Bob,
I think the fundamental disagreement between you and me is whether games against significantly stronger or weaker opponents are meaningful. I think they are, and reduce randomness in results, and you think they aren't or increase randomness. My good buddy and super-mathematician dude agrees with you much to my annoyance (why do smart people always disagree with me...hmmm) but I stubbornly persevere in my position. Given our different premise sets, there is not much to debate.
Now others may disagree on different issues (how long before you stop playing your peers for example) but in my mind that just stems from whether games against weaker or stronger opponents are meaningful. If they are, it becomes harder to predict how many games against "peers" you have because of possible early round upsets. If that doesn't happen all your arguments about round length make perfect sense.
Anyway, congrats on Crafty's improvements. The difference is quite noticeable. Are your finger notes on ICC left over from ACCA tourney, or are you really off your laptop and onto 8 processors on ICC now?
-Sam
CCT 2009 - Input Needed
Moderators: hgm, Rebel, chrisw
-
- Posts: 6401
- Joined: Thu Mar 09, 2006 8:30 pm
- Location: Chicago, Illinois, USA
Re: CCT 2009 - Input Needed
You argument is not compatible with the data in the link provided (CCT8).bob wrote:Now you begin to see my point. What happens after you play "your own half" (or more likely own 1/4) and there are more rounds to go. We usually have 6-8-10 programs that realistically can win a CCT (or ACCA) event. With 11 rounds, you definitely will run out of your group. In fact, with 8 top-level programs and 24 others, after the first 2-3 rounds, where the strong programs have only played weaker programs due to swiss pairings, now the top group runs through each other, and there are rounds left. With 32 players, it takes two rounds to recognize the top 1/4 of the field. But with seedings we already had a good idea. Once those play, the remainder of the games are just games.michiguel wrote:That generally never happens in swiss tournaments, where you have a more bell-like curve. See herebob wrote:My reasoning is valid when you have an event with two "classes" of players, one that is significantly better than the other. I'm not sure why you don't understand the point. We have seen it in _every_ tournament that was "too long". In every CCT we have played, we have a solid line of demarcation somewhere around the middle or higher. and there is usually a sharp line in those two groups to divide them into an upper and lower groups.michiguel wrote:Your reasoning is only valid in a tournament with 2 GMs and a bunch of patzers. A Swiss system does not work like playoffs if there are enough good competitors at the top. For instance, if there are 7-8 top players, the last 5 rounds become a quasi round robin among them. Many important swiss tournaments are played with a large excess over the log2(N), for instance, the World Junior championships (13 rounds) and it works just fine. If you are saying that there is only 3-4 engines worthy of winning all, that is another matter (but I would disagree). In addition, you are only looking at the top, but the middle of the pack benefits from more games. When you say that after games 5-6 everything is over, you are ignoring the interest of the ~70% of the participants who lay at the middle.bob wrote:And most would choose 11 not understandint the purpose of the Swiss system Too many rounds makes the last rounds pointless and we could just as well play random games on ICC for the last rounds. If you want to remove the luck (which is there regardless of what some might suggest) then a double-RR is the way to go. Otherwise there is a luck factor for who plays who and when, which can give someone a color advantage or disadvantage in key games...swami wrote:Majority vote is Ok, but many people would obviously choose 11 rounds, that too without knowing the final number of participants.BubbaTough wrote:Possible. Another option would be to go with majority vote. So far that is 1 for minimum rounds, and everyone else for maximum rounds.swami wrote: Perhaps we could settle for 9 rounds, Is everybody Ok with it?
or how about we use round formulas? For example, when we have
>50 participants 11 rounds
>30 participants, 9 rounds
<30 participants, 7 rounds
something like that?
-Sam
So, I'd prefer a fixed number or conditional cases to guarantee the number depending on total entrants.
http://cctchess.com/cct8/index.html
There is no solid demarcation of any kind.
Even if you have two halves, each player keeps playing in its own half.
In that CCT tournament (9 rounds), you can easily play two more interesting rounds. (Rybka here is out in its own league).
Look at the cross table. As an example, after 9 rounds, Zappa #2 did not play Hiarcs, Thinker, Diep, Crafty, Glaurung etc. Same for the rest of the top 10 engines. There is plenty of room fore more interesting games. Crafty was exactly in the middle of the pack with 50%. You cannot tell me that there is a selected elite of programs and the rest are patzers.
Miguel
Miguel
You need go no farther back than the last online tournament we had where we played 10+ rounds, or the past WCCC events that had enough participants to not create a round-robin...
I would go for the top four and play 4 games, where both sides play black and then white with same opponent. Or 6 games with a double-RR after a 6 round qualifier...
If you want to make it interesting to attract spectators, you have to increase the luck factor with a mini play-off system. For instance, after 9 rounds (regular rounds), the play-offs start like this:
#1 (white) plays #4
#2 (white) plays #3
(if draw, #1 and #2 advances)
Then, winners play each other and losers for 3rd/4th place. Always, the one that finished with better record plays white and advances with a draw (Sort of home court advantage as play-offs in the NBA etc. to reward a better regular season).
You can do the same for #5, #6, #7, #8 etc.
Miguel
Another idea is to do something like a reverse-accelerated pairing where you seed the programs, and the top group plays the bottom group in various combinations for the first half of the event, then normal swiss is used to that the top half mainly plays itself since most should have near-perfect scores, and the bottom programs play themselves. It just swaps the first half of the vent with the last half so that the tournament is won or lost in the last games...
.
-
- Posts: 2056
- Joined: Mon Mar 13, 2006 2:31 am
- Location: North Carolina, USA
Re: CCT 2009 - Input Needed
I think we can please all - we are having two tournaments.
The primary tournament has contention issues between TCs and TZs.
It is a worldwide tournament. So, lets keep the TC as is and go for
9 rounds.
Now, lets use the extra time gained from not having 11 rounds and make
the blitz tournament round robin. The blitz tournament will go fast and
one of the bots can handle the pairings. So, we could have a blitz
tournament with 40 or so rounds.
Or, we can add rounds to the long TC tournament and drop the blitz.
The primary tournament has contention issues between TCs and TZs.
It is a worldwide tournament. So, lets keep the TC as is and go for
9 rounds.
Now, lets use the extra time gained from not having 11 rounds and make
the blitz tournament round robin. The blitz tournament will go fast and
one of the bots can handle the pairings. So, we could have a blitz
tournament with 40 or so rounds.
Or, we can add rounds to the long TC tournament and drop the blitz.