After looking at the Berlin Bear Den Dash results (also an incredibly fast course) I have a better handle on how to compare the Jerome course to other courses. My regression model excluded really big drops or jumps in season's best times by athletes because those are usually a result of getting getting sick/injured or recovering from being sick/injured. It also struggles because I only have three years of data for this course at Jerome, and two years ago it was really muddy and times were not at all like last year and this year. I also changed baseline to be the average of the four regional courses, which effectively works out to be about 5-8 seconds slower than Troy, 21-24 seconds slower than Tiffin, 5-8 seconds faster than Pickerington and 14-17 seconds slower than Boardman. This was to make it easier for people who haven't run primarily central and SW Ohio courses over the last half decade or so to interpret.
So I forced it to accept all the too fast outliers, and with the Bear Den Dash results I think I have a reasonable handle for comparison. Please note that this is going to be a little conservative in terms of predicting the effect of the course that day because of the changes. For a 17:00 athlete (baseline) Jerome was 4.02% (.0402) faster than the baseline. Meaning a 17:00 baseline is 16:19 at Jerome. For a 21:00 athlete the course was 4.37% (.0437) faster, meaning a 21:00 baseline is a 20:05 at Jerome. If you look at the data, it is clear that some athletes and teams really ran fast even relative to the fast course generally. For what it is worth the conversions for the Bear Den Dash are about .0301 and .0337 respectively. So it is also blazing fast.
I am a believer on getting accurate measurements for the course the athletes ran, because it makes it easier to judge whether a performance was good or bad for any given athlete. For the "times don't matter, places do" folks, we have to remember that athletes are not always running against common opponents, or opponents who are close to them in ability. It is completely possible to get 3rd place one week running a bad race and get 10th the next running a great race, without any of the athletes finishing immediately ahead or behind you being the same. Or to get first or last two meets in a row and run well one time and badly the other. Times help us determine whether a race was good or bad, but they need to be comparable from race to race in order to do that.
So this brings us back to the whole "it's short/it's not short/times don't matter/those times are bogus" debate. (In general, not just about the Celtic Clash in 2021.) I get about 90% of the predictive accuracy from the model I use from the first 10% of the work done, which means a simpler, and more accessible, model for comparing courses is probably a better choice. Maybe this is even something that a group of coaches could hash out together and get posted someplace to allow us to more easily compare courses and then we wouldn't need to have the arguments quite as often. Or at least we could have more nuanced and productive arguments. (Since I genuinely don't like making people angry if it isn't necessary, and this past week was exhausting, this idea may be a little self serving.)