Swim_Bike_Rum wrote:
SteveCoz wrote:
I SEEMS longer races and races with more races get scored higher
I think that longer/tougher races DO result in higher top end scores, and there is a pretty logical reason for this. I'm going to make a few assumptions:
- Long course races and races in tough conditions (heat comes to mind) are likely to have a disproportionate number of blowups than drastic out-performances. Certainly much more than a sprint. The distribution of PAR times is going to tend to be more of a lognormal distribution (fat tail on the lower end), rather than a more normal distribution (relatively balanced around the mean, or bell shaped). I have not done any statistical analysis to prove this but I'd be willing to bet that it is in fact the case. There are just so many things that can go wrong in a 70.3 or IM, especially in tough conditions, that to cause you to finish an hour slower than projections and really virtually nothing that will allow you to finish an hour faster.
- Given a more lognormal long course par time distribution, and a more bell shaped sprint par time distribution, there will be differences in the relative position of the race par time.
- When the top and bottom 20% of PAR TIMES (not race times) are discarded before averaging individual par times to come up with a race par time, the average is likely to include a significantly higher proportion of relatively poor performances than good performances, resulting in lower race par time relative to the top finishers, and what seem like crazy high scores at the pointy end.
So I think the takeaway is that if you are aiming for a high USAT ranking you should aim to perform well in long/tough races where there is the potential for lots of other people blow up.
Your initial assumption is incorrect. It's not the top and bottom par times that are removed. It is the top and bottom most variable times that are removed. So if someone scored an average of 95 the previous year and scored a 75 on race day, they would be removed from the pool as an outlier because they would be in the bottom 10% of deviations. Same for the person who averaged 70 the previous year and knocked out a 91 on race day, they would be removed as well. Your bullet points are exactly why USAT does that. It takes out people that are just "training" through a race, or having a really bad day or having a really abnormally good day for example, as those aren't good data points to base an average on.
As stated previously, the par time is based on the majority. Harsh conditions or long races push the thick part of the bell curve back and affect slower races much more than faster racers so the scores get skewed a bit. A full IM on a tough course on a tough day will give the leaders much higher scores than a sprint on a nice day.
The initial assumption is correct I believe. I think we are saying the same thing....maybe different nomenclature.....the top 20% and bottom 20% of calculated par times for each individual are removed and the rest are averaged to get a race par time. A very high or low calculated par time indicates high variability and is removed as you correctly state. So if my ranking is 80 and I finish a race in 100 minutes the calculated individual par time is 100*0.8=80 minutes...if your ranking is 90 and you finish in 90 minutes the calculated individual par time is 90*0.9=81 minutes. If another racer with a 90 ranking has 3 flats and finishes in 120 minutes that calculated individual par time is 120*0.9=108 minutes. That 108 minutes calculated individual par time is likely in the top 20% and will be thrown out. The middle 60% of individual calculated par times are averaged to get the race par time. You are also on point in that the "thick part of the bell curve" as you say is well to the left of the median in tough races, resulting in high scores for those above the median