Page 1 of 1

What if the ICT prelims were card-paired?

Posted: Sat Apr 16, 2011 5:33 pm
by The Friar
I've become fascinated lately with the card system(s) used to pair teams at HSNCT during the prelim rounds. In sort of parallel to the discussion on the high school side about the PACE NSC format, I've been thinking about ways to use such a system at other tournaments. The event that most recommends itself seems to be NAQT ICT. A variant of the card system suggested for 32 teams at PACE's own Web site would seem to preserve the most important features of ICT's current bracketed round-robin prelims while providing some other desiderata, including no games with playoff implications for only one team, probably better robustness to unavoidable seeding errors stemming from roster changes between SCT and ICT, and opportunities to collect data from actual games on the difficulty/performance difference between Division I and II packets/teams. The only trade-off seems to be that such a card-paired format would need to be yoked to a playoff format that would be one round plus perhaps one tiebreaker round longer.

Here's how the prelims would work.
  • Let all teams with a winning record (four or more wins in seven games) qualify for the playoffs.
  • In rounds 1-5, teams play other teams with equal records. By the binomial theorem, this gives one 5-0 team, one 0-5 team, five each at 4-1 and 1-4, and ten each at 3-2 and 2-3.
  • In round 6, pair 3-2 and 2-3 teams against each other. This can be done without repeating matchups and gives five 4-2, five 2-4, and ten 3-3 teams.
  • In round 7, pair 3-3 teams against each other, giving five 4-3 and five 3-4 teams. This gives a total of 16 teams left in contention after the prelims.
  • In rounds 6 and 7, pair any team with at least four wins or at least four losses -- any team already eliminated or qualified -- against a team with the same record in the other division. Play, perhaps, games for four-win teams in round 6 on the Division I packet and those in round 7 on the Division II packet, and vice versa for four-loss teams.
There would need to be no tiebreaker round after the prelims. In the playoffs,
  • Break the 16 qualified teams into four pools of 4 and play a bracketed round robin, followed by tiebreaker if needed;
  • Advance the top two from each pool together into one of two pools of four and play a bracketed round robin, counting the game from the previous phase between the two teams, followed by tiebreaker if needed;
  • Advance the top two from each pool together into one pool of four and play a bracketed round robin, counting the game from the previous phase between the two teams, followed by tiebreaker if needed;
  • Play a NAQT-style final between the top two teams.
This would result in a seven-round playoff phase where ICT currently has six, and a total of three rounds available for tiebreakers where ICT currently has two (one after prelims and one after playoffs).

Is there a reason we couldn't pull off a scheme like this, or that it would be inadvisable? Would it be valuable enough to have some actual data on how Division I teams performs on Division II packets and vice versa (say, for calibrating the D-Value conversion factor) that it would be worth trying?

Re: What if the ICT prelims were card-paired?

Posted: Sat Apr 16, 2011 5:52 pm
by Matt Weiner
Like all power matching formats, this adds massive schedule imbalances, penalizes teams for winning, magnifies any seeding errors, and produces almost meaningless results. I see absolutely no advantage to moving from the current ICT format to this one.

Re: What if the ICT prelims were card-paired?

Posted: Sat Apr 16, 2011 7:33 pm
by Whiter Hydra
The Friar wrote:This would result in a seven-round playoff phase where ICT currently has six
There are already seven rounds used in the playoffs. It's just that one of them happens during the prelims.

Re: What if the ICT prelims were card-paired?

Posted: Sat Apr 16, 2011 9:07 pm
by The Friar
Matt Weiner wrote:Like all power matching formats, this adds massive schedule imbalances, penalizes teams for winning, magnifies any seeding errors, and produces almost meaningless results. I see absolutely no advantage to moving from the current ICT format to this one.
I don't... I don't follow how seeding errors and schedule imbalances get worse rather than better under this kind of format than a bracketed round robin.

Re: What if the ICT prelims were card-paired?

Posted: Sat Apr 16, 2011 9:26 pm
by Matt Weiner
Well, the fundamental flaw in power-matching is that you punish teams who win by making them play better teams in a format where, at the end, you are ranked solely on your win-loss and statistical records. I don't see what else needs to be said about how this magnifies rather than eliminates the potential problems of seeding in bracketed tournaments.