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

        Arimaa >> Events >> 2008 Prediction Contest Format
        
(Message started by: Fritzlein on Oct 5th, 2007, 7:36pm)
Title: 2008 Prediction Contest Format
Post by Fritzlein on Oct 5th, 2007, 7:36pm

on 10/05/07 at 17:16:06, Fritzlein wrote:
Random thought: will the prediction contest be only for the finals, or for the two tournaments combined?  One advantage of predicting only the finals is that the eight finalists can be excluded from predicting, eliminating the awkward question of how to handle prediction on ones own games.  However, I still would rather have the prediction contest span both legs, because the greater number of games makes the contest a more accurate measure of predictive skill.

Ah, what would we do without the daily carpool?  During this evening's ride, Omar floated the idea of totally changing the format of the prediction contest.  Instead of having an accumulation of points across all the games, we could have a winner-take-all prediction prize on each game.  Let's say it is $5 per game for each of the 14 or 15 games of the eight-player final.  To win the prize, you would have to pick the right player to win, and be the closest predictor to the actual number of moves.  There would be no more percentages involved: you either call the result or you don't.

My first reaction to the proposal was negative.  I like the old prediction contest, because it is a game of skill, and I'm good at it.  The old format rewards judicious judgment, not just getting lucky.  Why would we replace a battle of calculation with a mere lottery?

But the more I think about it, the more a lottery fulfills the objectives of the prediction contest.  The point of having the prediction prize at all is not to have parallel game of skill, it is to encourage non-players to watch the games and get emotionally involved.  Imagine if you know little about Arimaa, but you have five bucks riding on every game!  You would root for your guy to win, but not too fast, and not too slow, so you are trying to figure out not just who is winning but also how soon it will end.  And if your guess was wrong, well, for the next game you would have just as good a chance of winning, so you would watch that one with interest too.  Contrast this with last year's contest where twenty-five people entered but only ten or so were still making serious predictions by the end.

Instead, the interest should be rising as the tournament progresses.  We could foster this with a progressive jackpot: if two or more players tie for the closest prediction (i.e. both picked the winner and were off by the same number of moves) or if nobody picked the winner (perhaps it was a big upset) then nobody gets the prize, and it rolls over to the next game.  Furthermore, the prize money for the later round games could be $10 each and $15 for the championship game (on top of money that might be accumulating).

I like this idea enough that I'm already speculating about the details.  The two participants in each game would of course be ineligible, but that still lets 20+ spectators predict each game.  Of course the final game couldn't result in a push, so at that point any prize money would have to be split equally between any tied predictors.

Should the predictions become visible as soon as the game begins, or stay secret until the end?  I guess it heightens the suspense to not know, for example, whether anyone else duplicated your prediction.  On the other hand, knowing that one person predicted 49 moves and another predicted 55 moves could have everyone else rooting for the game to end exactly on move 52 to create a tie and push the money forward.

Well, anyway, it's just a possible experiment to make the prediction contest a bigger draw for spectators by making it more exciting, and also to encourage people not to drop out when they get behind.  What does everyone think?
Title: Re: 2008 Prediction Contest Format
Post by chessandgo on Oct 6th, 2007, 3:41am
I agree with Omar and you, this winner-takes-all formula sounds funnier than the big calculations of previous years.
Title: Re: 2008 Prediction Contest Format
Post by 99of9 on Oct 6th, 2007, 3:43am
I think that sounds worth a try.  I expect the winners would be more diverse.
Title: Re: 2008 Prediction Contest Format
Post by 722caasi on Oct 6th, 2007, 12:44pm
I like this too.
Title: Re: 2008 Prediction Contest Format
Post by RonWeasley on Oct 6th, 2007, 1:19pm
I like this in that it makes the late games more interesting.  I like that there will be more winners.  I like pumpkin juice.
Title: Re: 2008 Prediction Contest Format
Post by omar on Oct 7th, 2007, 9:23am
I have to give credit to Naveed for this idea. He had mentioned to me something like this years ago.

There might be some problems with carrying the prize for a game forward to the next game in the case when there are ties. It isn't clear which game in the next round it should carry forward to (maybe to the game of the winning player). But also as the number of players in the prediction contest increases so does the chance that there will be a tie. Also if it happens in the last round there is no way to carry it forward. Maybe we should just keep it simple and always split the prize among the players that tied and in the case where no one picks the correct winner then no prize is awarded for that game.

I might also throw in a fixed amount for everyone that predicted the correct winner in close games where the divide is like 60%-40% or closer.
Title: Re: 2008 Prediction Contest Forma
Post by Janzert on Oct 7th, 2007, 2:33pm
This does sound like it would help increase participation especially through to the end of the tournament. But I would personally much prefer and really miss the old style predictions. I found it really interesting seeing what the community thought the percentages were for each game. I also enjoyed the "big calculations" for each game last year (the only year I actually participated instead of bumming someone into telling me what the predictions were); but then I'm wierd (http://arimaa.janzert.com/fame.html) like (http://arimaa.janzert.com/gamegraphs) that (http://arimaa.janzert.com/bf_study/). :P

On a semi-related note something that I think would be interesting is to go back and see how a ratings based predictor would have done in the past prediction contests. Sort of a test to see if the individual predictors were able to "beat the market".

Janzert
Title: Re: 2008 Prediction Contest Forma
Post by Fritzlein on Oct 7th, 2007, 4:28pm

on 10/07/07 at 14:33:08, Janzert wrote:
On a semi-related note something that I think would be interesting is to go back and see how a ratings based predictor would have done in the past prediction contests. Sort of a test to see if the individual predictors were able to "beat the market".

I actually did this after the 2006 prediction contest.  I used the official rating calculation to predict percentages, and created a move predictor optimized to two variables: X(rating sum) - Y(rating diff).  The bot thus created would have come in about fifth, if I remember correctly.  I didn't try it on the 2007 results, but since there was a sharp dropoff in score after the six predictors in contention for first place, I'm guessing the formulaic predictions would have come in seventh.
Title: Re: 2008 Prediction Contest Format
Post by Fritzlein on Oct 8th, 2007, 2:19pm
I see the new rules are up.  Now it is $10 to play, but you get $1 per game just for picking the right winner each game.  That should be worth over $7 of the entry fee right there; $8 or $9 if you pick only favorites.

The new winner-take-all per-game prize money is

Round 1: 4 games times $5 = $20
Round 2: 4 games times $5 = $20
Round 3: 3 games times $10 = $30
Round 4: 2 games times $15 = $30
Round 5: 1 game times $20 = $20
Round 6: (maybe) 1 game times $25 = $25

For a bonus prize pool of either $120 or $145.  Unless we get fifty or more registered predictors, the average participant will make money.  And if we do get fabulously lucky and have more than fifty spectators, maybe that will put Omar in such a good mood that he sweetens the pot.  What do you say, Omar?

By the way, the tournament final rules still allow for a 2nd/3rd place game to happen after the championship game, although it will be rare.  I expect we want to explicitly exclude that game from the prediction contest this year.
Title: Re: 2008 Prediction Contest Format
Post by Janzert on Oct 8th, 2007, 3:22pm

on 10/08/07 at 14:19:10, Fritzlein wrote:
By the way, the tournament final rules still allow for a 2nd/3rd place game to happen after the championship game, although it will be rare.  I expect we want to explicitly exclude that game from the prediction contest this year.


Since there is no carry over of scoring I expect it won't be problematic for the participants to have it included this year. But it would be slightly weird to have the prize amount increase for the 2nd place game.

Janzert
Title: Re: 2008 Prediction Contest Format
Post by Fritzlein on Oct 8th, 2007, 3:54pm

on 10/08/07 at 14:19:10, Fritzlein wrote:
By the way, the tournament final rules still allow for a 2nd/3rd place game to happen after the championship game, although it will be rare.

I just calculated that if every game between the final four players is a coin flip, there is a 25% chance that an additional game will be necessary to break a tie for 2nd/3rd place.  The more the final four players are unequal, the lower the this percentage becomes.
Title: Re: 2008 Prediction Contest Format
Post by omar on Oct 8th, 2007, 10:22pm

on 10/08/07 at 14:19:10, Fritzlein wrote:
And if we do get fabulously lucky and have more than fifty spectators, maybe that will put Omar in such a good mood that he sweetens the pot.  What do you say, Omar?

Yes, definitely.


Quote:
By the way, the tournament final rules still allow for a 2nd/3rd place game to happen after the championship game, although it will be rare.  I expect we want to explicitly exclude that game from the prediction contest this year.


Thanks for noticing this. I've updated the web page to set the spectator prize for the tie breaker game at $15.
Title: Re: 2008 Prediction Contest Format
Post by Fritzlein on Nov 1st, 2007, 1:05pm
I see the registration deadline is currently set to February 8.  However, I think the deadline should be shortly after the preliminary Swiss completes.  I personally will not sign up for the prediction contest if I make the finals, since I can't predict on my own games during the finals.  Other people who feel like me might still be "on the bubble" in the qualifying tournament as of February 8, and would like to see whether or not they make the finals before registering to predict.  Perhaps the deadline could be fixed as midnight GMT the Monday after the last round of the qualifier, whenever that turns out to be.  Once we know how many people sign up, the date could be set explicitly.
Title: Re: 2008 Prediction Contest Format
Post by omar on Nov 2nd, 2007, 12:07am
Added a spectator prize of $10 for the 3rd place tie breaker game.

Also changed the rule so that if there are ties, the spectator who entered the prediction earlier wins. Spliting the prize can get to be a pain, this makes it easier for me.
Title: Re: 2008 Prediction Contest Format
Post by Fritzlein on Dec 29th, 2007, 9:54pm
Is anybody signing up for this?  I can't see a list of registered predictors.
Title: Re: 2008 Prediction Contest Format
Post by omar on Dec 30th, 2007, 9:05am
Added a link for this: http://arimaa.com/arimaa/wc/2008/predict/players.cgi


Arimaa Forum » Powered by YaBB 1 Gold - SP 1.3.1!
YaBB © 2000-2003. All Rights Reserved.