We froze our projection on Thursday evening, two days before the gun, using every Maine result through Sep 26. Then 1,341 of the Maine runners we had projected went out and raced in Belfast. The boys went close to script. The girls had a different winner. In both races, 3 of our top 5 Maine teams finished in the top 5, and the order of the whole field held up well (rank correlation 0.97 for boys and girls).
Boys: the top two exactly as called
Yarmouth won the Maine team race as projected, and Portland was second, also as projected. Below them it got interesting. Thornton Academy ran better than we had them (5th to 3rd), and Scarborough and Gorham both jumped two spots into the top 5. They tied on 284 points, and Scarborough took 4th on the sixth-runner tiebreak. Going the other way, Houlton/GHCA (our 3rd) finished 7th and Winthrop (our 4th) finished 6th.
The times were the best part. The typical boy finished 25 seconds from his projected time, better than the 33 seconds we said to expect, and with no lean either way: the middle runner was 3 seconds fast. Eight of our projected top 10 finished in the real top 10.
Girls: Hampden Academy flips the top three
We had Falmouth first by a clear margin, with Portland and Hampden Academy close behind. The same three teams made up the top three, in a different order: Hampden Academy won with 151 points, ahead of Falmouth (198) and Portland (217). The two teams that broke into the top 5 were Mt. Desert Island (our 8th, finished 4th) and Cape Elizabeth (our 7th, finished 5th). Presque Isle, our 5th, finished 10th.
Up front, the individual race went almost exactly to form: all 10 of our projected top 10 finished in the real top 10, and we had the winner. The typical girl finished 32 seconds from her projected time, right on the 33 we said beforehand, with a small lean slow (7 seconds).
Keep reading: deeper in the field, what we got wrong, and what changes, next up
Deeper in the field
The biggest team surprises came from the middle of the pack, where a team's score depends on its fifth runner and a few seconds move a team several places. Boys: Waynflete finished 13 places higher than we had them, and Carrabec and Noble finished 13 lower. Girls: Deering finished 14 places lower than projected and John Bapst 9 higher. Those are the ones we will look at first: who raced, who did not start, and how much race history each team had in our data.
What we got wrong, and what changes
Runners we barely knew. Runners with four or more races in our data were usually close; runners with one race, or only a seed time, were often a minute or more off. Their ranges should be wider and say so.
We projected some runners who never raced. About one boy in five and one girl in four that we projected did not start: late scratches from the entry list, roster runners who were never entered, and a few schools that did not bring a varsity team. The "Only runners who started" switch below shows each team's result with those runners taken out. Take them out and the team order gets closer: in the boys race, teams landing exactly where we said rise from 7 to 11 of 68 (within one place, 17 to 19); in the girls race, within one place rises from 13 to 15 of 55. Houlton/GHCA is the clearest case: without its one no-show we would have had them 7th, exactly where they finished.
Whole teams have days. Some squads ran ahead of their projections almost to a runner, others behind. Our team ranges treat every runner's day as separate; for the regionals they will allow a team to rise or fall together.
A correction. On race night we first said the likely-place ranges were far too narrow. That came from our grading tool, which checked them with an older, tighter setting than the page used. Checked properly, the ranges held for more than 8 runners in 10 who started, as designed.
The full story, with what we learned for the regionals, is on the next page.
Next up
Regionals, Saturday, Oct 24. Same method, same promise: we freeze the projection before the race and put it next to the result afterwards, misses included.