Wednesday, November 21, 2012

State Meet Projections by Sstoz Tes

27 comments:

Andrew said...

woah, Hagen Reedy beating Sarah Baxter...

but thanks for doing all of this Sstoz! it definitely helps

Anonymous said...

i disagree with so much of this i dont even know where to start. what are these based off of?

Albert Caruana said...

Section finals I believe.

Anonymous said...

yeah NCS teams really got shafted. The course was absolutely drenched. But it doesn't matter. States in a couple of days anyways, we'll see what happens then.

Anonymous said...

from SStoz blog:"S.J.S. qualifiers, who typically run quite a bit faster at the state meet course will in 2012 more likely run a fair amount slower than they did at the qualifier"...

Based on what and why? Willow Hills is one of the toughest qualifying courses. Times from Willow Hills to Woodward Park average from 28-30 seconds faster.

Anonymous said...

Your site is usually worthy of a read. These projections are so far off they are laughable. I would expect more out of CCExpress than to post these projections.

Anonymous said...

A lot of these teams have already run the Clovis course this year, yet these projections don't even take these results into consideration. What ??? SJS teams all messed up.

Anonymous said...

Agreed. There is some thoughtful logic in the predictions on prepcaltrack.com, not just number-crunching.

Anonymous said...

At least for girls, the Southern Section times are way too slow - probably off by 30-40 seconds.

Sstoz Tes said...

Oh boy. I was hoping folks would read the lengthy list of caveats before chiming in. The top of each projection, for example, lists my confidence in the ratio for each section. Reedy beating Baxter is unlikely. The top d. 1 C.S. male running only 15:45 is unlikely. The left side of the curve at the C.S. qualifier is typically uncompetitive for the top individuals (in most divisions), and so they often run to qualify or use the race as a tough workout. Each section has its foibles, some of which are accounted for in the projections (when the foibles are field-wide), many of which are not.

These projections are based purely on statistical relations between courses. In that sense, there is no human involved. If someone goes easy or has a bad day at his/her state meet qualifer, these projections take no account of that.

I did attempt to take into account the rain. I am unsatisfied with the results. I think the S.S., in particular, is short-changed by my projections. I think a ratio of 1,085:1 would be more appropriate, though nothing in the numbers indicates this.

Other courses, such as Willow Hills and Toro Park, ran much faster than normal in 2012 across all divisions. This indicates in both cases that there is some independent variable driving the results. Weather is always the first suspect, but so far as I know the weather was fine for both the C.C.S. and S.J.S. finals. Toro Park has been affected by weather at least once in the past (heat, in that case), making those results difficult to use for projections.

The next suspect is the course itself. Of the 12 state meet qualifier courses used since 2005, Willow Hills has been the 2nd most variable course (Toro Park has been the most variable). In 2005, 2006 & 2008, the course ran 4,8% - 4,9% slower than the state meet (around 0:50 for a 17:00 runner), but in 2011 it ran only 1,9% slower (about 20 seconds for a 17:00 runner). Based on historical norms (the median qualifiers' times at the state meet qualifer), I expect that Willow Hills will be only 1,25% slower than Woodward Park this year (this is around 0:13 for a 17:00 runner). I don't think S.J.S. runners will run slower at the state meet. That was a mis-reading and mis-statement on my part.

Inexplicably out-of-norm times at a given course could be explained by the level of quality of the runners themselves. This could easily hold for a set of qualifiers in one race and would not be eyebrow raising if it happened in a couple of races, but it is highly unlikely that this would hold across both sexes and all divisions. Even the much-touted improvement in quality over the past 13 years (using 1999 as the nadir) has been uneven between divisions and the sexes.

This year's projections have been weakened by several factors. The rain, of course (I'm not sure I'm going to even try to project-out the l.A.C.S.), but also the out-of-norm performances in the C.C.S. and S.J.S. Even the N.S., which should have been affected by weather (a hail storm and subsequently slippery course), is in its way anomalous -- it stuck to historical norms despite the conditions.

Humans are variable (talent, training, psychology and all the rest), as are the things they create (courses) and the climes in which they perform (weather). Statistics are a blunt instrument by which to measure those variables. The more specific the results need to be for a given data point, the less useful statistics are. This is why one uses confidence intervals (even/especially for things like election polling). Over the whole of the state meet, I think my projections will hold up fairly well, but for a given team, much less a given individual, they will provide only a general guideline. Team-times, which have a 61% accuracy rate (by far the highest of any measure and, when incorrect, are typically off by + or - 1 place), have tightened up in recent years, making even slight under-performance by a runner (particularly the 3rd or 4th on a team) more consequential than in past years.

Andrew said...

I'd like to see all of the haters project out all the races and see what they come up with. What else would you base the projections off of Anonymous 1:23 PM?

Prepcaltrack takes forever to do anything. They've only got 3 previews for the state meet done, and while they are well thought out, no one can say they are perfect. There is 7 to go and only 2.5 days before the meet. At this rate, we'll get the D5 previews before Footlocker West.

So thank you, Sstoz, for putting in the work for these numbers. No, they are not perfect and will have plenty of disagreement, but if you're going to criticize him, at least have an argument of why and have something substantial to back it up.

Anonymous said...

Who cares. It's all about the race!

Baxter is the top girl no doubt but Reedy is on her home course that she practices on weekly and if she doesn't watch out could get her. But they're in different divisions... So we never really have a true state champion.

Maybe one day we'll have an open division for the best to race the best?

Greg Beal said...

I tried to begin this discussion last year when CCE linked to your section/state ratios but you nodded at my comment and then didn't follow up.

I believe you should look at the boys and girls separately on each course and see what the numbers look like. And possibly look at time zones (15-16 minutes, 16-17. 17-18, etc.) regardless of sex. Are the conversion ratios the same regardless of sex and race time?

The other factor to consider is what happens to slower runners on tougher courses when compared to Woodward. For example, at Mt. Sac the traditional conversion for boys is about 18-19 seconds and for girls about 23-24. These tend to be close for the better runners but are way off for the girls deeper into the field. The slower a girl runs at Mt. Sac the closer she will be to her time at Woodward. I suspect the same may be true for other tough, slow sectional courses. It's this phenomenom that has, at least in part, caused your Mt. Sac to state conversion to be far lower for the faster runners than SS analysts expect.

Having spent many hours working initially on SS D2 girls prelims and finals and then on State D2, I have an idea how much time is involved in generating 10 race projections. While relying solely on sectional times will save time, I think a cross check of selected runners in a field against other races might have suggested how accurate your problematic sectional conversions were. For instance, you might have looked at the Great Oak and Simi Valley girls at the Mt. Sac Invitational to see whether your finals rain course conversions were near the expected conversions from the invitational to state. Similar comparisons for girls in rain-affected races might also have pointed you in a particular direction.

In any case, thanks for all the stats and projections you've assembled. They've been interesting and helpful (even when I disagree with some results).

Anonymous said...

Carlmont will not finish second. They had the best race they have ever had at CCS Finals and it will be a long shot for them to get on the podium.

Anonymous said...

Yeah i don't even know what to say after looking at these....

Anonymous said...

Look if history proves true again Bellarmine tops Carlmont at State. I don't know if they don't like the course, are just unfamilure with the course or get flustered not being up front in a stacked field but state never has been their thing. Running ccs with 2 good teams is much different than running state with 23 plus talented individuals. Best of luck at state as only time will tell.

Anonymous said...

Btw, the SJS section meet ran a shortened course this year--the start line and finish lines were pushed up. Best guess is that the course was 75m shorter than it usually is. Surprised that no one has mentioned this before.

Anonymous said...

Full disclosure, I know very little about NCS, NC or SJS sections or their courses, and much more familiar with SS and their courses. Your statistical analysis is mind blowing and great for stimulating discussion. However, trying to predict times for SS teams at Woodward Park based on the SS finals at the MT. sac course is an exercise in futility. I believe a much better predictor would be the times those individuals/teams ran at SS prelims or at the MT. Sac Invite. As an example both Redondo and Dos Pueblos ran at both the Clovis invite and the Mt. Sac Invites, both their times were well over two minutes faster than you project. Individually your projections show the first DP runner at 15:40+ when in fact he's already run 15:09 at the Clovis invite and 14:34 at The MT. sac Invite on the "real" MT Sac course. The same is true for just about every runner on these two teams as well as Westlake, Ayala, etc. Having said all that, while the times may be off, I think the actual order of finish is fairly accurate.

Anonymous said...

I meant to say an exercise in futility based on the Mt Sac "rain" course.

hank said...

I was going to post my State Meet projections based on Section Meet results and XCStats conversion formulas but then my computer blew up last night and now I'm just trying to save the data after TG passes. I too agree that the SS rain course made it tough to come up with a comparison (before my computer blew it had Estifanos finishing in 32nd place in the D1 race). So, the SS was going to fare really well (taking the all the top 10 spots I believe). So Section meets (and their conversions) really took a beating this year (LA Section was tough with TWO sites). These "predictions" are all for fun and to fill the 7 days with no XC racing. I'm going to wait til after States to post my "predictions" and I hope that I do better than most everyone else.

hank

Anonymous said...

The SJS Section meet was NOT shortened this year. It was setup exactly as last year and previous years. I have the pictures to prove it.

Anonymous said...

To anonymous at 8:28am: The finish line was moved up 50-60m and the start line another 15-20m. The timing company said so as have several coaches. The normal finish line is by the discus/hammer cage and the start line comes off of the manhole cover about halfway down the field. Unlike previous years, it is also my understanding that the course was not wheel measured--something George Paddeck usually did. What does it all mean? Everyone still ran the same course in the same conditions. Just don't use the marks for the All-Time list without an asterisk.

Jack Powless said...

while it is true the sjs section finals were only very slightly shortened, this race has been set up the same way the last three years. The finish needed to be pushed forward to accomodate the timing mats. All other previous races on this course are slightly longer, but all have been consistant over the past three years and will continue to be the same.

Anonymous said...

Isn't Kimberly Avalos, D5 girls, from Saint Joseph Notre Dame not Saint Joseph (Santa Maria)? It probably doesn't make a difference, just curious.

Anonymous said...

Thank you Jack Powless for clearing this up. I have pictures from the previous years showing the exact setup at Willow Hills over the past years.

Marty Beene said...

I agree with a previous post that this is great to stimulate discussion. I looked at the actual D2 boys results compared to the predictions, and was pleased to see that Alameda finished the most places higher than the prediction, at 7 spots higher. Our friends at Maria Carillo tied with Redondo for the next best, at 6 spots higher. The biggest differences the other way were Del Campo at 9 places lower and Whitney and Jesuit at 7 places lower. Fourteen of the teams were within three places of the predictions.

Nils said...

Anonymous 1:12, yes Kimberly Avalos D5 Girls is from Saint Joseph Notre Dame, of Alameda, CA.

Popular Posts