LACCTiC Returns

I hadn't seen that site until today - well done, keep up the good work.

It is well done. Could a team ranking be done based on top 5 runner ranking totals? The average time ranking can be misleading and too heavily weight 1 or 2 fast runners. Just a suggestion to maybe improve this great tool. BTW it is Sunday morning and the data already has yesterday's results in it.

The race importance seems skewed. You’ve got random invitationals scoring higher than regional championships in terms of importance.

Besides that, job well done.

Bijan, may I ask why the 2nd #, the current estimate of a runner's 5k ability, is different from the actual race performance? For instance, for nationals preview, here are the top results:

Patel and those below are ranked at 15:01 and below for this performance, yet the estimate of current level is quite different, presumably based on previous performances, so that Patel is said to be CURRENTLY 11 seconds worse at 5k than his teammate, whom he beat yesterday by 4 seconds. The third place guy ran 2 seconds slower than the 2nd place guy but is said to be 35 seconds slower for current race shape (15:22 to 14:47). Maybe the second stat should just be redescribed as something like season- or college career- (depending on what you're including) average 5000m equivalent performance, since obviously the performance from that day is going to tell us more about their current fitness level (unless they tempo it or, say, are well ahead of the rest of the field) than their average times.

1 3
Ishaan Patel
26:04
15:01
14:58 (14:57 - 15:01)
JR
North Central (Ill.)
2 1
Nathan Halbrader
26:08
15:03
14:47 (14:45 - 14:50)
SO
North Central (Ill.)
3 47
Josh Koehnen
26:10
15:04
15:22 (15:06 - 15:39)
SO
Wis.-La Crosse
4 3
Nigel Neibel
26:11
15:04
14:59 (14:54 - 15:06)
JR
North Central (Ill.)
5 4
Ben Stangel
26:26
15:13
15:11 (15:10 - 15:13)
SO
Wis.-La Crosse

Just dropping in to say I love your site. Thanks for putting up a detailed faq.

The team average time of the top 5 is not intended to be a ranking. Instead you should use the simulator. To simulate a specific field.

The current estimate of the 5k ability is the average of the runners top two performances. If a previous season or track performance helps their score those are included as well (they are forced to be included if the runner has not run twice yet this season).

The actual race score could be better than the estimate of their 5k ability because it is an average. If you beat me last weekend, but I beat you this weekend, who is better? Thats what this captures.

What does the system use to assess deltas in difficulty between course X, Y or Z?

Essentially, if race A and race B have some shared runners and they average 25:30 at race A and 25:00 at race B, we can infer that race B was 30 seconds faster than race A. The system relies on this intuition (but does it in terms of percentages, since a 50:00 8ker will probably slow down by a minute, not 30 seconds). Linking these together allows us to get comparisons for races that have no shared runners. More details are on the website FAQs.

Do you have a specific example for me to check? It's worth noting that the race importance is normalized for the season so far, so the scores you see for the current season will go down as more important races come in.

Generally, some invites like Nuttycombe etc are actually very important races, while regional races can be less predictive because people are running to qualify. The importance system currently relies on the idea that races with lots of runners who don't usually race each other have higher stakes. I tried many other systems for this, including a noise-whitening process and weighting races by their predictive power, but they were all numerically unstable.

What year and season did you start to include super shoes in the conversions, or notice a difference in times? How do super shoes factor into cross country times of track 5k times now?

thanks

I never actively did anything for super shoes. The way I convert to track is by including track PRs as another race and adjust everything to that. When I include "Track PRs," I can choose the season before or the season after. Those conversions should be the same, and they are for all years except the 2019-2020 season. All track times before the 2019 track season were around 1.5% slower than in the 2020 season and beyond.

I used to have both a "modern TiC" and a "season TiC", but the clutter was confusing people, so right now I just convert to whatever the previous track season was. That means 2019 XC season and before will be around 1.5% slower.

Just curious as to why some of the races suddenly had a way slower LACCTiC value over the weekend.

For instance, sub 24:00 at the UC Riverside course was worth 14:10-14:15 or so and not its worth only 14:25-14:30. That seems like a massive adjustment all of a sudden

There’s a notice on the front page about the shift.

After a weekend of explaining a nuanced interpretation of the ratings, I decided to shift them to be actual track equivalents. Previously they only indicated “this is an average performance for someone with this PR.” Now it’s an unbiased estimator for a track equivalent.

Bijan, thank you for your work.
GSC Cross Country Championships results show that all male and female TiC are UNSTABLE.

The Northwest Missouri St. Women are in your system as Missouri Western. Those are two different Universities.

How do you deal with the fact that many runners have a track 5k or 10k that’s not a good predictor of their ability? 5ks and 10k are more likely to go tactical or have an “off” race, and more importantly a lot of athletes will only race the event once or twice, often in a tactical race at the conference meet, even if they’re focusing on it. Anecdotally, I don’t know a single runner who has a lacctic rating this season that’s better than their best track PR at 5k/10k. While I agree with the change in principle, it seems like an over-adjustment that is assuming away a lot of things about what a track PR actually says about a runner’s ability.

You have missed the Parker Valby discussions....which is probably a good thing. Track pr of 15:20. 2 races this fall - 15:03 and 15:00 ratings I believe

Most people have at least one 5k or 10k PR that is representative. I take whichever is better and that forms a “race” of PRs. I can adjust everything according to those terms.

Of course, if one race is a list of track PRs, that race is going to be super fast because nobody is having bad days. This is exactly the difference between a PR performance and an average performance. We want to add an adjustment factor so that, on average, someones BEST (not a Fager) performance in XC is around equal to their best performance in track.

One final point is that in the middle of the season a lot of people will not have run their “best” yet. To make this adjustment more stable I used historical data.