NAQT card-basd power-matching system

[rchschem]

I'm interested in using the "win and take the lower-numbered card" system NAQT used at last year's HSNCT. I know there was discussion about this on the board sometime in the last year. What I'm wondering is how complicated the pairing system is. It seems to me that it must be more than just random combinations of team numbers in rooms each round in order to guarantee that every team gets a chance.

I'm working with 19 teams in 6 prelim rounds, and only interested in getting the top 16. I know it doesn't take much work to cull that field, but I'd like to try something other than just a random draw, and I don't have the staff or time for a proper Swiss pair.

Thanks,
Eric

[btressler]

Here is one resource:

http://www.charterschool.org/clubs/acad ... sintro.php

I only have the first four rounds because the repeat matches become prevalent once you hit the round near your power of 2 teams.

We did three tournaments using the cards in Delaware this year. I've found that it works well for the first four rounds. Then I send the teams to lunch and either pair the next round by hand or use a computer. (our tournaments tend to have between 16 and 32 teams)

[rchschem]

Thanks. It's better than the kludge I was trying to generate by hand. How did you assign from round to round? Did you use any kind of algorithm I could apply or extrapolate from to populate the next 2 rounds?

Eric

[dschafer]

Effectively, the first log_2(n) rounds are just a straight, pre-planned Swiss pair; for an algorithm, recursively do a n/2 team, r-1 round Swiss pair on the odds and evens, then play 1v2, 3v4 in the round that you add; this has the added benefit of maintaining effective standing order in the cards after a given round (if a team A has a lower card than team B, team A has no fewer wins than team B).

After that, you just have to keep track of the standings and try and do a Swiss Pair as best you can; repeats of early matches are difficult to avoid (since any team could be anywhere at this point), so I just try to avoid repeating a card-vs.-card match in any given three-round stretch. The standings are also not strictly determined anymore, since you have to play teams with different wins against each other.

[btressler]

To extend what Dan said:

http://www.charterschool.org/clubs/acad ... et3940.php

I did an analysis of TJ's January tournament using this bracket. They had 40 teams, which should allow 5 "safe" rounds, since 2^5 = 32 < 40.

In their rounds 6 and 7 the repeat rate was 20%, which isn't the end of the world, but also isn't small. I imagine with a field smaller than 40, the challenge multiplies since there are fewer teams with each record.

[Howard]

Thanks. It's better than the kludge I was trying to generate by hand. How did you assign from round to round? Did you use any kind of algorithm I could apply or extrapolate from to populate the next 2 rounds?

I don't believe Bill (stat74) assigned pretournament seedings. Rather I think he assigned random numbers. One of the things that can happen if you do this is that the best (or worst) teams can get stuck playing each other. If I were to utilize this system, I'd assign pretournament seedings. While no pretournament seedings are perfect, a reasonable job of preseeding should help to minimize the chances of undesireable matchups.

[rchschem]

I'd assign pretournament seedings.

I plan to.

Thanks for the information!

[btressler]

I done it both ways.

I've decided I like the random pairings better. I do make a minimum effort to make sure that a school's teams don't play each other in round one. But other than that I figure that any "hard" or "easy" in the early rounds should balance out.

[AKKOLADE]

I've considered introducing something similar to this for the WV State Tournament. I take it that the Charter system is well received?

[dschafer]

How do people who have designed card-swap systems deal with rounds after the W-L records stop being completely determined? As an example, in a 6-team tournament, there will be three 1-0 teams and three 0-1 teams after the first round. In the second round, some 1-0 team has to play a 0-1 team, but whoever gets the winner card could be either 2-0 or 1-1, depending on who won. In a normal Swiss pair, this is easy to accommodate, since the schedule is generated after the standings are known; in a card swap system, how do you deal with this team?

[btressler]

I assume the higher record is going to win the round, and proceed to the next pairing on that assumption.

That assumption's veracity varies. On Saturday, Charter A (1-0) did indeed beat Concord A (0-1). At the December tournament, Walter Johnson (2-1) beat Moravian A (3-0), causing WJ to then have to play Kellenberg A (4-0).

The latter example convinced me that the errors will multiply as the rounds proceed when you do not have a number based on a power of two. I usually use the computer or other means after round 4.

At nationals this works really well because you have five blocks of 32 teams. But in round six, some 5-0 will have to play a 4-1.

And "Charter System" may not be accurate. The idea comes from NAQT. The first time I saw the cards given out randomly was TJ (and I decided I liked that).

[dschafer]

Yeah, having the number of teams be a multiple of a large power of two is rather helpful for these schedules to work out nicely.

[BobGHHS]

Matt Bruce actually sent me a 32-team excel file two years ago that does all the pairings for you that I use for Ohio NAQT States. The only thing I had to make up were the cards, which was the easy part.

You punch in the room numbers you will be using and it does the rest of the work, even giving you read-outs of which teams are in what rooms during which round, what should be on the cards, etc.

If anyone would like to see a copy of it, shoot me an email at theamazingbobbo <at> aol <dot> com and ill send it over.

[Howard]

I take it that the Charter system is well received?

The card system seems to work well if, as Bill says, the number of teams is greater than or equal than two raised to the number of rounds.

[Whiter Hydra]

Ideally, you'll want the number of teams to be greater than or equal to the number of rounds. In addition, it's much better if the number of teams are a multiple of a nice power of two (16, 32, 48, etc), since the number of mis-matched games go down considerably. If the number of teams are not divisible by four, then you'll have a 1-0 team play a 0-1 team.

I only have the first four rounds because the repeat matches become prevalent once you hit the round near your power of 2 teams.

You can generally go one or two more rounds past, but after that things break down very quickly. Plus, you'll have the same two good teams play each other over and over again.

I've decided I like the random pairings better. I do make a minimum effort to make sure that a school's teams don't play each other in round one. But other than that I figure that any "hard" or "easy" in the early rounds should balance out.

My thought on this is to spread the talent out such that the best (and worst) teams won't end up playing each other until the later rounds. That way, you only have to look at the last few rounds when trying to avoid repeats. We didn't do that in our tournament and ended up with a 20% repeat rate, which was too high for my liking.

How do people who have designed card-swap systems deal with rounds after the W-L records stop being completely determined? As an example, in a 6-team tournament, there will be three 1-0 teams and three 0-1 teams after the first round. In the second round, some 1-0 team has to play a 0-1 team, but whoever gets the winner card could be either 2-0 or 1-1, depending on who won. In a normal Swiss pair, this is easy to accommodate, since the schedule is generated after the standings are known; in a card swap system, how do you deal with this team?

What I did was I just gave each team half a win. So if a 1-0 team played an 0-1 team, the winner would be 1.5-.5 and the loser .5-1.5.

I have a feeling that no one understood what I said.

[dschafer]

Nah, I follow... half a win seems to work pretty well, but won't that lead to those teams being in the same Swiss Pair group, so they'll play each other over and over?

[Whiter Hydra]

Not really. When I made my brackets, I didn't use any Swiss Pair groups. I just had teams play other teams with as close of a record as I could while trying to avoid repeats.

[rchschem]

What I ended up doing was dividing the field into 4 subgroups of about equal size (frakkin' 19 teams...) and then having each subgroup play each other over the course of 4 rounds. This way I felt there were opportunities for high seeded (weak) teams to have a chance to take a low number, while at the same time low seeded teams (strong) had to play good teams as well. I then paired within groups, usually high versus low, but I mixed it up.

I haven't looked at the results yet, but there were a couple of glaring errors (not necessarily flaws, though they exist, I'm sure). First, I accidentally made the 10 seed play up all day and they got hosed and went 1-5. Then I accidentally paired a repeat across rounds 1 and 2, which was just dumb.

The teams were generally satisfied, and it seemed to do a good job of presenting adequate challenges for teams while not presenting anyone with a schedule of death or a plate of cupcakes.

Thanks for all the advice. Next time there will be a proper Swiss-pair, but for the purposes of having a tournament that would basically run itself this worked.

Eric