The whole design of those percentages seems way off.
Start with - history seems to show a 20% chance that any given team will implode and just have a bad day. There’s just a range of how teams will do.
I think it starts with the accuracy of the TiC Rating range, and then the percentages are crunched from there.
For example, Mia Barnett’s Rating range is 15:49-16:09.
Mia Barnett Oregon (SR) LACCTiC Rating: 15:58(15:49 - 16:09) Status: Inactive 2024 Season Weight: 100% Season Fitness Estimate: 15:58
Having no results yet this season, 100% of this athlete’s Rating is based upon her 2024 season.
So sure, a team could implode. The math in effect, takes into account depth, the percentages give an idea how much a team can withstand hiccups. Abnormal TiC performances will stand out.
The whole design of those percentages seems way off.
Start with - history seems to show a 20% chance that any given team will implode and just have a bad day. There’s just a range of how teams will do.
The current simulation probabilities just come from individual variance. Deep teams are rewarded here, because a single person’s implosion is covered by a strong 6th runner.
This doesn’t take into account the probability that the whole team implodes (which is a breakdown of the independence in performance). I haven’t added this because I felt like it was adding unnecessary noise (since the same blowup probability would be applied to each team). I don’t think there is enough data to estimate per-team variance with any statistical confidence. I could be wrong though. Maybe I’ll test it out in the off season.
The whole design of those percentages seems way off.
Start with - history seems to show a 20% chance that any given team will implode and just have a bad day. There’s just a range of how teams will do.
The current simulation probabilities just come from individual variance. Deep teams are rewarded here, because a single person’s implosion is covered by a strong 6th runner.
This doesn’t take into account the probability that the whole team implodes (which is a breakdown of the independence in performance). I haven’t added this because I felt like it was adding unnecessary noise (since the same blowup probability would be applied to each team). I don’t think there is enough data to estimate per-team variance with any statistical confidence. I could be wrong though. Maybe I’ll test it out in the off season.
You could consider adding a correlation factor between teammates, as sometimes the success or failure can be continuous. In case you don’t think it’s necessary, just set the factor as zero.
The current simulation probabilities just come from individual variance. Deep teams are rewarded here, because a single person’s implosion is covered by a strong 6th runner.
This doesn’t take into account the probability that the whole team implodes (which is a breakdown of the independence in performance). I haven’t added this because I felt like it was adding unnecessary noise (since the same blowup probability would be applied to each team). I don’t think there is enough data to estimate per-team variance with any statistical confidence. I could be wrong though. Maybe I’ll test it out in the off season.
You could consider adding a correlation factor between teammates, as sometimes the success or failure can be continuous. In case you don’t think it’s necessary, just set the factor as zero.
Yeah exactly. You can sum/average the difference between Lacctic rating and TiC performance for an entire team, then calculate this variance per team. Since the variance of each individual is known, you can back out the team-level variance.
LACCTiC very well may award its first 14:52 (or better) TiC for a cross country performance after it crunches the data from this weekend’s Conference results. I say this after a quick check of the following numbers:
The numbers in parentheses indicate Nuttycombe ran greater than 25seconds slower (approximately) than Rim Rock; this implies Jane might have gone well under 19min at Nutty (maybe 18:55), more than 30sec faster than Hartman’s winning time of 19:30. Hartman was awarded a TiC of 15:16 for that performance.
1 G. Hartman 15:04(14:56 - 15:14) NC State 2 S. Kennedy 15:21(15:13 - 15:31)Stanford 3 S. Jelego 15:22(15:14 - 15:32) Clemson 4 Amy Bunnage 15:25(15:16 - 15:35) Stanford 5 Sjoberg 15:27(15:19 - 15:37) North Carolina 6 Hannah Gapes 15:28(15:19 - 15:37) NC State 7 Napoleon 15:28(15:20 - 15:38) NC State 8 A. Perez 15:30(15:21 - 15:39) Wake Forest 9 B.-Dalton 15:33(15:24 - 15:43) Notre Dame 10 Michalak 15:38(15:29 - 15:48) NC State 11 A. Aramini 15:41(15:32 - 15:51) Notre Dame 12 Chisholm 15:43(15:34 - 15:53) Notre Dame 13 Blade 15:45(15:36 - 15:55) Florida State 14 Brown 15:47(15:38 - 15:57) North Carolina 15 Kate Putman 15:50(15:41 - 16:00) NC State 16 Schipperen 15:55(15:46 - 16:05) FState 17 Rose Mburu 15:55(15:46 - 16:06) SMU 18 Engelhardt 15:56(15:47 - 16:06) NC State 19 Olson 15:57(15:47 - 16:07) Notre Dame 20 Moss 15:59(15:50 - 16:10) Duke
We saw a significany jump today for Napoleon and Bonnor-Dalton to edge out Jelego. Bonnor-Dalton’s last k, in 3min, quite likely was the fastest last k from all top contenders in all the conference meets.
