Postal Tournament
Post by Fritzlein on May 5^th, 2008, 8:47am

Ah, thanks for fixing this, Omar. You see that Yahoo hasn't made a programmer out of me yet. I had to come up with the "times ten plus one" theory because it didn't even occur to me that "+1" could have another meaning. :)

Title: Re: 2008 Postal Tournament
Post by omar on May 5^th, 2008, 10:05am

You're certainly more of a mathematician than a programmer :-)

Title: Re: 2008 Postal Tournament
Post by RonWeasley on May 14^th, 2008, 8:34am

With most games coming up on 20 moves, I'm really impressed with how well bot_OpFor is playing. I expected it to be non-competitive. Instead it's winning against some good players. Maybe the extra time makes a difference for it. Early congratulations to Janzert.

Title: Re: 2008 Postal Tournament
Post by Janzert on May 14^th, 2008, 11:21am

Thanks, although I've been quite surprised by how well it's doing also.

Something I would interested in hearing as people finish up the games with OpFor is what rating player does it feel like you're playing, i.e. if the opponent was unknown does it feel like you are playing an 1800 rated player or 1500 or what approximately? Similarly how "weird" does its play seem, does it feel basically like playing a similarly rated human, does it make an occasional weird move or does it just have a plain weird or maybe bot like play style? I'd be really interested in hearing peoples thoughts.

Janzert

Title: Re: 2008 Postal Tournament
Post by Fritzlein on May 14^th, 2008, 11:43am

The game room ratings of OpFor's opponents in the Postal Mixer were

1601
1683
1725
1803
1890
1922
2090
2090
2109
2177
2177
2297
2297
2304
2304
2340
2491

We can calculate OpFor's performance rating based on the number of games it wins against this opposition:

Wins / Performance
1 1437
2 1592
3 1699
4 1787
5 1863
6 1933
7 1998
8 2060

Recall that Zombie had a performance rating of 1409 in the 2007 Postal Tournament (1-14) and Bomb had a performance rating of 1716 in the 2005 Postal Tournament (4-6). OpFor only needs to win four this year (4-13) to set a record postal performance for a bot.

Of course this could touch off another thread about rating inflation, but please recall that at the time the 2005 Postal tournament started, we didn't have Bomb-bashing down to a science, and the human participants weren't all veterans of the bot ladder, because the bot ladder didn't exist yet. There may be some human rating inflation, but that isn't the only thing going on. Blue22's rating now is about what mine was three years ago, and I'll wager he is as strong a player now as I was then, given how much all of us have learned in the mean time.

OpFor's style is IMHO more suited to postal play than Zombie's wildness and preference for small pieces. If OpFor outdoes Bomb as well it may be due to having a more strategic style, although Bomb's strategy is pretty good.

Title: Re: 2008 Postal Tournament
Post by Fritzlein on May 14^th, 2008, 12:04pm

on 05/14/08 at 11:21:03, Janzert wrote:

Something I would interested in hearing as people finish up the games with OpFor is what rating player does it feel like you're playing

In my completed game against OpFor, I felt much more like I was playing against an inexperienced human than against a bot. OpFor seemed to be playing with strategy by taking my attacking horse hostage and opposing my camel on the other wing with its own camel, but it was messing up tactically by letting me get its two cats offside and win them both for a rabbit. Decent strategy with bad tactics is very human-like. At the end OpFor made a colossal strategic mistake in underestimating my goal attack, and a tactical mistake in overlooking my goal in two. The latter mistake seems very un-computer like, but the former mistake could have been either a computer or an inexperienced human.

In my ongoing game against OpFor, I was surprised how strongly it resisted my EH attack. It stopped my camel on the opposite wing, and then activated its horse on move 7b, a strategically strong move that was an unwelcome surprise from a bot. I was about to post my compliments at OpFor's excellent strategy when it played a horrendous move 14b. Using only two steps is bizarre in itself (can't OpFor think of any active plan to improve its position with the other two steps?), but what's worse the two steps it did play invited an elephant smother that it didn't resist on the following moves either. OpFor just sat there and let the wave roll down. I didn't begin the game intending to win with a boring formula tested many times against Bomb, but ultimately OpFor begged me to do it. This was not a human loss: even an inexperienced human will see that I am up to something and try to stop it, or if he isn't paying attention to what I am doing, he will at least try to do something himself. Humans naturally operate by the maxim that it is better to have a bad plan than no plan at all.

Title: Re: 2008 Postal Tournament
Post by mistre on May 14^th, 2008, 2:54pm

I don't expect to lose against Opfor, but it is putting up quite a fight in my postal game. After a very strange opening that resulted in its Elephant being stationed behind his own trap and a relatively easy cat capture, We have since traded a rabbit. While I am ahead materially, Opfor has got me into several hostage situations, holding my camel with his E on the side of the board, my H with his M near a trap, and my D with his H.

While Opfor will make a puzzling move from time to time, I think its relative strength rivals Bomb (I said this before). Also, it is less predictable than Bomb (probably because no one has played it enough to notice any patterns). So a rating of 1850-1900 seems about right.

Title: Re: 2008 Postal Tournament
Post by Fritzlein on May 14^th, 2008, 4:07pm

on 05/14/08 at 14:54:55, mistre wrote:

I don't expect to lose against Opfor, but it is putting up quite a fight in my postal game.

That's quite an interesting position you've got against OpFor, mistre. OpFor seems to have good practical chances despite your material advantage. OpFor has a couple more games that are unclear like yours, and also has a clear material advantage in four games, so at the moment I would project OpFor to win five or six. Thanks for pointing out OpFor's success so far, Ron. I wouldn't have noticed otherwise.

It is quite remarkable, because bots are supposed to get weaker relative to humans at postal speeds. Even if OpFor is 1900 strength at live games, it should only be 1750 or so against humans who have lots of time to plan and avoid blunders. Janzert, if I didn't know that you hadn't altered OpFor between the Computer Championship and the Postal Mixer, I would insist that you must have souped it up somehow.

Looking at the individual games in which OpFor is doing well, I can't quite explain it. It does seem that OpFor has a knack for complicating the position when it is in trouble, and (just as importantly) for doing well in messy positions. The general affinity of OpFor's elephant for the opposing camel stands it in good stead when pieces are flying all over the place. Yet, as you point out, mistre, it makes some moves that are quirky and others that are just bad. My impression from having looked at games besides my own is that OpFor is perhaps like an early naveed, capable of a brilliant game next to a pathetic one, and sometimes both at once.

I wonder if OpFor isn't benefiting from a bit of complacency on the part of humanity. I recall that in 2005, we were still afraid of Bomb, and many of us spent more time thinking on our games against Bomb than against each other so as not to be humbled by a loss. Also it was conventional wisdom to play especially defensively against bots to avoid their relative strength in open positions. Now we know that we can out-punch Bomb even in slugfests, so we don't take care. A postal loss to a bot would be even more of a blow than three years ago, but perhaps we think it too improbable to be worth worrying about until we get a wake up call over the board. How many people will want another postal game against OpFor now that they have had fair warning?

Title: Re: 2008 Postal Tournament
Post by mistre on May 14^th, 2008, 4:25pm

I had played Opfor twice when Janzert put it up in the gameroom and won both times, but neither game resulted in what I would call an easy win.

I have to admit that I was taking it for granted in Postal figuring it would be no problem and sure enough I captured a cat early on. But then it made some surprising moves that caught be completely off guard. I was not expecting it to give up the horse frame and threaten my horse in the opposite trap -I don't know of any other bot knowing when to give up a frame. Opfor seems to have a knack of playing well between its home traps - at least in this game.

What is even more crazy is that Opfor is tougher for me than Arimaa_Master who I beat for the very first time (and convincingly!) I got my revenge for my loss in our Championship match.

Blue22 is very close to beating me using a very effective swarm strategy despite giving up a camel hostage.

The rest of my games are still early and close.

Title: Re: 2008 Postal Tournament
Post by Fritzlein on May 14^th, 2008, 7:32pm

on 05/14/08 at 16:07:23, Fritzlein wrote:

A postal loss to a bot would be even more of a blow than three years ago, but perhaps we think it too improbable to be worth worrying about until we get a wake up call over the board.

How ironic that I wrote this just before bot_sharp hit me with a move I totally didn't see on my last turn, and which forces me into an unfavorable camel trade. I am going to have to start taking that game seriously now that I've had my own personal wake up call...

Title: Re: 2008 Postal Tournament
Post by arimaa_master on May 15^th, 2008, 2:44am

I must admit that Opfor is constantly doing surprising moves (but also it is true that I don´t pay enough attention to my postal games either).

Opfor is playing very good arimaa in both my games (despite horse loss at move 12 in one of them). So far it found always ways to complicate the battle. But I am quite sure I am going to win both games :).

Title: Re: 2008 Postal Tournament
Post by arimaa_master on May 15^th, 2008, 2:46am

on 05/14/08 at 16:25:57, mistre wrote:

What is even more crazy is that Opfor is tougher for me than Arimaa_Master who I beat for the very first time (and convincingly!) I got my revenge for my loss in our Championship match.

Yeah you smashed me there!

Title: Re: 2008 Postal Tournament
Post by omar on May 15^th, 2008, 8:26pm

I think OpFor is going to win the game with me. If so this will be the first postal I lose to a bot. It is playing surprisingly well. It manage to take my camel hostage and then really surprised me on 15w by advancing the camel (I would never expect that from Bomb). I thought this would be a chance for me to trade camels, but I overlooked that it could save it's camel and block me from saving mine. The game is starting to fall apart for me. OpFor feels like a human opponent to me; probably because I am not familiar with it's style and the surprising moves its been making. I would say it feels close to a 2000 rated player. Nice job Brian.

Title: Re: 2008 Postal Tournament
Post by Fritzlein on May 16^th, 2008, 10:04am

Openings this year:

Elephant File	2005	2006	2007	2008
d	98.1%	79.5%	85.3%	89.9%
c		13.7%	9.4%	9.5%
b	1.9%	6.3%	5.3%
a		0.5%		0.6%
Rabbits Forward	2005	2006	2007	2008
none	14.4%	22.6%	10.0%	6.0%
a	2.5%	15.3%	4.1%	0.6%
c	3.1%
ac				0.6%
ah	55.6%	56.3%	76.5%	59.5%
cf	23.1%			8.3%
ag		0.5%	1.2%
adh		2.6%	4.1%	10.1%
abgh	0.6%
acfh	0.6%			14.9%
adeh		2.6%	4.1%
Setup Balance	2005	2006	2007	2008
Symmetrical	56.9%	50.5%	54.1%	50.0%
Balanced	22.5%	30.0%	34.1%	42.3%
Unbalanced	20.6%	19.5%	11.8%	7.7%
Gold Move 2w	2005	2006	2007	2008
E up 4	68.8%	26.3%	25.9%	13.1%
E up 3 over 1	11.3%	11.6%	1.2%	3.6%
E up 3; X up 1	6.3%	41.1%	40.0%	45.2%
E up 2; X,Y up 1	3.8%	10.5%	23.5%	27.4%
E, X, Y, Z up 1	3.8%	3.2%		1.2%
Other	6.3%	7.4%	9.4%	9.5%

Last year I thought the 99of9 setup was gradually driving out all other options. I'm glad that hasn't happened yet. In fact, the "rabbits forward" statistic conceals some extra variation this year: many people, even when they put rabbits forward on a2 and h2, don't necessarily put pieces on d1 and e1. If I included all the variations of the back-row rabbits, I believe we would see that this year's setups are the least homogeneous ever.

Title: Re: 2008 Postal Tournament
Post by Fritzlein on May 16^th, 2008, 2:11pm

Janzert, when I compiled the opening statistics, I noticed that Opfor was the most experimental of any player. What is your setup algorithm?

Title: Re: 2008 Postal Tournament
Post by Janzert on May 16^th, 2008, 6:31pm

Thanks for the comments everyone. I really appreciate and find them all very interesting. Sorry it took so long to respond (left town on a business trip right after posting and ended up without internet access at the hotel).

I'm encouraged to hear that it sounds like there are still big gains to be made in tactics, those should be the easy part to fix.

on 05/14/08 at 16:07:23, Fritzlein wrote:

It is quite remarkable, because bots are supposed to get weaker relative to humans at postal speeds. Even if OpFor is 1900 strength at live games, it should only be 1750 or so against humans who have lots of time to plan and avoid blunders. [edit: moved intervening comments to seperately reply to below - Janzert]... I wonder if OpFor isn't benefiting from a bit of complacency on the part of humanity. I recall that in 2005, we were still afraid of Bomb, and many of us spent more time thinking on our games against Bomb than against each other so as not to be humbled by a loss. Also it was conventional wisdom to play especially defensively against bots to avoid their relative strength in open positions. Now we know that we can out-punch Bomb even in slugfests, so we don't take care. A postal loss to a bot would be even more of a blow than three years ago, but perhaps we think it too improbable to be worth worrying about until we get a wake up call over the board.

Yeah, I agree. I think humans should, and actually do, "scale" better with more time than OpFor. But OpFor has been putting a minimum of 2 hours 'thought' into every move. I know by the move times that at least many of the humans certainly aren't. One thing of note related to scaling and that I realized earlier this week, the postal tournament speeds OpFor is running at are approximately 6 doublings away from championship/challenge speeds (think of moore's law and time to end of the challenge).

Quote:

Looking at the individual games in which OpFor is doing well, I can't quite explain it. It does seem that OpFor has a knack for complicating the position when it is in trouble, and (just as importantly) for doing well in messy positions. The general affinity of OpFor's elephant for the opposing camel stands it in good stead when pieces are flying all over the place. Yet, as you point out, mistre, it makes some moves that are quirky and others that are just bad. My impression from having looked at games besides my own is that OpFor is perhaps like an early naveed, capable of a brilliant game next to a pathetic one, and sometimes both at once.

I wonder how much we as a community have come to equate bot like play with bomb's style. Especially after your further comment about sharp catching you by surprise as well. In fact initially in my post where I said "maybe bot like play style?" I had put "bomb like" then changed it after I realized "bot like" was what I really meant.

Quote:

How many people will want another postal game against OpFor now that they have had fair warning?

I'm really hoping people will.

Janzert

Title: Re: 2008 Postal Tournament
Post by Janzert on May 16^th, 2008, 6:51pm

on 05/16/08 at 14:11:47, Fritzlein wrote:

Janzert, when I compiled the opening statistics, I noticed that Opfor was the most experimental of any player. What is your setup algorithm?

First some motivation, ;) I wanted to give people variety in OpFor's openings so decided to add more piece setups. I also don't think at OpFor's current level of play that the opening setup matters much, at least as long as it's fairly reasonable.

OpFor randomly chooses between 3 setup styles, Omar, 99of9 or Fritzlein. It then places the major pieces (EMH) and rabbits in the usual positions making sure the elephant is on the same half as the opponent's unless that would be directly across from the oppoent's.¹ Then it randomly fills in the dogs and cats in the remaining four slots.

A small bit of self play seemed to show that this at least didn't drastically hurt its play. So I decided to throw it in to mix things up a bit as I was somewhat afraid that people might end up bored by there games with OpFor. I'm glad to see that my fear was completely unfounded and hope that the added setups still gave a little extra spice. :)

Janzert

¹ As can be seen in the game against sharp, this actually had a bug in it and would not switch when necessary.

Title: Re: 2008 Postal Tournament
Post by mistre on May 16^th, 2008, 7:10pm

on 05/16/08 at 18:51:30, Janzert wrote:

¹ As can be seen in the game against sharp, this actually had a bug in it and would not switch when necessary.

This bug occurred in my game vs Opfor as well. I switched my gold camel/E and it setup its E across from my E. This resulted in a strange opening game, but it only netted a cat for me.

Title: Re: 2008 Postal Tournament
Post by Janzert on May 16^th, 2008, 8:17pm

Oops, meant to reply to this part above as well instead of accidentally cutting it out.

on 05/14/08 at 16:07:23, Fritzlein wrote:

Janzert, if I didn't know that you hadn't altered OpFor between the Computer Championship and the Postal Mixer, I would insist that you must have souped it up somehow.

Hmm, I hope I haven't given the impression that OpFor in the postal mixer is unchanged from OpFor in the CC. Besides the opening setup work I have made a few small changes (there have been 20 commits to the VCS since the CC version). But by small here I'm definitely measuring by code size/time to write it. At this stage in OpFor I wouldn't be surprised if small changes could still cause large improvements in its play.

The one change that I would not be surprised if it caused a jump in play strength is this. Shortly after the CC ended, it finally sunk into my thick skull what null move pruning actually is and why it helps (i.e. it only took me about a decade or so for realization to strike). I simultaneously realised that what I was doing in OpFor, that I had thought was basically equivalent, was not doing any where near the same thing. This led to me finally adding null move pruning to OpFor and testing showed that it did seem to help although how much I have no idea. This ended up being only 17 lines of code.

In chess I believe null move pruning often adds a couple of hundred elo to an engine. In other games it has not been shown to be useful at all. I have a feeling arimaa will land more on the chess end of the spectrum.

Janzert

Title: Re: 2008 Postal Tournament
Post by Fritzlein on May 17^th, 2008, 8:10am

on 05/16/08 at 20:17:53, Janzert wrote:

Hmm, I hope I haven't given the impression that OpFor in the postal mixer is unchanged from OpFor in the CC.

Now that you mention it, you implied in an early post in this thread that you were still developing, and I forgot that bit of exciting news. My comment that I "know" you haven't been making changes was based on your commit log here (http://arimaa.janzert.com/opfor/commitlog.txt), which has a last update of March 1. Since I no longer "know" that OpFor is the same as the one that played in the Computer Championship, my grasping for explanations of Opfor's strong performance in the Postal Mixer seems less likely than the obvious explanation that you made it significantly stronger.

Quote:

This led to me finally adding null move pruning to OpFor and testing showed that it did seem to help although how much I have no idea. This ended up being only 17 lines of code.

In chess I believe null move pruning often adds a couple of hundred elo to an engine.

Very interesting! I would believe that null-move pruning adds several hundred Elo points of strength to an Arimaa engine, on the theory that the vast majority of moves actually weaken a good position. This is in contrast to Go where very few moves weaken a strong position, and most have some positive value.

Before your explanation I was assuming that OpFor's strong postal performance was a fluke, but now it seems to me there is a genuine possibility that OpFor is 1900 strength postally, and by extension around 2000 strength for 2-minute games and 2100 strength for 30-second games. In my mind the effectiveness of null-move pruning has more ramifications for the future of the Challenge than the number of doublings of thinking time, since I expect that ultimately software breakthroughs will be more important than hardware speedup.

Feng-Hsiung Hsu thinks recursive null-move pruning will be a big deal for Go, but I'm wondering whether Arimaa search even gets to a depth whether the recursion pays off. Is your implementation recursive?

Do we know whether Bomb has null-move pruning?

Title: Re: 2008 Postal Tournament
Post by 99of9 on May 18^th, 2008, 4:16pm

on 05/17/08 at 08:10:49, Fritzlein wrote:

Do we know whether Bomb has null-move pruning?

Yes, I'm pretty sure David has previously suggested that it was quite valuable to him.

Title: Re: 2008 Postal Tournament
Post by Janzert on May 18^th, 2008, 11:15pm

on 05/17/08 at 08:10:49, Fritzlein wrote:

Feng-Hsiung Hsu thinks recursive null-move pruning will be a big deal for Go, but I'm wondering whether Arimaa search even gets to a depth whether the recursion pays off. Is your implementation recursive?

No. I don't remember for sure if I experimented with a recursive version or not. At least for current interactive search depths I don't expect that it will make much difference since it's pretty rare to get beyond 3 ply (12 steps) even at postal speeds it seems to be very rare to break 4 ply (16 steps). I did try various adaptive reductions but none of them seemed to do as well as a simple 4 step reduction.

Quote:

Do we know whether Bomb has null-move pruning?

Yes, Fotland has stated that he uses null move pruning as well as a few other pruning strategies (found posts here (http://arimaa.com/arimaa/forum/cgi/YaBB.cgi?board=devTalk;action=display;num=1148173288;start=15#15) and here (http://arimaa.com/arimaa/forum/cgi/YaBB.cgi?board=devTalk;action=display;num=1137089374;start=5#5)).

Title: Re: 2008 Postal Tournament
Post by Fritzlein on May 19^th, 2008, 9:40am

on 05/18/08 at 23:15:58, Janzert wrote:

I read each of those posts when written, but apparently it went in one ear and out the other. Thanks for the links. If Bomb was already using null-move pruning and had difficulty coping postally, then OpFor's great start to the Postal Mixer is once again mysterious to me. For now I'm just glad you are still working on it and your work appears to be paying off. I hope that after the postal games your latest version will be available for game room play at least some of the time. Especially I wish you well in the 2009 Computer Championship; your latest additions may already have taken you past Bomb!

Title: Re: 2008 Postal Tournament
Post by aaaa on May 20^th, 2008, 1:09pm

on 05/17/08 at 08:10:49, Fritzlein wrote:

Do we know whether Bomb has null-move pruning?

I had no trouble finding the answer to this question using the search option. You can restrict posts to specific members, but it's not enough just to put a name in the "By User:" field; you also have to select the appropriate option in the pull-down menu right below it.

Title: Re: 2008 Postal Tournament
Post by Fritzlein on Jun 8^th, 2008, 4:11pm

on 05/14/08 at 16:07:23, Fritzlein wrote:

I wonder if OpFor isn't benefiting from a bit of complacency on the part of humanity.

Well, there goes the excuse of being taken by surprise. OpFor has beaten both mistre and arimaa_master from unclear positions they thought they were winning, after each had responded to comments about OpFor's unusually good performance. I'm starting to believe that OpFor (including improvements between the Computer Championship and the Postal Mixer) is now clearly the strongest bot ever. I am curious to see, after the Postal Mixer is over, how the improved OpFor does against the old Bomb, and also how it does against humans in fast games.

Title: Re: 2008 Postal Tournament
Post by mistre on Jun 8^th, 2008, 5:22pm

What is interesting in both mine and Arimaa_Master's losses to OpFor is that we both had a horse framed. I don't know of any other bot that tries to frame horses and actually succeeds (Other bots attempt but usually don't know how to hold the frame). Later on, OpFor framed one of my dogs and actually made me give it up.

I can't help but wonder if I bungled the end game in my match as I was nearly equivalent materially. OpFor pulled off some surprisingly strong moves down the stretch. Regardless of how I underestimated OpFor, I still thought it would blunder eventually or make some weak moves, but it never did. It's 47b, which I originally thought was a weak move, was actually a very sneaky and strong move that caused me to blunder on 48w and pretty much sealed my chance for a comeback.

Title: Re: 2008 Postal Tournament
Post by RonWeasley on Jun 9^th, 2008, 4:48am

In my games against OpFor, I felt it left its camel too exposed to my elephant in the opening. I was able to use that to create lasting weakness. I can't say OpFor misdefended the E-H attack because it was trying to defend an offside dog at the same time. The good news is that OpFor set up opportunities for me to blunder that I sometimes didn't see right away. Bomb has that effect on me too, so I'm not sure OpFor is ready to beat Bomb. I would rate OpFor near 1900 and recommend it keep its camel safer in the opening.

Title: Re: 2008 Postal Tournament
Post by Arimabuff on Jun 9^th, 2008, 6:02am

I wonder how well would do the "old" OpFor against the new one. It'd be interesting to pit them against each other or even propose the two versions in the gameroom bot list.

Title: Re: 2008 Postal Tournament
Post by Fritzlein on Jun 15^th, 2008, 12:36pm

I just noticed the fascinating endgame RonWeasley and mistre are having. On move 72 it is down to EHDCCRR vs. EHDCRR. How much is an extra cat worth in such a position? How should one play when the board is so empty? With so little experience in endgames, I really have no idea, and should not say anything if I did, given that the game is ongoing.

Let me at least say, however, that it could make a huge difference whether draws are still in effect or not. I checked the tournament rules, and they don't say! I would imagine that draws were in effect at the start of the Postal Mixer, although this wasn't specified, and therefore they should be throughout. It would be best to clarify that in case it is affecting the play of either party.

Title: Re: 2008 Postal Tournament
Post by Fritzlein on Jun 15^th, 2008, 12:57pm

I count that exactly half of the Postal Mixer games have completed (42 of 84). Meanwhile the two games left that are least progressed are the two between chessandgo and myself; they are on moves 9 and 10 respectively. We did not plan this in order stage a climactic finish, but I hope that's how it turns out!

Title: Re: 2008 Postal Tournament
Post by mistre on Jun 15^th, 2008, 6:54pm

on 06/15/08 at 12:36:25, Fritzlein wrote:

Thanks for noticing. I really do feel fortunate that I have got to the position I am in. Ron has played a tremendous game. I have no idea how I am going to play this ending - but it should be interesting. I'll save any actual analysis until the game is over, of course.

Title: Re: 2008 Postal Tournament
Post by RonWeasley on Jun 16^th, 2008, 3:24am

Oh man! Now everybody's watching. And the Slytherins are singing that song again.

Title: Re: 2008 Postal Tournament
Post by Fritzlein on Jun 19^th, 2008, 8:11pm

on 06/08/08 at 17:22:11, mistre wrote:

What is interesting in both mine and Arimaa_Master's losses to OpFor is that we both had a horse framed.

Live by the frame, die by the frame. It looks like OpFor lost to sharp due to over-valuing the camel frame. It's weird how OpFor has done so well, yet lost to the lowest-seeded player in the tournament. Sharp appears poised to lose all games except against OpFor, and a gift (on time) from nbarriga.

OpFor can still win eight games of its seventeen by pulling out the victory in its final game against woh. So far it looks pretty even. Whatever the outcome is, I will declare it a fluke. :P

Title: Re: 2008 Postal Tournament
Post by Fritzlein on Jun 19^th, 2008, 8:12pm

on 06/16/08 at 03:24:06, RonWeasley wrote:

Oh man! Now everybody's watching. And the Slytherins are singing that song again.

Everyone is watching you take sole possession of first place, with five wins and no losses. Well done!

I'm amazed that, at this late stage of the tournament, there are still six undefeated players.

Title: Re: 2008 Postal Tournament
Post by RonWeasley on Jun 20^th, 2008, 4:16am

on 06/19/08 at 20:12:43, Fritzlein wrote:

Everyone is watching you take sole possession of first place, with five wins and no losses. Well done!

I'm winning! Well, let's not count acromantulas before they molt.

Title: Re: 2008 Postal Tournament
Post by Arimabuff on Jun 20^th, 2008, 4:27am

on 06/20/08 at 04:16:35, RonWeasley wrote:

I'm winning! Well, let's not count acromantulas before they molt.

Let's not chicken before the count's hatchet. ;D

Title: Re: 2008 Postal Tournament
Post by Fritzlein on Jul 13^th, 2008, 4:46pm

Still none of the players have completed all of their games. I feel sorry for woh. Four players (OpFor, TheJeh, ArifSyed, and ChrisB) have completed all of their games except against woh, so whenever the poor guy moves in any of those games he can expect a quick response.

Mistre, I don't know how you have such a knack for getting in unclear positions, but I noticed your endgame against Adanac is another exciting one. This tournament has produced some great fighting Arimaa!

Title: Re: 2008 Postal Tournament
Post by Fritzlein on Jul 14^th, 2008, 6:46pm

61 of the 84 games in the tournament have finished. The_Jeh is the first participant to finish all of his games over the board. Against an average opponent rating of 1956, he scored four wins in five games, for a performance rating of 2197. Well done, John!

Title: Re: 2008 Postal Tournament
Post by The_Jeh on Jul 14^th, 2008, 8:37pm

Thanks, Fritz. You were right, though - I should have taken a few more games. I'll be sure to take at least eight next year.

Title: Re: 2008 Postal Tournament
Post by Fritzlein on Jul 19^th, 2008, 7:15am

I thought my load of eight games was too few last year (except in a few moments of frazzledness), but this year I thought thirteen was too many. Maybe I'll get wise and take ten a year from now on; ten used to be the mandatory standard.

We have just passed the 100-day mark, and only three undefeated players remain: chessandgo, RonWeasley, and myself. Given that the first three postal tournaments each had an undefeated player, it's natural to ask who will go undefeated this year, but I wouldn't be surprised if nobody runs the table, because we are each in trouble in at least two games.

Sharp has also finished all of its games, with a record of 1-8, not counting a time win from an equal position. Against sharp's particular opponents, that's a performance rating of 1568, more in line with what I would expect from a bot playing at postal speeds, although a bit on the low side.

Lightvector, thank you very much for entering sharp in the Postal Mixer this year. I hope the games are useful to you in your quest to make the evaluation more sophisticated. Certainly it was a treat for me (and I hope others) to get a serious game against the new bot on the block.

Title: Re: 2008 Postal Tournament
Post by mistre on Jul 19^th, 2008, 10:53am

The biggest surprise of the tournament for me is how poorly Blue22 is faring especially since he beat me pretty easily. His 4 wins came from Sharp, myself, ChrisB, and a timeout.

In one of his remaining games, it looks like he is trying to swarm Woh the same way he swarmed me, but Woh's E is mobile and centered, so I don't know how that is going to work. Should make for an exciting finish.

Title: Re: 2008 Postal Tournament
Post by aaaa on Jul 19^th, 2008, 11:11am

I understand blue22 has a bit of a reputation of seemingly cavalierly advancing rabbits for its own sake. He's like the opposite of me in that respect.

Title: Re: 2008 Postal Tournament
Post by Fritzlein on Jul 19^th, 2008, 1:24pm

It seems that postal play tips the advantage towards defensive play. For example, robinson's slash and burn tactics that won him the 2006 World Championship were only good enough for four wins and six losses in the 2006 postal tournament.

In that light, it is too bad that blue22 doesn't play in the World Championship. His reckless attacks would be very difficult to refute in live games; I would have bet on him to at least make the final eight this year.

What (still) surprises me most this tournament is OpFor's excellent handling of wide-open games. Blue22 has shredded Bomb many times after losing his camel for inadequate compensation such as a pair of dogs, so it's no wonder he keeps attacking all out. For some reason, though, it didn't work out so well against OpFor.

Title: Re: 2008 Postal Tournament
Post by aaaa on Jul 19^th, 2008, 2:51pm

on 07/19/08 at 13:24:17, Fritzlein wrote:

What (still) surprises me most this tournament is OpFor's excellent handling of wide-open games. Blue22 has shredded Bomb many times after losing his camel for inadequate compensation such as a pair of dogs, so it's no wonder he keeps attacking all out. For some reason, though, it didn't work out so well against OpFor.

Fritzlein, I think that, ironically due to your overwhelming dominance of the game of Arimaa (with my rating system currently considering you over 140 Elo points stronger than chessandgo alone), it could be that you're not in the position to fully appreciate that, despite its selling point, the view of the fabled computer ineptitude at this game is currently in need of a bit of a nuance.
I believe that there are only a handful of people out there that can beat the strongest few bots while truly being oblivious about the nature of his or her opponent. The rest of us lot will struggle mightily in open, tactical games against the top bots, while having much better luck when resorting to well-known (outworn?) anti-computer strategies.
Perhaps a combination of subconscious human overconfidence in his natural advantage against bots in postal games, leading to open games, and OpFor's relatively newness on the scene, too short to expose its weaknesses, might be enough to explain its excellent results in the tournament.

Title: Re: 2008 Postal Tournament
Post by Fritzlein on Jul 19^th, 2008, 4:24pm

That could all be true (except for the overwhelming dominance bit :P).

As long as I am significantly better than OpFor, it will be difficult for me to tell how wide the gap is by playing OpFor. Kasparov was overconfident going into his match with Deep Blue too; he didn't clue in based on the successes of computers against increasingly strong humans.

I have been thinking that we had better not ascribe too much significance to OpFor's results across a mere 17 games, and that we won't know how good OpFor really is until Janzert puts it on line for open play after the Postal Mixer is over. However, it may be that we won't know how good OpFor is even then, because then we will learn new formulas to beat this new bot, which will lead to a new round of overconfidence on humanity's part.

If you think computer ineptitude at Arimaa is over-hyped at the moment, imagine how I felt when I joined in the summer of 2004 and saw only three players rated higher than the top bot. And it was even worse in the summer of 2003. At that time it wasn't clear whether there was even one human better than Bomb! It seemed to me that Omar had put out a lot of confident assertions about his game that reality wasn't backing up.

Interestingly, however, Omar's hype has become slightly more true every year since the first. We humans are still rapidly getting better, and not just by memorizing anti-bot strategies. I don't know when I last learned something by playing a bot, but I'm confident I'm 100 Elo points stronger than I was last year.

Leading up to the 2005 challenge match, there was much more debate about the strength of computer players than there is now. Fotland was still improving Bomb, and it wasn't obvious that new human strategic insights would translate into a lasting human advantage. Perhaps, we thought, every time a human learns something new, bot developers will just feed that insight back into the evaluation function. That's what happened with frames, elephant blockades, and camel hostages, so why not with everything else?

After a much-improved Bomb was crushed for a second time, Fotland worked on it a bit more, but basically threw in the towel on trying to win the Arimaa Challenge. I have taken this as evidence that whatever humans learn about Arimaa really does tip the scale against machines. It helps us more than them.

From that experience, I extrapolate that as long as we are learning more and more about Arimaa strategy, the gap between humans and machines will continue to expand. Meaning no disrespect to possible software improvements, I don't think they will come mostly from me sitting down with a developer and telling him how I think about the game. To a certain extent knowledge of the game is useful and necessary to the programmer, but at some point our fuzzy way of thinking doesn't fit into code.

So, although it is hard for me to get a good read on how much better than OpFor I am, I remain confident that the trend is still in our favor. Even if OpFor has taken a quantum leap and lopped 200 Elo points off of our lead, I think we will build it back up faster than Janzert can keep improving OpFor.

At some point, human learning about Arimaa has got to plateau. The game can only have a limited amount of depth that is within our comprehension. If I ever start to feel that I play just the same this year as I did last year, then I will spend more time looking in the rear-view mirror, and start to calculate the number of years before faster hardware alone overwhelms us.

The wild-card in my mind is improved/different programming techniques. Developers are a long way from having tried everything yet, and some new approaches could turn out to be stronger than anyone expects. Without some software breakthrough, we're set to defend the Challenge until 2020 even if our strategic advances start to plateau tomorrow.

So yes, the fabled computer ineptitude needs a bit (a lot!) of nuance. We're not fantastically far ahead today, and nobody knows what is going to happen next. For the present, however, it looks rather like Arimaa has some genuine intrinsic computer resistance, and the more players and programmers have a go at the game, the more solidly the fact is established.

Title: Re: 2008 Postal Tournament
Post by aaaa on Jul 19^th, 2008, 5:07pm

Excuse me, but I don't want to be misunderstood, so I would like to point out an important distinction that tends to be blurred:
I personally agree fully with the notion that the odds are considerable that, at least till there is a truly major breakthrough in AI, no engine will ever be better than all humans in the game of Arimaa; make no mistake about that. So when it comes to Omar Syed's money, don't expect it to depart from him in the foreseeable future.
However, this given, and this needs to be stressed, tends to be warped into the different notion that every bot is actually bad at the game and this is what I disagree with.
It's well possible that one can improve a bot till it reaches a plateau of, let's say, the fifth percentile of all Arimaa players, making it not that great a threat to the defense of the challenge, but certainly not something I would label as inept.

Title: Re: 2008 Postal Tournament
Post by aaaa on Jul 22^nd, 2008, 5:03am

When bot_OpFor is done playing its last game of the tournament against woh, I would, if possible, love also to play a postal against it. I think I can safely make the claim that there are not that many other people out there (perhaps maybe Arimabuff) whose knack for playing against bots compares so favorably to their general skill (i.e. that against normal-playing humans); so how bot_OpFor performs against an otherwise only moderately strong player like myself could help establish a more clear "floor" of its performance, i.e. against more "optimized" opposition.

Title: Re: 2008 Postal Tournament
Post by Fritzlein on Jul 22^nd, 2008, 6:36am

Perhaps Janzert could arrange immediately for you to play OpFor postally, especially if you intend to move several times a day to get the game over quickly. Whatever machine OpFor is running on is surely idling most of the time, since it is woh's move most of the time. If I understand how the bot script works, OpFor could join a second postal game without changing anything.

Of course it is up to Janzert whether to have OpFor play postal games rather than live ones, if it plays at all. I guess I could see the draw of playing live games only, so as to get a bigger body of games. Also, I'm still curious about one of my games against OpFor (http://arimaa.com/arimaa/gameroom/comments.cgi?gid=76045) as to whether there was only one non-losing move at one point, and how long it will take OpFor to decide that is the case. That position could also be a test position for static goal extensions: how much faster will OpFor understand the situation with that performance tweak? I don't think I have seen the "stop thinking and play the only hope" behavior from a bot, nor am I sure how often the situation would even arise, so I'm curious what happens in one position where it actually did (apparently) arise.

Janzert, one drawback of having created (potentially) the best bot ever is that you are going to be in high demand. :)

Title: Re: 2008 Postal Tournament
Post by Janzert on Jul 23^rd, 2008, 1:01am

I'm fairly reluctant to start any more postal games with OpFor right now. Games at live time controls and BvB games are actually what I'm more interested in at the moment.

I am currently building a file backup server for the house network. Thanks to moore's law even though I went for fairly low cost it will actually be slightly more powerful than my current desktop. Once I have it completely set up and operational I'm planning to set OpFor up to play live games on the server. If I get really ambitious I might try and set up something so people can start games at different time controls.

Unfortunately probably the best user interface to do that would be to allow OpFor to accept game invitations, but it appears that bots can't currently receive invites. There is a comment in the current bot interface that looks like omar started to implement it at one point but decided not to add the extra complexity.

Janzert

Title: Re: 2008 Postal Tournament
Post by Fritzlein on Aug 6^th, 2008, 2:55pm

Now that Adanac has survived a thrilling endgame against mistre, a quick check reveals that he could still run his remaining games. Yes, many of the positions are unclear and unstable, so he could lose four more as easily as lose no more, but nowhere is Adanac clearly losing already. That would raise an interesting possibility of Adanac and The Jeh being the only one-loss players this year. What a change that would be from having an undefeated player every year so far, and two last year!

Title: Re: 2008 Postal Tournament
Post by Fritzlein on Sep 16^th, 2008, 7:45pm

I have noticed that I am getting an automatic e-mail every five days to remind me that it is my move in the games where I am not moving. How long has that feature been around? I love it! Thanks, Omar, for the reminders.

Title: Re: 2008 Postal Tournament
Post by mistre on Sep 17^th, 2008, 5:47am

I also like how the games are now sorted on the postal games page by number of moves from largest to smallest. It makes it easy to find certain postal games.

I also noticed that there are a lot of postal matches that are still on move 1 (I am guessing that these were set up with the automatic pairing and then the player playing gold hasn't logged back in for a while).

Title: Re: 2008 Postal Tournament
Post by Fritzlein on Sep 17^th, 2008, 6:18am

on 09/17/08 at 05:47:14, mistre wrote:

I also noticed that there are a lot of postal matches that are still on move 1 (I am guessing that these were set up with the automatic pairing and then the player playing gold hasn't logged back in for a while).

Yeah, automatic registration for postal games is great for increasing participation, but the non-started games are lame. It harkens back to the first World Championship where registration was free, but there were a ton of no-shows. How do we balance encouraging registration with encouraging commitment? I like that the Postal Mixer and the World Championship now have registration fees that are not refunded in case of forfeits. It doesn't make sense for the auto-postal pairing to have a registration fee, but it's starting to look like it should be opt-in, not opt-out. Having all these non-started games is only going to teach newcomers that the feature is lame and turn them off from using it later on.

Title: Re: 2008 Postal Tournament
Post by Fritzlein on Sep 17^th, 2008, 12:46pm

OpFor has just finished off its last game, against woh, with another victory. That leave OpFor with an 8-9 record for the tournament, and a performance rating of 2060.

This tremendous result must be regarded as some kind of fluke, because OpFor has attained a lower rating than that from live play in the game room over the last few weeks. However, the fluke can hardly be explained by our collective overconfidence, because woh, arimaa_master, and mistre all had fair warning that OpFor was lighting up the scoreboard while their own games were still unclear. I assume they all tried extra hard to win, but lost anyway. Perhaps we didn't know at first how to cope with OpFor's style, and we are learning better since. Alternatively OpFor's huge result is just one of those statistical outliers that will happen from time to time by chance.

Even if OpFor isn't really a 2060-strength postal player, it is obviously very strong, and in my mind the favorite to be the 2009 Challenger.

Janzert, now that OpFor's Postal Mixer is over, are you still willing to test OpFor on this game (http://arimaa.com/arimaa/gameroom/comments.cgi?gid=76045) to see whether there was only one non-losing move, and if so, how long it takes OpFor to play that one move?

Title: Re: 2008 Postal Tournament
Post by mistre on Sep 17^th, 2008, 1:01pm

Is it possible that OpFor is stronger postally than live? I just defeated it live with 1 min time control, and while it is the toughest live version of OpFor that I have faced, I thought it played better in the postal game. Having said that, it has been performing very well the last two weeks in the live format, beating some higher ranked players. It has lost to a few bots which has lowered its ranking, but it is still a very respectable 1947. I would imagine the difference live vs. postal is smaller than 100 pts and probably no more than 30 pts - so a performance rating of 2060 postally sounds high, but I wouldn't call it a fluke.

I would like another shot against OpFor postally for sure (especially because I consider myself a stronger player in that format than live.)

Title: Re: 2008 Postal Tournament
Post by Fritzlein on Sep 17^th, 2008, 1:08pm

on 09/17/08 at 13:01:33, mistre wrote:

Is it possible that OpFor is stronger postally than live?

Hmmm... Postally OpFor has a time advantage that it doesn't have in live games. Most humans won't (can't) invest the two hours per move necessary to be on an even footing time-wise. But that same time advantage didn't help Bomb or Zombie. In fact both Bomb and Zombie had a worse performance rating postally than live. The conventional wisdom is that giving humans enough time avoid short-sighted blunders helps them more than giving computers a whole bunch of extra time to search deeper, because computers already are tactically excellent, and their weak strategy is barely helped by an extra ply or two.

Title: Re: 2008 Postal Tournament
Post by Fritzlein on Sep 17^th, 2008, 1:14pm

I count only eight remaining games:

Adanac vs chessandgo
chessandgo vs Adanac
Fritzlein vs chessandgo
RonWeasley vs Fritzlein
chessandgo vs RonWeasley
Adanac vs RonWeasley
omar vs arimaa_master
woh vs ChrisB

Astonishingly, depending on how these games turn out, RonWeasley, chessandgo, and I could each still end with the best record outright, while Adanac could still end tied for the best record. Also omar vs. arimaa_master could reach 150 moves. It looks like the movie people planned it this way for greater drama. :)

Title: Re: 2008 Postal Tournament
Post by mistre on Sep 17^th, 2008, 2:59pm

In case you missed it, In my postal match vs Chessandgo, I played him virtually equal in material almost all the way up to the end (C vs RR). Eventually his attack was too much for my defense even though I was up a piece. I thought that with 8 rabbits still alive that I would get some kind of shot to break one through, but Chessandgo effectively blocked any potential breakaway attempts. I didn't see where I could have played this one any differently.

http://arimaa.com/arimaa/games/jsShowGame.cgi?gid=82970&s=b

Title: Re: 2008 Postal Tournament
Post by Fritzlein on Sep 17^th, 2008, 7:50pm

Yes, I noticed the game before. I discussed the material imbalance with aaaa in the chat room. After move 40 when you have RR for C, I thought you must be ahead materially, although chessandgo's attack is well underway. I was surprised chessandgo gave up the framed rabbit so easily, and not sure the attack fully compensated.

I went so far as to say that RR should always be worth more than C materially, to which aaaa proposed the possible counter-example of elephant, cat, and six rabbits versus elephant and eight rabbits. This inspired an experimental game on our part, which instead of proving which side was stronger, proved to me that I don't know how to play the endgame.

After move 40, I'm not sure how you should have played exactly, but I think going after his invading horse with your elephant should have been a lower priority relative to getting your own rabbits rolling. So on 41s, why not use the last two steps of your move to advance your g7-rabbit to g5 rather than relocating your elephant from f5 to e6? Actually your elephant might have landed on a weaker square in addition to using up the two steps. From then on you were in such a defensive bind I'm not sure what moves would have been better.

To play the reigning postal champ so tough is indeed an accomplishment to be proud of.

Title: Re: 2008 Postal Tournament
Post by mistre on Sep 18^th, 2008, 10:49am

on 09/17/08 at 19:50:12, Fritzlein wrote:

So on 41s, why not use the last two steps of your move to advance your g7-rabbit to g5 rather than relocating your elephant from f5 to e6? Actually your elephant might have landed on a weaker square in addition to using up the two steps. From then on you were in such a defensive bind I'm not sure what moves would have been better.

To play the reigning postal champ so tough is indeed an accomplishment to be proud of.

This seems to be a recurring theme among my postal losses. Omar and Adanac both made comments to the effect that they were worried about potential rabbit advances that never came. Personally, I did not see the opportunities, but I think it warrants being a little more aggressive in my postal play even when I don't see the immediate benefit.

I was very proud of my play vs Chessandgo in our postal game, in fact I think I caught him a little off guard which was satisfying. While I held out hope for a win, I certainly wasn't expecting it and was pretty sure he would put the hammer lock on me at some point.

Title: Re: 2008 Postal Tournament
Post by chessandgo on Sep 18^th, 2008, 11:19pm

Good game, mistre ! It certainly was a tough one all along. I fear many people found it easier than usual to play the current postal champ this year :)

I'm not sure what to say about the sequence from 7 to 19. The following moves remind me that don't understand the start of endgames, and especially whether advanced (or even pinned) rabbits are an asset or a liability ... I guess I have to adjust my views toward liability ... i any case I messed up with the sequence from 21 to 27, and by letting you play 39b. After 39b I'm starting to feel behind, but I'm not sure what I would do with silver. In any case, that was a fun an atypical endgame ...

Thanks for the game, and looking forward to a rematch

Jean

Title: Re: 2008 Postal Tournament
Post by mistre on Sep 19^th, 2008, 5:54am

on 09/17/08 at 19:50:12, Fritzlein wrote:

So on 41s, why not use the last two steps of your move to advance your g7-rabbit to g5 rather than relocating your elephant from f5 to e6?

I actually looked at that move! I can't remember why I decided not to play it. Also, I think a stronger 4th step on 40s would have been rh7s instead of rf7e preventing his rabbit from occupying h6.

Title: Re: 2008 Postal Tournament
Post by Janzert on Sep 19^th, 2008, 12:21pm

I don't know if I mentioned it in a post here, but some of you may remember that at the beginning of the postal mixer I was hoping that OpFor would have 1 or 2 wins. So winning just under 50% of its games is way beyond my expectations.

My current thoughts on why it did so well are a mixture of it surprising people, being an unfamiliar bot and the disparity in actual time spent thinking about a move.

Regarding how much stronger it could be at postal speeds than live, I think arimaa will fall within the same general area as has been shown for chess and go as regards to search scalability. Where, against fixed strength opponents, doubling the search time increases the strength between 50-100 elo. Also while there are people that argue against it, I think humans do the same if not better with longer times. So in an absolute sense I think postal opfor could be a couple hundred elo stronger than the live version. But of course the humans weren't constrained to live time controls either and were also doing proportionately better. Although since they weren't spending as much extra time per move probably were not gaining quite as much strength as opfor.

In regards to the game with you Fritzlein (76045 (http://arimaa.com/arimaa/gameroom/comments.cgi?gid=76045) move 21). I actually did some fairly deep analysis back in August (two of the runs lasted over 5 and 10 days respectively). I haven't posted anything on it only because I don't completely believe the results and in order to double check it I need to add the ability to get information from opfor on more than just what it considers to be the current best move. But if the current results are to be believed then there are somewhere around 2000 moves that would not lead to a forced goal in 2.

Janzert

Title: Re: 2008 Postal Tournament
Post by Fritzlein on Sep 19^th, 2008, 1:44pm

on 09/19/08 at 12:21:36, Janzert wrote:

Regarding how much stronger it could be at postal speeds than live, I think arimaa will fall within the same general area as has been shown for chess and go as regards to search scalability. Where, against fixed strength opponents, doubling the search time increases the strength between 50-100 elo. Also while there are people that argue against it, I think humans do the same if not better with longer times.

Yes, I think humans clearly gain more than computers per doubling of search time. Why else would blitz bots be harder to beat than slow bots? Are there people that argue human strength plateaus at some point, such that further doublings have no effect? People kept saying that about computers, and were always wrong.

Quote:

In regards to the game with you Fritzlein (76045 (http://arimaa.com/arimaa/gameroom/comments.cgi?gid=76045) move 21). I actually did some fairly deep analysis back in August (two of the runs lasted over 5 and 10 days respectively). I haven't posted anything on it only because I don't completely believe the results and in order to double check it I need to add the ability to get information from opfor on more than just what it considers to be the current best move. But if the current results are to be believed then there are somewhere around 2000 moves that would not lead to a forced goal in 2.

Is the move that OpFor prefers among those 2000 not-losing-in-two moves the same one that Bomb prefers, namely 21w Rd2e Re2e Re1e Rf1e?

Title: Re: 2008 Postal Tournament
Post by Janzert on Sep 19^th, 2008, 4:25pm

on 09/19/08 at 13:44:49, Fritzlein wrote:

I should probably first say that since I don't agree with this side of the argument that I may be misremembering or distorting their viewpoint.

My understanding is that they believe humans gain strength through extra time only by reducing the number of blunders they make. So at some increase in time, presumably somewhere between normal live game speeds and postal speeds, there are no further blunders to catch and that increased time won't provide the human with any new insight into the position.

Quote:

Is the move that OpFor prefers among those 2000 not-losing-in-two moves the same one that Bomb prefers, namely 21w Rd2e Re2e Re1e Rf1e?

Yep, which is why I'd like to see at least one of the other moves that stops goal in two.

Janzert

Title: Re: 2008 Postal Tournament
Post by tize on Sep 20^th, 2008, 11:29am

Quote:

But if the current results are to be believed then there are somewhere around 2000 moves that would not lead to a forced goal in 2.

I think some of these moves could lead to forced loss in 2, even if opfor doesn't realize it. Because the alpha-beta will prune a line as fast as it's worse than the best found, thus the move to be pruned can be a losing move without it being proved.

Title: Re: 2008 Postal Tournament
Post by Janzert on Sep 20^th, 2008, 2:41pm

Ah, yes of course. Modifying OpFor so it searches all the root moves with a full window allows it to find that all but 15 moves are losing with a 13 step search. So most likely it will turn out to be only the one move that is not losing.

[Edit: Yep it now shows all other moves to be losses after 45 minutes and a depth of 15]

Janzert

Title: Re: 2008 Postal Tournament
Post by aaaa on Oct 30^th, 2008, 11:31pm

Congratulations to Fritzlein for winning all his games in the mixer.

Title: Re: 2008 Postal Tournament
Post by chessandgo on Oct 31^st, 2008, 1:10am

Yep, well done Karl !

Title: Re: 2008 Postal Tournament
Post by RonWeasley on Oct 31^st, 2008, 3:41am

Fritzlein is the master of all owls.

Title: Re: 2008 Postal Tournament
Post by Fritzlein on Oct 31^st, 2008, 1:56pm

Thanks, guys. I certainly didn't expect to go undefeated after getting into a lost middle game as Silver against chessandgo. Also my swarm in my other game against him was looking very unpromising at first, and against Adanac I gave up my camel hostage in exchange for only a cat: both of those games could have gone either way. And the fact that I overlooked a forced goal in five against RonWeasley just before he gave me a forced goal in three gives new meaning to the phrase "double-edged".

I'm convinced that, as usual, I spent more time on the games than any of my opponents, which probably provided the extra edge in all those close games. Nobody complained that they were too swamped by real life to play well, but chessandgo in particular burned through his whole reserve and had to make hurried moves anyway. I'm as happy to take a win from commitment as from skill, though. I hope I keep my time advantage next year too.

I'm happy about winning my games, but even more happy that, for the first time in four Postal Tournaments, I finally played somewhat experimentally. For example, against mistre I managed to get into a position I completely failed to understand. We know too little for it to make sense to play every game the same way; we can count on what we think of as being the "right" way to play actually being proven wrong in time. It's much more fun to play loosely and take each position as it comes, rather than being committed to any specific idea.

Ron, I think your clear second place finish means you are simply going to have to play in the World Championship this year. We can't tolerate having a contender for the crown on the sidelines. If necessary, we will all chip in to buy you a computer that works for live games.

Thanks everyone for some fun and challenging games. I'll see you all again next year!

Title: Re: 2008 Postal Tournament
Post by camelback on Oct 31^st, 2008, 3:20pm

on 10/31/08 at 13:56:37, Fritzlein wrote:

Ron, I think your clear second place finish means you are simply going to have to play in the World Championship this year. We can't tolerate having a contender for the crown on the sidelines.

If chessandgo wins RonWeasley, then the second place would be ambiguous. In my opinion Adanac, RonWeasley and chessandgo have a 3-way tie for 2nd place.

I was also wondering why RonWeasley not playing interactive games and would certainly vote for "The Moderator" to show his wizardry in live games.
;D

Title: Re: 2008 Postal Tournament
Post by RonWeasley on Nov 3^rd, 2008, 6:03am

on 10/31/08 at 15:20:13, camelback wrote:

Second place is kind of hard to define. We had different numbers of games and different opponents. I was very pleased with my result as I escaped some bad positions and won against many good players. And I almost won against Fritzlein. But since there's no prize for second on the line, everybody can tie for second if they want to.

As far as interactive games, I can play them now. The issue for me is time. I am either at Hogwarts, where I'm supposed to be studying, or at The Burrow, where I have many family responsibilities. And there is competition for computer time. Finding a continguous hour to play games is very rare for me. That's why I play by owl. We are getting another computer at The Burrow soon and this may make it possible to play interactive there occasionally.

I think we would find that my interactive rating is not very high and I am certainly not a contender for the title. Still, I think I could provide interesting competition and the community would be entertained. Realistically, I may try to compete in the WC next year. Thanx for asking.

Title: Re: 2008 Postal Tournament
Post by Fritzlein on Nov 3^rd, 2008, 6:14pm

With the last two games drawing to a natural close, I thought I'd give one more measure of the great success this year, namely the number of games ending in resignation or timeout.

Year	2005	2006	2007	2008
Players	16	19	20	15
Games	80	95	85	84
Not Finished	20	31	18	8
% Not Finished	25.0	32.6	21.2	9.5

It's a new record for committed play! This is a lesson to me not to go around twisting everyone's arm to sign up for the tournament even if they aren't sure they want to and/or not sure they will have enough time. Yes, we had fewer people sign up this year, but we had the most completed games ever.

Title: Re: 2008 Postal Tournament
Post by omar on Dec 24^th, 2008, 2:33pm

I have sent the refunds of the registration fees for 2008 Postal Mixer. If you were expecting a refund and did not receive it, please let me know through the contact page.

Thanks everyone for taking part. Hope to see you again in the next one.