- Updated Tuesday, February 10th 2015 @ 13:15:17
There was a mail sent out already, but I'm putting it up here as well so you can easily look up the format for the finals again. It'll be largely the same as we did with the first Warlight competition. On Sunday February 1, 23.59 (GMT+1), we will reset all the ranks in the leaderboard back to default. From that time, until Sunday February 8, 23.59, Heads Up Omaha will be in lock down, meaning you cannot enter any new bots in to the competition.
At the exact time of the end of the lock down, the top 24 bots in the leaderboard will progress to the semi-finals. There the bots are matched up in a double-elimination tournament scheme, meaning that if you lose twice, you are out. Each match will be a first to 7 match, with a minimal difference of 2 wins. Each day of the week another round of the tournament will be played. At the end 8 bots will progress to the finals.
Sunday February 15 18.00, the finals will be played. This will be a single-elimination tournament, with best of 5 matches between the bots. Games can be viewed 'live': every 10 minutes a game is played that you cannot skip through, until the winner is known.
The 8 finalists will be awarded with a prize, but the grand winner get over a thousand euros! If you think you have the skills to reach those finals, go code, because when the lockdown starts it's out of your hands!
As a note: we are aware that this is not the most "fair" form of tournament imaginable, but we want to keep things exiting. Poker always has an element of luck after all. If you think you can reach the semi-finals, make sure your bot can handle the tactics of the other bots in the top 24.
Good luck with your bot and have fun coding!
- Created Friday, January 30th 2015 @ 11:12:06
Hi Jim, at the end of the lock down, (after selection of first 24 bots) will it be allowed to update bots before the next round of the finals (before starting matches to select 8 bots) ?
- Created Friday, January 30th 2015 @ 11:28:42
No bots can only be updated once the finals are completely over.
- Created Thursday, February 5th 2015 @ 14:20:24
Are you planning to match bots randomly in the elimination round ? Or is there a system like "1 vs 24, 2 vs 23, ...."
- Created Thursday, February 5th 2015 @ 14:36:38
We have the following order (numbers are a bot's final rank on the leaderboard):
1v5 9v13 17v21 23v19 15v11 7v3 4v8 12v16 20v24 22v18 14v10 6v2
- Created Thursday, February 5th 2015 @ 22:11:21
the ranks will most likely still be semi random especially for top 10 because we won't reach close to the equilibrium elo points in just 1 week of running. top ranks are harder to hold onto because as people in top 30 are gaining points per game on average toward equilibrium, the highest players play a lot less games and are therefore slightly handicapped. I'm not complaining about it though just pointing out some observations. it's not a big deal as long as the top 24 are more or less in the top 24 which should be the case
- Created Friday, February 6th 2015 @ 02:13:38
I'm sure this would make the first few games more entertaining, but if it's maximum excitement that you're going for, don't you think it might actually be better to do more typical seeding? Because then after the first set or two, the matchups would be really even in both the winners and losers brackets all the way to the end, which would be much more exciting, I would think. It would also be a lot more fair presumably, because it seems kind of unreasonable that the 5th seed will be knocked down first round but the 20th seed will stay in?! It just seems to leave a lot more up to chance so, I'm not sure if this is the best tournament model.
- Created Saturday, February 7th 2015 @ 17:07:02
I also think it will make the tournament a lot more random.
- Created Sunday, February 8th 2015 @ 00:52:51
I agree. This way there are 4 strong pools and 2 weak pools. From the weak pools at least two bots reach the finals, but they don't stand a chance in the finals. That's what happened in the Warlight I finals and I think it will happen in the poker finals as well. I would prefer 6 pools of equal strength, and the top 6 players in separate pools.
- Created Sunday, February 8th 2015 @ 01:33:10
I think the randomness from the match making based on seed is unimportant compared to randomness from playing to first to 7 wins. especially since the seed number will not be accurate anyways based on elo. if you win 55% of the games against someone you will only win a match with 65% chance. 60% leads to 78% win rate in match. given that there are multiple rounds to play the final result will still be very random without worrying about seed numbers.
while I don't think there are enough time for changes to occur. I would like to point out its weird that there's months of computation spent on practice games, 1 week of computation for an estimated seeding, but less than one day of total computation will be spent on semifinal and finals combined. I don't care about the prize so much but I want to know if it will be possible after the finals are over to later play the top 8 or 12 players against each other for 100 games each to see who is truly the top 8?
- Updated Sunday, February 8th 2015 @ 20:33:39
I have greatly enjoyed the format format of the Heads Up Omaha competition thus far and also appreciate how you increased the amount of games played in order to reduce randomness in the competition.The one thing I really still do not understand is the seeding. The way you have it set up, either the 17, 19, 21, or 23 seed is guaranteed to make the finals. The same goes for the 18, 20, 22, or 24 seed. Meanwhile, the 5, 9, and 13 seed all would have to beat the one seed (to advance out of the winners bracket)? I do not understand how this is fair. Would it not make the finals much more exciting if the teams in the finals were actually mostly representative of the top bots? Of course you should not just take the top 8 bots off the leaderboard for the finals, but I think you should at least reward the top 6 bots on the leaderboard by placing them in separate groups of 4, as AdsRiskbot mentioned above. May I propose an alternate seeding arrangement (If you are not willing to change to a conventional one i.e. 1v24 12v13 etc.) that would keep the first round exciting but also almost always reward bots for being higher ranked at most parts of the top 24?
1v7 18v24 6v12 13v19 3v9 16v22 4v10 15v21 2v8 17v23 5v11 14v20
Of course, if it is too late to change the seeding, I understand. However, if it is not too late, I hope that you consider either conventional seeding or the arrangement outlined above.
Thanks in advance for your consideration and thanks for the awesome competition regardless of seeding!
- Updated Sunday, February 8th 2015 @ 20:12:51
It's seems to be the same format as with the first WarLight AI challenge where 'Muli' and 'damian_warlight' got a free pass to the finals. Since WarLight isn't that luck dependant than poker they of course had no chance and got their but whipped by the serious bots in the finals. The only effect was that the serious bots that got paired up with them in the finals got a free pass to the next round. The only bot that could have gone for a surprise victory was 'Herz' since it was a poorly programmed bot that went for a hardcoded 'perfect strategy' on this limited map.
- Created Monday, February 9th 2015 @ 00:14:19
Congratulations to Shaftoe for first place on leaderboard! I am sure the other bots will treat you a lot more respectful now:)
- Created Monday, February 9th 2015 @ 00:18:18
The top 24 bots have been selected. Based on your suggestions I'll change the seeding a bit before the first rounds starts tomorrow.
- Created Monday, February 9th 2015 @ 02:32:13
Thank you so much Jim! Can't wait for the semifinals to start!