Poll for Michael's Tournament

[Dave Mitchell]

This is feedback for Michael Sherwin's Tournament, next year.

What do you want to see?

For divisions:
===========
1) unlimited machines (bring your cluster!)

and

2) one cpu (quad) system, all the same. Supplied by the tournament

or

3) Split these up, with a book-neutral section*, and a bring your own opening book sub-division, in each of the above.Total of four divisions.


Neutral book would be supplied by the tournament

[hgm]

I think for the application of money prizes to have any beneficial effect, it would be important that more than 4 or 5 participants have a realistic chance of winning a meaningful fraction of it. If the numbers 1-3 will walk away with 90% of the money, it might be appealing to Rybka / Rondo / Shredder, which would likely be there anyway, and the rest of the world will just shrug, knowing that they have no chance at all to win anything. So the turn-up would not be much larger than at a typical WCCC, i.e. perhaps 10 participants.

If I would have to plan an event like this, I would go for a Swiss pre-tourney to make a sub-division into groups, and then play a round-robin in each group, where there are prizes in each group. If it was a 2-day event (mkaing it longer is usually fatalfor attendance), you could play 5 rounds per day (1 hr sudden-death TC): 9-11am, 11am-1pm, lunch, 2-4pm, 4-6pm, dinner 7-9pm. So you could make finals groups of 6, playing a round-robin on Sunday, and have 6 rounds of Swiss on Saturday as qualifiers. (You could make the latter 5 if you want to start late.) In each group the first 2 could then get prizes (in the top-division perhaps the first 4).

The disadvantage of Swiss, if there is a wide range of strength of the participants, is that the first few rounds are wasted to confirm what you already know (if you pair the first round by Elo). When such a seeding
is used, there is an alternative scheme, which I used in my broadcasted Xiangqi engine tourney:

Sub-divide the field into qualifier-groups (of 6 players each, in this case, as there is time for 5 rounds), based on prior Elo. Play round-robins in these groups. Then make a new sub-division, where you join thetopfinishers in one group with the bottom finishers of the next higher group, to make the finals groups. You could even have the number one and last of the qualifier groups skip a group (i.e. exchange the number last of a qualifier with the number 1 two groups lower before you join lower half and upper half of the qualifier groups to make finals groups). This works better with groups of 8 than with groups of 6, though.

[Dave Mitchell]

Yes, we need to squeeze in as many rounds as humanely possible. Three per day, just isn't enough. Five is much better. More than five might not be possible without having some portion of it being automated to run late at night/early in the morning.

What is your opinion of having one or two rounds run automatically, while the competitors are sleeping?

We haven't discussed the prize money distribution, yet. I agree that spreading the money around a bit beyond first through third place, would be good. That is the custom with sports tournaments I'm familiar with.

I thought about a pre-tournament Swiss, but rejected it, for the simple reason that you don't know WHAT is really going on when you play a match over the net. I mean, you could wind up playing on a quad, and your opponent is using 2,000 hex core servers in Google's basement or something.

I like your idea about round robin within the bracket, and then making a final bracket with the best of both divisions included. The details will have to be ironed out later, but your model is a good one, imo.

Great to see that you have a lot of input. The more we discuss these idea's, the easier it will be for Michael, later.

[Michael Sherwin]

Yes, we need to squeeze in as many rounds as humanely possible. Three per day, just isn't enough. Five is much better. More than five might not be possible without having some portion of it being automated to run late at night/early in the morning.

What is your opinion of having one or two rounds run automatically, while the competitors are sleeping?

We haven't discussed the prize money distribution, yet. I agree that spreading the money around a bit beyond first through third place, would be good. That is the custom with sports tournaments I'm familiar with.

I thought about a pre-tournament Swiss, but rejected it, for the simple reason that you don't know WHAT is really going on when you play a match over the net. I mean, you could wind up playing on a quad, and your opponent is using 2,000 hex core servers in Google's basement or something.

I like your idea about round robin within the bracket, and then making a final bracket with the best of both divisions included. The details will have to be ironed out later, but your model is a good one, imo.

Great to see that you have a lot of input. The more we discuss these idea's, the easier it will be for Michael, later.

YES! -- Thanks.

[Dave Mitchell]

bump for Michael's tournament poll

[lucasart]

I believe you are in the wrong forum.

this is for people who are interested in technical discussions about chess programming.

there is another section for tournament and match results discussions.

[Don]

As far as prize money is concerned I like the idea of giving out money for points won. It could be done in a non-linear way so that doing poorly should not be rewarded too much. This is very easily done, for instance doubling the "credit" for each half point won. You divide the total points that can be scored by the prize fund to determine how much each credit is worth. So someone who scores 7 points wins twice as much as someone who scores 6.5 points, etc.

Of course it does not have to be double for each half point, it can be any multiplier and the choice determines how aggressively the winners are rewarded at the expense of the lower scoring programs.

I think for the application of money prizes to have any beneficial effect, it would be important that more than 4 or 5 participants have a realistic chance of winning a meaningful fraction of it. If the numbers 1-3 will walk away with 90% of the money, it might be appealing to Rybka / Rondo / Shredder, which would likely be there anyway, and the rest of the world will just shrug, knowing that they have no chance at all to win anything. So the turn-up would not be much larger than at a typical WCCC, i.e. perhaps 10 participants.

If I would have to plan an event like this, I would go for a Swiss pre-tourney to make a sub-division into groups, and then play a round-robin in each group, where there are prizes in each group. If it was a 2-day event (mkaing it longer is usually fatalfor attendance), you could play 5 rounds per day (1 hr sudden-death TC): 9-11am, 11am-1pm, lunch, 2-4pm, 4-6pm, dinner 7-9pm. So you could make finals groups of 6, playing a round-robin on Sunday, and have 6 rounds of Swiss on Saturday as qualifiers. (You could make the latter 5 if you want to start late.) In each group the first 2 could then get prizes (in the top-division perhaps the first 4).

The disadvantage of Swiss, if there is a wide range of strength of the participants, is that the first few rounds are wasted to confirm what you already know (if you pair the first round by Elo). When such a seeding
is used, there is an alternative scheme, which I used in my broadcasted Xiangqi engine tourney:

Sub-divide the field into qualifier-groups (of 6 players each, in this case, as there is time for 5 rounds), based on prior Elo. Play round-robins in these groups. Then make a new sub-division, where you join thetopfinishers in one group with the bottom finishers of the next higher group, to make the finals groups. You could even have the number one and last of the qualifier groups skip a group (i.e. exchange the number last of a qualifier with the number 1 two groups lower before you join lower half and upper half of the qualifier groups to make finals groups). This works better with groups of 8 than with groups of 6, though.

[Dave Mitchell]

I believe you are in the wrong forum.

this is for people who are interested in technical discussions about chess programming.

there is another section for tournament and match results discussions.

Yes, quite right.

This is however, the forum Michael chose to put his tournament questions into. Obviously, the poll needs to be in the same forum as the discussion.