DeveloperCreated Thursday, November 26th 2015 @ 14:46:17
The dates for the finals are known! Lockdown, starts from Sunday January 10, 23.59 CET. The semi-finals will start exactly one week after that, namely Sunday January 17, 23.59 CET. During lockdown you cannot update your bot anymore, and your faith will be in the hands of the bot you created.
The finals will be in the format we usually have, except the amount of games per round will be increased to get a better estimate of the better bot. Of course we have some great cash prizes again for the 8 finalists, with the winner taking over € 1000!
- Created Friday, November 27th 2015 @ 21:37:36
Great news! I appreciate that you will increase the number of games in the finals.
- Created Monday, January 11th 2016 @ 00:07:42
Good luck to you all in the semi-finals and finals!
- Created Monday, January 11th 2016 @ 00:12:02
I'm hoping for the best. Just started my attempt for this one 4 days ago, and I'm afraid I'm way of the cut-off, but maybe the last one was a doable one. :) Good luck for those who have better changes. :)
- Created Monday, January 11th 2016 @ 03:40:23
Good luck guys
- Created Monday, January 11th 2016 @ 05:36:37
Best of luck to you all!
- Created Monday, January 11th 2016 @ 15:12:50
Great work on staging all this! I do have a concern. Looks like games take 2 minutes on average. Hence about 720 per day or roughly 5k for the week. With 147 participants this means 35 games on average per bot. Omaha had about 1/3 the field so maybe 3 times as many games (100+). With the variance in the game I'm not sure 35 games for each bot is enough to get a very accurate rating distribution. With such a large field it may have made more sense to keep the initial ratings and let the weeks worth of matches simply sort out any minor discrepancies, even if someone made significant improvement in their last version.
- Updated Monday, January 11th 2016 @ 17:45:28
5k games for the week means 67 games per bot, not 35
- Created Monday, January 11th 2016 @ 18:31:26
Correct. I realized that soon as I had posted (2 bots play in each game). However, the entire ELO rating system on which this is based awards pts based on the likely hood of one opp defeating another given their rating difference. This becomes meaningless for the initial rounds when all have been set to an arbitrary value. It is possible that this is enough games to converge on a reasonable estimate. I used to have Elo's book and seem to recall it had some graphs of confidence levels in the ratings based on number of contests, but no longer have the book and could not find on the web. However, starting the bots with reasonable estimates to their 'actual' ratings will converge much more quickly than starting everybody at 0 as it were. The USCF has provisional rating for less than 20 games, but that is 20 games against established ratings, I would think against 'air' ratings it would have to be double that. Anyway, I am new to this and it just seemed very odd to do it in this fashion.
- Created Monday, January 11th 2016 @ 19:17:24
I was wondering about similar details.
@jim: Could you guys elaborate on how the finals will be run specifically? (i.e.: how many games per bot to determine rank, etc).
DeveloperCreated Tuesday, January 12th 2016 @ 08:46:52
To clarify: The lockdown lasts 1 week after which we have determined 24 top bots, those will progress to the semi-finals in a double elimination knock-out scheme. Each match between two bots will be a first to 10 wins match, with at least 2 win difference. After the semi-finals, 8 bots remain which will be in a single elimination knock-out scheme with a first to 6 wins match. The 'live' finals will be watchable on the finals page.
- Created Monday, January 18th 2016 @ 21:53:22
Ugh, looks like the matchups were determined randomly. I'm not at all thrilled about my first round match, but, I guess it is what it is. Best of luck to my first round opponent, "HowDoesItWork"! But not too much ;)
- Created Thursday, January 21st 2016 @ 02:12:25
- Created Thursday, January 21st 2016 @ 10:09:35
If I look at the finals it is clear how the winner is determined, but how are the second to eighth place determined?
- Updated Sunday, January 24th 2016 @ 16:34:42
Are the games in the finals being recorded correctly? It seems like there's a mix-up between the names of the bots and the actual bots playing (during the live preview) similar to what happened during the Warlight 2 finals.