Arimaa Forum - Print Page


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

        Arimaa >> Events >> 2009 Postal Mixer
        
(Message started by: Fritzlein on Mar 23^rd, 2009, 11:40am)

Title: 2009 Postal Mixer
Post by Fritzlein on Mar 23^rd, 2009, 11:40am

I have signed up. I was thinking of playing fewer games this year because I might not have time, but who I am kidding? I will always make time to play Arimaa.

It looks like the time control is the same as last year, which is great since there seemed to be no time control issues. Last year we had a record low number of timeouts/forfeits/resignations, so we must be doing something right.

Two rules that have provoked lots of discussion in the past:

(for bots) "Between moves the bots algorithm, configuration or even the hardware on with it runs may be changed."

and

(for humans) "A player must not use a computer program to suggest or evaluate moves."

I don't care much for rules that can't be enforced, so the permissiveness of the first rule is more to my taste than the restrictiveness of the second rule. Since we can't stop people from doing whatever they want anyway (such as players getting hints from their bots or developers giving hints to their bots), I vote we make it a freestyle tournament. Cyborgs and teams of humans welcome.

Instead of trying to restrict how the moves are generated, let's just ask people to be candid if any entrant in the tournament is not an unassisted bot or an unassisted human. If the developer says, "bot_Bust was going to play a terrible move, so I changed the evaluation to prevent it," let's explicitly allow that and praise the developer's openness. Similarly if a player says, "I played by myself until move 27, but then I consulted a bot to make sure I wasn't missing a goal threat," let's be cool with that too, and indeed be glad it is out in the open instead of happening out of view because we have banned it.

I expect most humans will prefer to play unassisted (because it's more fun), and most developers won't force their bots to make/avoid specific moves (because they would rather learn than win), and there will be no teams of players (because consultation is a hassle). However, since we can't prevent any of these from happening, we should invite them all to happen, and see if there are any takers.

Incidentally, I'm not suggesting that the World Championship or the Computer Championship or the Arimaa Challenge become freestyle. Those three events each have a specific purpose: determining the best human, determining the best bot, and comparing the two. Relaxing the rules in those events would undermine the point of even holding them. But for the Postal Mixer I don't see why we shouldn't get funky.

Title: Re: 2009 Postal Mixer
Post by chessandgo on Mar 23^rd, 2009, 1:52pm

why not indeed.

Title: Re: 2009 Postal Mixer
Post by Simon on Mar 23^rd, 2009, 3:33pm

Hmm. The rule that really surprised me (from the 2008 rules according to the page - is there any change this year?) was "A player must not imitate the moves from one game over to another game." Even if this means literal exact copying as in GnoBot's book, I don't see the reason for it at least in H v. H games. After all, you're going to leave the book immediately unless the other player also copies the same game - and why would they do that if they are going to lose by doing so?

I'm less concerned about the no-using-bots rule, because:

a) if I sign up I will probably not use a bot, whereas I would likely copy my opening setup and initial moves from other people if that is allowed

and

b) If the mixer includes people who can't do better than a bot even at postal speeds, they may feel like they are essentially forced to have a bot do their moves for them, and thus feel like they can't meaningfully participate.

Title: Re: 2009 Postal Mixer
Post by Fritzlein on Mar 23^rd, 2009, 3:48pm

Good points, Simon. The rule about copying games is not directed at repeating moves from historical games. As you point out, it takes two to tango, and either player in the game can deviate from the historical game on any move. The hypothetical concern being addressed is rather relaying moves. Suppose you get one game against me and another against chessandgo, and you realize that if you play his moves on my board and my replies on his board, you are guaranteed to beat at least one of us! Chessandgo and I would effectively be playing each other through you, which is silly. That hasn't happened in any past postal tournaments; it is just in the rules out of caution. Apparently the rule needs to be made more clear.

As for needing to consult bots to be competitive, I understand the theoretical worry, but I don't think it will come up in practice. Most people will not be consulting bots, and there is usually a wide range of skills represented in the tournament. Since you are paired with the players closest to you in rating, you should get several competitive games. I should add, however, that this is only true for your first four to six games. If you sign up for ten or twelve games, then you will be paired in some mismatches because there won't be enough opponents of similar rating.

Title: Re: 2009 Postal Mixer
Post by camelback on Mar 23^rd, 2009, 4:47pm

I agree and welcome your proposal Fritz.

I request to modify the Postal Mixer status page, so that there are separate columns for Won, Lost and/or In-progress instead of "win - lose".
It would be really helpful to track the games.

Title: Re: 2009 Postal Mixer
Post by Simon on Mar 23^rd, 2009, 5:51pm

I hadn't thought of that. Better clarify the rule and leave it in then.

About the bots: I'm still not entirely convinced; I think if there are multiple very weak players and even one is using a bot, others will be strongly tempted to follow suit even if that means that it is basically the bot doing the moves. Besides, suppose the rule change becomes a standard, and bots become strong enough to beat even strong humans postally?

I'm also not that pleased by the pair-by-rating system - I'm not convinced the ratings are that accurate, and skill levels can change over a long postal game. And I think I'd rather play at least slightly stronger players than equal players anyway.

Title: Re: 2009 Postal Mixer
Post by arimaa_master on Mar 24^th, 2009, 2:23am

on 03/23/09 at 11:40:56, Fritzlein wrote:

I totally agree!

Title: Re: 2009 Postal Mixer
Post by Fritzlein on Mar 27^th, 2009, 9:46am

Out of curiosity, are any developers planning on entering their bots this year?

Title: Re: 2009 Postal Mixer
Post by Tuks on Mar 27^th, 2009, 12:23pm

if they do, i will try much harder on those games than others, gotta keep my dignity, blunders arent acceptable in postal times

btw, omar, can i use the spectator prize to pay for my postal participation, otherwise i have to renew my paypal banking info so that i can pay

Title: Re: 2009 Postal Mixer
Post by Janzert on Mar 27^th, 2009, 6:23pm

Right now I'm still conflicted about entering OpFor. On the one hand I really want to see bots continue to participate in this. On the other, I was surprised last year on how resource intensive* and for me at least how much of a psychological block to further developing opfor it was.

At the moment though I'm leaning toward entering, but with a max of 8 or 10 games.

Janzert

* It kept opfor thinking constantly 24 hours a day from the start of the mixer till sometime in August last year.

Title: Re: 2009 Postal Mixer
Post by Fritzlein on Mar 27^th, 2009, 6:45pm

One thing to consider in addition to reducing the number of games is reducing the thinking time to only an hour per move. With ten games and an hour per move, OpFor could get in all its moves overnight (not all opponents will have moved), leaving the computer it is running on free all day, every day.

I understand about the psychological block to development, but perhaps it would work differently if OpFor were only on-line while you were sleeping. Not only do the rules allow mid-tournament changes, after OpFor's incredible performance last year you have nothing prove in terms of won-lost record. If you made experimental changes that weakened OpFor for for one night, and decided the next morning you had to roll back, I personally would be excited by the active development rather than disappointed by the weak move.

Even running a computer all night is an electricity expense, so I understand if you think twice before committing to that. I just wanted to suggest the option in case it tips the balance towards participation. Just looking back at the Forum post from the last Postal Mixer, OpFor was the talk of the tourney.

Title: Re: 2009 Postal Mixer
Post by 99of9 on Mar 28^th, 2009, 4:57am

I am unlikely to enter Gnobot in the postal. It will continue to play in the continuous tourney when that restarts.

Title: Re: 2009 Postal Mixer
Post by Fritzlein on Mar 28^th, 2009, 5:25am

Hopefully I will be restarting the Continuous Tournament in May; thanks for GnoBot's past and future participation in that event.

Title: Re: 2009 Postal Mixer
Post by omar on Mar 29^th, 2009, 9:44am

on 03/27/09 at 12:23:52, Tuks wrote:

Yes, definitely.

Title: Re: 2009 Postal Mixer
Post by omar on Mar 29^th, 2009, 10:24am

on 03/23/09 at 16:47:14, camelback wrote:

I'll make this change. Thanks for the suggestion.

Title: Re: 2009 Postal Mixer
Post by omar on Mar 29^th, 2009, 12:07pm

I don't mind allowing computer assistance, but there are more aspects of it that we need to consider.

First I think it would be nice if anyone intending to use computer assistance would announce it before the start of the tournament. If I get matched with someone who is a few hundred rating points below me I will probably spend less time on the move then moves with stronger players. But if my lower rated opponent is using assistance he may defeat me because I underestimated my opponents ability. So I would at least like to know before hand if my opponent is planning to use assistance or not.

Can a player use computer assistance in some games and not in others? Or if they announce they will use computer assistance they have to use it in all games? Ideally a player should be able to use assistance in some games and not in others. I may chose to use assistance in games where my opponent is using assistance and not use it in games where my opponent is playing alone. I'll need to setup some way to allow players to indicate (after they have been paired) in which games they will be using computer assistance and have some kind of a deadline for announcing. Humm... but what if my opponent chooses to use assistance right before the deadline and I get stuck playing without assistance :-)

Also should games where computer assistance is used be unrated? Otherwise it will make our rating system inaccurate. Even the WHR rating system using only HH games will be effected by this. Upsets where the lower rated player used assistance to defeat a higher rated player that did not use assistance will effect the accuracy. Remember a rating system is only as accurate as the games you put into it. Just when we thought we had a more accurate rating system that can be used to seed tournaments, we would start corrupting it if we counted the games where computer assistance was used. At the very least I would suggest playing assisted games with a different account. But a new account won't have established ratings so I'll have to manually set them to initially match the rating of the established account.

It is nice to be able to have rules that are enforceable, but sometime you just can't avoid it and the best you can do is ask the players to follow the unenforceable rule. I prefer to avoid unenforceable rule if possible, but as long as we don't have the players sitting at a table where the TD can observe them, I think we will have some rules that can't be enforced. I don't think that we can in general adopt a policy that unenforceable rules are not allowed. The rules page of the Postal Mixer tries to nicely ask the players to not use any kind of assistance:

Quote:

Since these are postal games and there is ample time for a player to think about the moves, the players must be careful to make sure that the moves they make are truly moves that were produced by them. A player must be careful not to discuss the game or a particular move with anyone else. A player must not imitate the moves from one game over to another game. A player must not use a computer program to suggest or evaluate moves. Although a player may use the 'Plan' option provided in the game window to plan out a move without input from any other source.

Ideally I should just drop the whole paragraph from the rules because I can't enforce people not getting assistance from other humans either. Humm... I don't think I would want the Postal Mixer to be like this; especially if the games are going to be rated and played with our regular accounts.

At the very least I need more time to think about this and will not make any changes for this years Postal Mixer.

Title: Re: 2009 Postal Mixer
Post by Fritzlein on Mar 29^th, 2009, 2:49pm

Your raise good questions, Omar. I hadn't carefully thought through some of the implications of freestyle play.

For ratings I think the simplest solution is to have all games be unrated if they include one or two non-traditional players (cyborgs or teams). The seeding could be based on the highest-rated component of the compound team. These two rules would solve all rating-related issues, right?

Also for simplicity, I think it would be best for each entrant to be the same in all games. If I'm going to use Bomb on any move of any game, I should be considered Fritzlein/assisted for all moves of all games. With this simple rule you would probably choose to be unassisted for all games, and lose the option of using computer assistance against only opponents who used it; do you see a problem with that? It seems to me that people who want to play unassisted unless the opponent doesn't must be motivated by a feeling that computer assistance is "unfair". I think that feeling misses the point of freestyle play.

I agree with you that ultimately we have to say what the rules are and trust the players to follow them. Even in a freestyle tournament, nothing prevents me from claiming to be playing unassisted when I'm not. So my motivation in advocating to allow cyborgs isn't purely that I want to get away from an unenforceable rule; it's also that I think that cyborg players are cool in their own right.

I understand your decision to leave the rules unchanged for the present year. I doubt anyone will be disappointed, because nobody stepped forward during this discussion to say they would be interested in playing with assistance. Thus the rule debate has been entirely moot. :)

I'm jazzed that we already have 11 players signed up for the Postal Mixer, and nearly two weeks of registration to go. We could have a great turnout this year!

Title: Re: 2009 Postal Mixer
Post by mistre on Apr 6^th, 2009, 5:55am

I hope to see some new faces join the postal mixer - so far we only have the returning vets.

Title: Re: 2009 Postal Mixer
Post by Simon on Apr 7^th, 2009, 11:47am

I would like to join the postal mixer, but I would need to delay payment of the deposit for a few days. I started the process to associate my bank account with my paypal account on Sunday and verified the account and transferred money to paypal today, but it says it takes 6-8 business days for money to be transferred from the bank account to paypal. Then it might take time to transfer to omar... I could pay instantly if omar would accept credit card-based payments though.

I also can't play too many games because of concerns about available time, at least until the end of April (but I suppose I could just do the startup move, then burn my reserve...).

Title: Re: 2009 Postal Mixer
Post by omar on Apr 7^th, 2009, 3:28pm

Simon. It is OK if the payment gets a little late, since you are still setting up the payPal account. Thanks for letting me know.

You can sign up for just one or two games if that's all that your schedule will allow. Also you will find that the time control is very relaxed.

Title: Re: 2009 Postal Mixer
Post by Simon on Apr 8^th, 2009, 7:07am

Thanks. But, I've just checked to see what happens if I start the procedure to make a payment right now, and it seems to say that it will take the money from my bank account, and only use my credit card if my bank account doesn't have enough money (it does have enough money). I didn't try this before since the postal mixer registration page said to fund the paypal account first. But it looks like I don't have to wait for the funds to reach paypal, I can directly use paypal to send money from my bank account already.

However, before I do so, I would like to ask if I should wait for the paypal account to be funded, since that's what the registration page seems to suggest?

edit - It seems logical that it would take the same time for money to be taken from my bank account whether it's going to my paypal account or someone else's. The transfer between paypal accounts probably works immediately. In that case, I will save 1 day by waiting for money to reach paypal, since I sent the money to fund the account yesterday.

The dilemma about the number of games is in part that I will be strongly tempted to spend unwise amounts of time on them instead of studying, and the temptation will be stronger the more games there are. On the other hand, I will have relatively more free time after exams are over, and then I will probably want to have several games.

Title: Re: 2009 Postal Mixer
Post by Fritzlein on Apr 8^th, 2009, 9:29am

on 04/08/09 at 07:07:58, Simon wrote:

The dilemma about the number of games is in part that I will be strongly tempted to spend unwise amounts of time on them instead of studying, and the temptation will be stronger the more games there are. On the other hand, I will have relatively more free time after exams are over, and then I will probably want to have several games.

Let me encourage you to start with three to five games. That way the whole tournament won't be wrecked by a blunder in one game, but at the same time the commitment won't be overwhelming. If you have more time later, then you can just analyze more deeply in your few games.

I'm feeling a bit overwhelmed at the moment, and am regretting having signed up for 13 games. Maybe Omar will be kind enough to reduce the number for me. :)

Title: Re: 2009 Postal Mixer
Post by mistre on Apr 8^th, 2009, 12:15pm

I think I was a bit ambitious as well with 12. I would like to reduce mine to 8. Also I would like my current rating to be reflected when my opponents are selected.

Title: Re: 2009 Postal Mixer
Post by Fritzlein on Apr 8^th, 2009, 2:26pm

on 04/08/09 at 12:15:21, mistre wrote:

Also I would like my current rating to be reflected when my opponents are selected.

I believe the seeding will be based on the Whole History Ratings generated by woh.

Title: Re: 2009 Postal Mixer
Post by Janzert on Apr 10^th, 2009, 6:33pm

Just dropping a note here in case Omar doesn't notice the PM or email, that I did send a registration for OpFor to Omar yesterday.

Janzert

Title: Re: 2009 Postal Mixer
Post by Fritzlein on Apr 10^th, 2009, 8:29pm

Yay! Thanks for joining again this year, Janzert. I'm eager to see if we humans can put in a better showing against OpFor this year.

I guess your entry raises a question: if we are using HvH Whole History Ratings to seed the tournament, where do we seed OpFor? I guess for bots the game room rating is the only alternative.

Title: Re: 2009 Postal Mixer
Post by omar on Apr 10^th, 2009, 9:30pm

I've setup the games for the postal mixer. Look under "Postal Games" or "View Games" from the postal mixer main page. Let me know if there is any problem. I used the ratings April 10th snap shot of the WHR ratings. For bot_OpFor I used the gameroom rating.

Title: Re: 2009 Postal Mixer
Post by Fritzlein on Apr 11^th, 2009, 10:55am

That was a great idea to schedule all the games in advance so that we can see what the pairings are, in case there are errors. Thanks, Omar! I don't see any problems.

Compared to last year we have the same number of humans but only one bot instead of two. Everyone who asked for twelve or fewer games got as many as they wanted, but the maximum for game-hungry players went down to fourteen from seventeen because this year more players signed up for a more moderate number of games.

Omar, I can't remember if you implemented my feature request from bygone years: Will the clocks of the Gold players start at the scheduled game time?

Title: Re: 2009 Postal Mixer
Post by omar on Apr 11^th, 2009, 3:14pm

on 04/11/09 at 10:55:30, Fritzlein wrote:

Omar, I can't remember if you implemented my feature request from bygone years: Will the clocks of the Gold players start at the scheduled game time?

Yes the clocks will start on all the games tomorrow at the scheduled time. At least it is supposed to; keep your fingers crossed :-)

Title: Re: 2009 Postal Mixer
Post by Simon on Apr 12^th, 2009, 4:21am

So, I am assuming from the discussion that I can copy moves from Fritzlein's games in this same postal mixer, given that I am not playing him.

edit: and the time has passed but the clocks don't seem to be ticking...

edit 2 - I moved against opfor before checking to see if it was vulnerable to bait and tackle - oh well, I guess I will have to work for my victory.

Title: Re: 2009 Postal Mixer
Post by Fritzlein on Apr 12^th, 2009, 6:03am

on 04/12/09 at 04:21:46, Simon wrote:

So, I am assuming from the discussion that I can copy moves from Fritzlein's games in this same postal mixer, given that I am not playing him.

Yes, only relaying moves between two of your opponents is forbidden. There is no problem with you copying my moves. I'd be flattered you copied me rather than chessandgo! But your opponent will quickly deviate and force you to think for yourself anyway.

Copying me would only be akin to relaying moves if your opponent also copied my opponent. In that case, however, you and your opponent would be equally responsible for creating a non-game, and one of you would be sorry for backing the wrong horse.

Title: Re: 2009 Postal Mixer
Post by Fritzlein on Apr 12^th, 2009, 6:13am

on 04/12/09 at 04:21:46, Simon wrote:

edit: and the time has passed but the clocks don't seem to be ticking...

Yes, the clocks are definitely messed up. Not only are the clocks not ticking for the Gold players, they aren't ticking for the Silver players either. I just made my setup in response to Tuks' setup, and my window shows that I used 0s to make that move.

Title: Re: 2009 Postal Mixer
Post by Simon on Apr 12^th, 2009, 6:33am

odd. I haven't experienced such problems on the games once they were started, and the clocks seem to be ticking now even on the games not started (the clocks seem to have started ticking around half an hour ago).

edit - but yes, your game with tuks does seem to have that bug... that is quite an impressive feat of speed arimaa, Fritzlein :)

Title: Re: 2009 Postal Mixer
Post by Fritzlein on Apr 12^th, 2009, 6:52am

Tuks' clock is ticking now, so apparently the issue is resolved.

Title: Re: 2009 Postal Mixer
Post by Fritzlein on Apr 12^th, 2009, 9:45pm

Despite the reduced game load of some players, there are still quite a few mismatches this year, where a mismatch is defined as a gap of 400 rating points or more. However, the number is slightly exaggerated because this year the Whole History Ratings were used for seeding, and those ratings are slightly more spread out than the game room ratings, which I used to calculate mismatches in previous years.

	2005	2006	2007	2008	2009
Players	15	19	20	15	14
Games	75	95	85	84	70
Mismatches	9	5	3	21	19

I'm afraid we are entering a vicious cycle where lower-rated players don't sign up because they see that lower-rated players don't sign up. If the participation wasn't so biased towards the top players, then other players wouldn't have to fear getting clobbered. Maybe next year we can try to actively counteract this phenomenon, and get newer players to sign up en masse.

Title: Re: 2009 Postal Mixer
Post by Janzert on Apr 13^th, 2009, 7:01am

Just a note that http://arimaa.com/arimaa/tourn/postal/2009/showGames.cgi seems to not have linked up with the actual games yet.

Janzert

Title: Re: 2009 Postal Mixer
Post by camelback on Apr 13^th, 2009, 11:52am

on 04/12/09 at 21:45:53, Fritzlein wrote:

May be we can have a cap, restricting the players to a optimal number of games to avoid mismatch. This will ensure newbies having close games.

In my opinion, auto postal would be a better alternative for game hungry players. I hope that having a cap will not modify the tournament atmosphere. From another perspective, only difference between postal tourney and auto postal is that a dedicated tracking page and a required commitment. Most top players already committed.

May be a better algorithm would avoid having a cap.

Just my 2 cents.

Title: Re: 2009 Postal Mixer
Post by camelback on Apr 13^th, 2009, 11:56am

on 04/11/09 at 10:55:30, Fritzlein wrote:

the maximum for game-hungry players went down to fourteen from seventeen because this year more players signed up for a more moderate number of games.

Is counting icons in show games page the only way for finding number of games for each player?

Title: Re: 2009 Postal Mixer
Post by Fritzlein on Apr 13^th, 2009, 12:40pm

Camelback, I'm not so worried about the mismatches that game-hungry players get. They should know what they are getting into. I'm more looking at 2007 when we had the fewest mismatches in part because we had so many players. If we had had twenty players signed up this year as well, I'll bet we could have had just as many games per player on average (10) without as many mismatches. For this year having ten opponents is not just playing the half of the players who are near you, it is playing nearly the whole field!

Title: Re: 2009 Postal Mixer
Post by ChrisB on Apr 13^th, 2009, 7:10pm

Perhaps one way to increase participation, especially among newer players, would be to let players specify a maximum rating difference between them and their opponents (like done in auto postal), in addition to specifying a maximum number of games.

Unlike auto postal, though, I would strongly prefer that one has the option of specifying an unlimited rating difference, to accommodate the more game hungry players.

Title: Re: 2009 Postal Mixer
Post by RonWeasley on Apr 14^th, 2009, 9:11am

Just when the Owl Mixer started, I find I have to click twice on the send button to send my move. The first click gives me the message to click again to really send my move. Omar, don't you trust me anymore? Or did you just add a feature to discourage players from making dumb moves?

Title: Re: 2009 Postal Mixer
Post by Fritzlein on Apr 14^th, 2009, 10:11am

on 04/14/09 at 09:11:42, RonWeasley wrote:

For me that send confirmation happens only when I have expert mode turn on, i.e. when I am using the game window itself for planning rather than using a separate plan window. When I don't have the distinction between a game window and a plan window, I rather like the send confirmation to prevent me from sending a move that I only meant to be analyzing.

Title: Re: 2009 Postal Mixer
Post by omar on Apr 14^th, 2009, 10:30am

on 04/14/09 at 09:11:42, RonWeasley wrote:

Omar, don't you trust me anymore? Or did you just add a feature to discourage players from making dumb moves?

Actually I don't trust myself :-) I almost hit the send button in a postal game when I didn't have expert mode turned on. I wanted to turn on expert move, but wandered my mouse over to the send button. So for postal games, I figured it might be good to always confirm the send even if expert mode is not on.

Title: Re: 2009 Postal Mixer
Post by Fritzlein on Apr 15^th, 2009, 1:15pm

I had four games with Gold in which I opened with a flank camel on g2 and my elephant on e2. I am surprised that nobody tried to out-flank the flanker as Silver with elephant on g7 and camel on b7. Blue22 came closest with elephant on f7 and camel on b7; I guess I should have expected it since his play inspired me to try flank camels in the first place.

I was planning to open with a flank camel as Gold for a least a month, but I saved it for my World Championship game against chessandgo so that nobody would have time to realize the potential in setting up the silver elephant across from the gold camel and the silver camel away from the gold elephant. It seems that I need not have worried. Either the out-flank reply is no good, or nobody saw it anyway!

I was eager to further test my hypothesis that Silver has a slight advantage in the opening, but I guess first I have to demonstrate that flank camels are superior to centralized ones. Only after I have induced other people to starting using my opening setup as Gold can I try to demonstrate that Silver has an out-flanking refutation.

Title: Re: 2009 Postal Mixer
Post by Janzert on Apr 17^th, 2009, 7:56am

Would it be possible to get the View games (http://arimaa.com/arimaa/tourn/postal/2009/showGames.cgi) linked up to the actual games like it was last year? This was quite handy for me to track what opfor was actually doing in its games last year.

Janzert

Title: Re: 2009 Postal Mixer
Post by omar on Apr 17^th, 2009, 4:29pm

on 04/17/09 at 07:56:59, Janzert wrote:

I'll check it out.

Title: Re: 2009 Postal Mixer
Post by Janzert on Apr 17^th, 2009, 7:12pm

Thanks Omar. Not sure if it's related but I also noticed that the Postal Games (http://arimaa.com/arimaa/gameroom/currentgames.cgi?id=7653) link from a player's page shows the games but not the turn number or who's turn it is, I'm don't know if that is the way it has always been though.

Janzert

Title: Re: 2009 Postal Mixer
Post by RonWeasley on Apr 20^th, 2009, 9:19am

It's always hard to keep track of so many games. One thing I notice about the first several moves: When I make a move, I think I'm winning. When I look at my opponent's response, I think I'm losing. Makes me afraid to open games. Maybe I should move without looking (as I do whenever I play Fritzlein).

Title: Re: 2009 Postal Mixer
Post by jdb on Apr 20^th, 2009, 12:00pm

on 04/20/09 at 09:19:32, RonWeasley wrote:

Easy fix, just make sure you're the last one to move in the game.

Title: Re: 2009 Postal Mixer
Post by mistre on May 7^th, 2009, 10:48am

Games have finished already. Blue22, who performed poorly last year considering his ranking, is off to a fast start at 3-0.

Title: Re: 2009 Postal Mixer
Post by RonWeasley on May 20^th, 2009, 5:04am

Opfor hasn't moved in 5 days. Did it go on vacation?

Title: Re: 2009 Postal Mixer
Post by Janzert on May 20^th, 2009, 8:03am

Oops, looks like it decided to take an unscheduled leave. I've sat it back down now and gave it a stern talking to. Hopefully it'll stick with it now.

Janzert

Title: Re: 2009 Postal Mixer
Post by arimaa_master on May 21^st, 2009, 12:21am

on 05/20/09 at 08:03:58, Janzert wrote:

Oops, looks like it decided to take an unscheduled leave. I've sat it back down now and gave it a stern talking to. Hopefully it'll stick with it now.

Janzert

LOL

Title: Re: 2009 Postal Mixer
Post by Fritzlein on May 31^st, 2009, 11:01am

Some stats on openings:
2009[/td]

Elephant File	2005	2006	2007	2008	2009
d	98.1%	79.5%	85.3%	89.9%	90.0%
c		13.7%	9.4%	9.5%	10.0%
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%	2.1%
a	2.5%	15.3%	4.1%	0.6%	5.0%
c	3.1%
ac				0.6%
af					0.7%
ah	55.6%	56.3%	76.5%	59.5%	57.1%
cf	23.1%			8.3%	1.4%
ag		0.5%	1.2%
ach					1.4%
adh		2.6%	4.1%	10.1%	10.0%
abgh	0.6%
acfh	0.6%			14.9%	22.1%
adeh		2.6%	4.1%
Setup Balance	2005	2006	2007	2008	2009
Symmetrical	56.9%	50.5%	54.1%	50.0%	35.7%
Balanced	22.5%	30.0%	34.1%	42.3%	42.9%
Unbalanced	20.6%	19.5%	11.8%	7.7%	20.7%
Gold Move 2w	2005	2006	2007	2008	2009
E up 4	68.8%	26.3%	25.9%	13.1%	20.0%
E up 3 over 1	11.3%	11.6%	1.2%	3.6%	14.3%
E up 3; X up 1	6.3%	41.1%	40.0%	45.2%	45.7%
E up 2; X,Y up 1	3.8%	10.5%	23.5%	27.4%	17.1%
E, X, Y, Z up 1	3.8%	3.2%		1.2%
Other	6.3%	7.4%	9.4%	9.5%	2.8%

All of us except blue22 centralized the elephant in the setup. Fritz rabbits gained popularity at the expense of no rabbits forward and rabbits only behind the traps, but two rabbits forward on the flanks retained a majority. Unbalanced setups gained sharply at the expense of symmetrical setups, while balanced setups held steady in the middle. Elephant forward four steps for Gold's first move made a comeback, but Elephant up three and some other piece up one retained a plurality.

Title: Re: 2009 Postal Mixer
Post by Fritzlein on May 31^st, 2009, 11:05am

on 05/07/09 at 10:48:45, mistre wrote:

Games have finished already. Blue22, who performed poorly last year considering his ranking, is off to a fast start at 3-0.

Make that a 7-1 start. Bravo blue22!

Title: Re: 2009 Postal Mixer
Post by Fritzlein on Jul 3^rd, 2009, 8:44am

Heads up to spectators: ChrisB and blue22 have played a wild and entertaining game, which has miraculously traded down to an endgame with even material. Check it out while the fun is still ongoing!

Half the games are now finished. All players except Simon have finished at least one game, but no player has finished all games.

Title: Re: 2009 Postal Mixer
Post by Fritzlein on Jul 23^rd, 2009, 3:38pm

Unless Simon times out, he will finish demolishing OpFor for an impressive victory. Well done, Simon!

This will give OpFor a 3-7 record for the tournament and a performance rating of 1798. That persuades me that OpFor's performance in 2008 was a statistical outlier.

Here is the performance of bots in postals so far:

Bot Year Record Postal Performance
--- ---- ------ ------------------
Bomb 2005 4-6 1716
Zombie 2007 1-14 1409
Sharp 2008 1-8 1568
OpFor 2008 8-9 2060
OpFor 2009 3-7 1798

If we take human performance as the benchmark, and say that a bot's "true" strength is how strong it is at the Arimaa Challenge time control of two moves per minute, then I guess the rating adjustment based on slower/faster time controls should be approximately:

Speed Bot Strength
----- ----------
Postal -100
Challenge 0
Fast . +100
Blitz . +150

Probably if humans and bots used equal thinking time in postal games, then the advantage to humans would be even greater than 100 rating points, but bots apparently make up some of the gap by thinking longer per move than their human opponents.

Title: Re: 2009 Postal Mixer
Post by Fritzlein on Aug 16^th, 2009, 7:47am

The game ChrisB vs. blue22 (http://arimaa.com/arimaa/gameroom/comments.cgi?gid=113833) has finally ended on the eighty-first move. This incredibly dramatic game should be reviewed by anyone who doesn't think Arimaa is exciting. It had more back-and-forth than a swingset. Congratulations to both players on a hard-fought, creative, take-no-prisoners, to-the-death battle.

Title: Re: 2009 Postal Mixer
Post by ChrisB on Aug 16^th, 2009, 11:37am

Thanks, Fritzlein. It certainly was an exciting and interesting game for me!

I found it interesting that a large number of blue22's moves were very different from those I thought he might make. Given the amount of tactical interplay, I would have thought that it would be easier to predict the lines of play.

Congratulations blue22 on the impressive win!

Title: Re: 2009 Postal Mixer
Post by Fritzlein on Aug 18^th, 2009, 9:25am

Now that chessandgo has finished thrashing me, I recall the historical trend: In 2005, 2007, and 2008 (i.e. every year except 2006) the Arimaa World Champion went on to be undefeated in the Postal Mixer. Are we witnessing history repeat itself?

Title: Re: 2009 Postal Mixer
Post by Fritzlein on Nov 9^th, 2009, 6:03pm

Adanac's victory over chessandgo today has insured that I will top the standings on the official page. I cleverly chose to play nine games as opposed to chessandgo's eight, so that even though we each lost once, my final score is plus seven to his score of plus six. Incidentally, Adanac also finished on plus six. My choice to play nine games ultimately insured me top billing; that's what I call long-term strategic planning!

But in another sense, chessandgo, Adanac, and I all finished near the middle of the pack. A truer measure of who had a good tournament or a poor tournament is how we each did relative to expectations. Given chessandgo's dominant rating in the wake of the 2009 World Championship, for him to win seven and lose one was a par score. Meanwhile Tuks's six wins and eight losses against opponents rated (on average) well above him was a stellar performance.

ChrisB's win over woh today establishes ChrisB as the clear tournament winner, while Simon will take second. I am assuming Simon will complete his victory over woh, leaving the only game with an uncertain outcome ChrisB vs. Simon. Thus the standings are:

Player . Rating Games Wins Expected Delta
---------- ------ ----- ---- -------- -----
ChrisB . 1803 12 7(8) 3.15 3.85 (4.85)
Simon . 1554 4 3(4) 0.71 2.29 (3.29)
Tuks. . 1904 14 6 4.48 1.52
omar. . 2014 4 3 1.99 1.01
Fritzlein. 2537 9 8 7.49 0.51
chessandgo 2604 8 7 6.92 0.08
blue22 . 2267 14 9 8.94 0.06
Adanac . 2451 12 9 9.50 -0.50
mistre . 2002 8 3 3.61 -0.61
bot_OpFor. 1897 10 3 3.89 -0.89
woh . . 1720 12 1 2.15 -1.15
The_Jeh . 2036 8 2 3.27 -1.27
RonWeasley 2279 12 6 8.01 -2.01
arimaa_master 2093 13 2 5.88 -3.88

One could argue that if Simon prevails in the final game for a perfect score of 4-0 he should rightfully be considered the tournament winner. I certainly tip my hat to Simon and acknowledge an impressive performance from the pre-tournament underdog. But he only played four games simultaneously, while ChrisB was taking on twelve against much tougher average competition. Given his greater time pressure and greater scale of undertaking, I have to rank ChrisB's performance as the more amazing feat.

It is fitting that, while all the other games are long decided, ChrisB vs. Simon is presently, after 62 moves, just as unclear as it was to begin. This is a game that could easily go 100 moves and have many more dramatic swings before it is over. I truly hope that this clash of the over-performers is not prematurely terminated by the 300-day game-time limit. If time is what ultimately decides this epic battle, I will vigorously argue that the 2010 Postal Mixer should have a 360-day game-time limit instead.

Title: Re: 2009 Postal Mixer
Post by Tuks on Nov 10^th, 2009, 7:19am

im satisfied with that, even though, without sounding too petty, my game against chris was a blunder loss, im pretty sure i was in a position with a forced goal but instead i committed a suicide move.

other than that, i look forward to next year, ill keep reminding myself not to move too fast, i got through my games this year months before most other people and i had more games to get though than most other players too

Title: Re: 2009 Postal Mixer
Post by ChrisB on Nov 12^th, 2009, 9:33pm

on 11/09/09 at 18:03:10, Fritzlein wrote:

Player . Rating Games Wins Expected Delta
---------- ------ ----- ---- -------- -----
ChrisB . 1803 12 7(8 ) 3.15 3.85 (4.85)
Simon . 1554 4 3(4) 0.71 2.29 (3.29)
Tuks. . 1904 14 6 4.48 1.52
omar. . 2014 4 3 1.99 1.01
Fritzlein. 2537 9 8 7.49 0.51
chessandgo 2604 8 7 6.92 0.08
blue22 . 2267 14 9 8.94 0.06
Adanac . 2451 12 9 9.50 -0.50
mistre . 2002 8 3 3.61 -0.61
bot_OpFor. 1897 10 3 3.89 -0.89
woh . . 1720 12 1 2.15 -1.15
The_Jeh . 2036 8 2 3.27 -1.27
RonWeasley 2279 12 6 8.01 -2.01
arimaa_master 2093 13 2 5.88 -3.88

Interesting analysis Fritzlein. It nice for the low-to-middle-ranked players to have an additional chance to get a top score.:) It's tough, though, for the top ranked players to win this metric. Had any of the three highest ranked players (chessandgo, Fritzlein or Adanac) gone undefeated, they still would not have had the highest score.

I set up a spreadsheet which duplicated your analysis (I got identical scores) and would be interested in doing this analysis again for next year's postal mixer.

Title: Re: 2009 Postal Mixer
Post by Fritzlein on Nov 13^th, 2009, 7:59am

on 11/12/09 at 21:33:53, ChrisB wrote:

It's tough, though, for the top ranked players to win this metric.

It's only fair to add a metric by which all but the top-ranked players can do well, since it is tough for all but the top-ranked players to go undefeated and win by the traditional metric. Perhaps chessandgo, Adanac, and I could not have won by this metric, but we have the consolation of topping the standings in the "official" scoring of wins-losses.

Of course, one would never want to give prizes on out-performing expectations, because then the easy way to win is to lower expectations. Chess tournaments have suffered from people sandbagging to lower their ratings and thereby get into easier divisions.

Personally, I liked calculating this metric because I was so disappointed to lose even one game this year. Using the ratings helps to remind me that my par score was losing 1.5 games, so that my "disappointing" performance was actually slightly better than my average performance.

I have noticed chess players are prone to this disease too. If you listen to chess folks talk, they are almost always performing below their true ability because of (fill in excuse). Only when they are performing brilliantly, far above their usual level, do they consider it an "OK" performance. I always wanted to tell chess players that playing flawlessly is not normal, it is a wild aberration.

Quote:

I set up a spreadsheet which duplicated your analysis (I got identical scores) and would be interested in doing this analysis again for next year's postal mixer.

Awesome. The job is officially yours. :)