Learning from Ottawa

The bracket from the City of Ottawa Men’s Bonspiel tourney discussed in the last post had a number of features that make it look like it would be fun to play. In this post, I’ll consider whether a truncated version of the Ottawa bracket might make a good double-elimination bracket for a 91-team tourney in some other event.

This post will show the way the Ottawa-style brackets would look, both (nearly) as drawn by the Ottawa organizers, and with a few of my own alterations to eliminate some small sources of unfairness that seem to me gratuitous. In a later post, we’ll test these designs against a more conventional bracket.

The result, I’m guessing, is that the Ottawa bracket gives up a good deal of fairness (C) in exchange for its other virtues. Chief among these virtues is the way the tourney is experienced by the vast majority of players who lose in an early round.

In an ordinary double-elimination tourney, losing in an early round is very dispiriting. You know that you have a path back to glory, but you also know that that path is a long one, and your chance of success small. But you have a different prospect in Ottawa. Here the A, B, and C rounds losers all drop into a pristine new bracket, a bracket in which you have as good a chance as anyone else to win a named trophy.

Here, again, is how the double-elimination part of the Ottawa bracket works. It begins as a pretty conventional 128 upper bracket, with mostly well-spread byes. This bracket plays out in seven rounds, identified A through G.

The A, B, and C drops from this bracket each go to a separate bracket: a 32 for the A’s, another 32 for the B’s, and a 16 for the C’s. For convenience, and to emphasize their relationship, I’ve drawn them all on the same sheet, using the same round identifiers, H through L: Ottawa345.

Finally, drops from both of these go to a single bracket for the Open Grand Aggregate trophy: OGA. This is slightly altered from the one actually used in Ottawa in that I’ve swapped back the drops for the G1 winner and loser. (This is necessary, for tedious technical reasons, because of a limitation in my simulator.) The rounds here are M through R.

I’m also going to test an alternate version of the Ottawa bracket. This differs from the other in only two respects. First, the byes will be evenly spread, breaking up the cluster of four byes that caused byes to move into later rounds. Second, I’ve interleaved the lines in rounds M, N, and O, which taker the D, E, and F drops. It’s of the essence of the design not to interleave the A’s, B’s, and C’s, so we’ll live with the fairness problems that creates. But while there may be some reason to segregate the drops to help with the demands of scheduling a curling tourney, I think that’s a gratuitous source of unfairness for current purposes. Here’s what the OGA bracket looks like with the revised architecture: OGAalt.