Lacctic is bad

I disagree on "so off on everything". "everything" by definition means every single thing. I think the vast majority of relative rankings are very, very good. I can see a few individuals rankings which seem off some, but given the available data, and assessing all data consistently the results for the free toll are good. If I had one wish it would be the ability to override the calculated rating for a runner as the logic seems to occasionally miss big changes. What tool would you prefer with wich to assess runners/teams who have not competed?

i think it does well enough with rankings, but the adjusted times or whatever they call it isn't really worth much.

I like it a lot. It is directional... the 5k times may be a bit off but the relative performance abd rankings do a good job overall. For some top runners (like Tuohy) it is probably under estimating due to effort (this happened last year too) but overall it is pretty good.

Just saying it’s bad doesn’t really help me make it better. If you can direct me to what you believe to be bad, I can look into why it’s making mistakes and tweak the algorithm. Idk why you’d go through the effort of making this thread just to hate on the website - nobody is making you use it.

The only reason I do 5k times is so that you can understand the relative differences. Imho I don’t really know how to think about what the difference between a 160 and 180 speed rating is. But I do know the difference between 15:00 and 15:30. I also think the 5k equivalents do pretty well at predicting early season indoor times from what I’ve seen.

I’ve been planning on adding an override feature in the simulator for a while - just haven’t gotten around to it. Managing tfrrs data errors is more work than people realize. That, and I do have a real job as well…

Here are the “Season 2022 LACCTIC Fitness Estimates” of some athletes prior to the December 2022BU Indoor, and what they actually ran there:

Athlete………LACCTIC……….BU
Tuohy…………..15:19………….15:16
Olemomoi…….15:30…………15:18
Chelengat…….15:28…………15:18
Cook…………….15:38…………15:24
Chmiel………….15:29…………15:28
Baran……………15:56…………15:34
Stearns…………15:36…………15:34
Seymour………..15:51…………15:36

It estimated Tuohy, Chmiel and Stearns very well.

Terrible thread, Lacctic is awesome. Thank you.

Just to chyme in- I think Lacctic Rating system is an awesome resource. I feel the thing it does VERY well, is comparing XC course to XC course, so you can track what is a better or worse for an athlete. I use it for coaching, in minimal ways (just evaluating how hard a course was compared to another course).

I do think it's not super accurate at predicting races, but it is as accurate as you can be given humans are running the races. The 5k track times do seem much faster than the XC TIC ratings though.. For example, take how many people run under 14min for 5k on the track, compared to how many are rated under 14min or have a sub 14min TIC performance at all: The difference appears significant & that is not factoring in that, on average, you'd assume someone might have gained 1-2% fitness from Spring to the next Fall.

Some intangibles that I'm not sure you can factor into race predictions would be depth on a team- For example: you might have one team with 5 studs, who have all have a great TIC rating based on 1-2 performances, and another team with 7+.. In an actual race, I think the odds of 1/5 people being off is likely and more points are accumulated at the 4/5 spot than the 1/2, so you'd really see a lot of change in outcome based on this- just a thought.

But to confirm: I do love the system and think it's amazing for comparing XC courses. I use it often. I realize there are some limitations, but that's more due to evaluating humans, which can be hit or miss on the day. Maybe the one area to look at would be the 5k times being exactly the same TIC rating- they just seem 20+ seconds too fast, or maybe the XC TIC ratings are 20s too slow? I appreciate so much all the work you put into this website though, and you are totally right- people complaining about it aren't forced to use it. It's easier for people to hate than to love.

2022

Athlete………LACCTIC……….BU
Tuohy…………..15:19………….15:16

It is awesome. I know some college coaches who look at it with their teams. Don't change a thing.

Thank you for your service, Washed-Up Grad Student

You’re right about this and I actually think it should be incorporated in the simulator. When I simulate a race I sample from a distribution of possible performances. Having no good 6th or 7th runner means the team has a lot of bad days when their 5th has a bad day. Having a good 6th or 7th generally gives the team a more robust distribution of finishes, because a lot of the bad days for the 5th man are picked up by decent or good days from the 6th.

Regarding the time difference - I actually add a 1.3% to the xc ratings (around 12 seconds for a 15:00 5k). This was the average difference between “Rating PRs” and actual 5k PRs in the old algorithm, across all runners.

The reason for this adjustment was to correct for the fact that I was treating the list of PRs as one big race. Of course, a big race where everyone runs their lifetime best is pretty unlikely. So, instead of equating xc performances to “an average performance for someone with a PR of 15:00,” they should now be closer to equivalents.

I do think this new system throws people off, though. They think that people with 15:00 PRs will spend their whole season running 15:00 equivalents when the reality is they will probably spend most of the season running 15:10s and pop one 14:55 at the end.

I said I think lacctic is good so I hope this isn't a response to me... i.am not the OP

It has some good data and a good idea to be an objective ranking. But the formula does have some flaws, like it does not account for racing at altitude, and if a course is long or short but TFRRS says it isn't (like Roy Griak is actually 8,150m but is labeled as an 8k). So the idea is great but like anything in life it is far from perfect.

I was responding to the OP - just hit reply on a random post

Go read about the algorithm. It doesn't use course distance or elevation - it looks at relative times, so all of those should be accounted for.

Might just be contributing to an echo chamber at this point, but I love the idea of Lacctic. As a stats nerd myself, I cannot imagine how difficult it is to establish an accurate rating system for NCAA XC, considering things like injury, illness, effort, and other day-to-day variation in performance on top of TFRRS data errors. Considering all of that, Lacctic does a great job and is a lot of fun to use.

I like the TiC rating as a 5k equivalent—much easier to interpret than a speed rating. The problem lies with people interpreting them at date track 5k fitness, not considering that some runners are better at XC relative to track and vice-versa. I doubt Parker Valby could run 14:49 right now, but looking at who she beat and by how much at SECs, it is not unreasonable to claim she could challenge a 14:49 runner with average XC strength.

One thing I’m curious about—it seems like the simulations are based on the individual TiC confidence intervals, which are roughly the same width (~16 seconds for the top runners).

1. Does the simulation account for some runners being more volatile than others? For example, a team might rather have a reliable 13:55-14:05 equivalent runner than a 13:45-14:15 runner if the race is thinner near the front, since the second runner will cost more points on a bad day than he will contribute on a good day. That being said, I can imagine it would be difficult to automatically distinguish a runner’s off-days from races treated as
workouts.

2. Does the simulation assume a uniform distribution in the TiC interval, or one that puts a greater likelihood on performances closer to the average?

Nope. It accounts for ALL factors. Course length, altitude, weather, etc.

It accounts for them by normalizing all races to the same scale. Those factors affect all the runners in the race so they get washed out in the normalization process which processes individual results across all races.

A simplified example of this would be for a race at altitude where the average runner is 30 seconds slower than average then their rating for the race is adjusted by those 30 seconds. It does not matter whether that 30 seconds was due to altitude, course length, weather, etc.

The guy that created it told us he has zero way to adjust for altitude so he acknowledged that it isn't perfect

I would not say it is bad. It is tool. The issue is that unlike a high school speed rating there just are not enough data points. With the high school speed rating you have courses that are run year after year, allowing for a course par. Plus the athletes run more races, as opposed to just 2-3, and the races tend not to be tactical. All of that provides a lot more information. With Lacctic an unvetted course like the SEC course last year or likely the regional course this year will throw it off. There is just not enough information to go on.

Yes it does. But not in the way that your brain expects. It doesn't matter if at altitude of if the course is 7K or 8K or 9K. It doesn't matter if the course is rock hard or very muddy.