Simple (ahem) Race Prediction

This week's training

I'll post the numbers later, but I'm not doing what I'd like. I look at last year's calendar and see that my long runs were 35-38 miles and I was running a minute per mile faster than this year. This winter's looking long and cold - fortunately, my best years of racing came after terrible winters; unfortunately, those years were 1984, 1983 and 1978.

The lakes have frozen over, so the bald eagles have switched from eating fish to eating rodents and birds. Yesterday, what I thought was snow falling off a tree branch onto my head turned out to be a duck carcass.

Race prediction made easier

All the math in my last post doesn't help in predicting how one will do in a race. So, here's what I use.

For road and track races: What one does on an average day is 80% of what one can do. So, if one's running 10 miles in 90 minutes, as I seem to be doing lately, that suggests a 10 mile race in 72 minutes (assuming I train for a 10 mile race). For those who do a lot of speed workouts or extremely long long runs, instead of average, one should substitute "typical" - what one would do if one ran the same thing every day. One's marathon time can be predicted by the fact that one's marathon pace is about a minute per mile slower than one's average training pace (give or take 15 seconds per mile, predicting a marathon within a 15 minute range). One's 24 hour race distance is about 320 minus one's marathon time in minutes; this is accurate to about 5 miles for experienced 24 hour runners.

So, 9 minutes per mile training is a marathon at 8 min./mi. or 210 minutes. That then gives 110 miles in 24 hours.

From 5K to 50K, most runners times fit a straight line on a log-log plot of time versus distance. With 10 miles in 72 and marathon in 210, the intermediate times can be found. For the math-impaired, this is the same as the Leger nomogram that can be found in Tim Noakes' "Lore of Running." One's times in the ultramarathon range also tend to fit a straight line on a log-log plot (though not the same line as above). If one doesn't have recent times to use for prediction, the best substitute is to assume a slope for the line; using age- and sex-class records for Minnesota runners is best, but one can hope one slows like the world's best... Kouros and Godale have personal records fitting 1.216xlog(miles)=log(minutes)+constant. That slope, plus the 24 hour race point will give one the other times for various distances.

For trail races: Last time, I mentioned that the number of finishers formed a normal distribution when plotted against the log. of time. Within 0.3 standard deviations of the mean, this becomes a straight line, which simplifies things. If one plots finishing place versus log. time, one should expect a straight line somewhere.

In practice, one finds TWO. My theory here irritates people to no end. One can divide runners into four groups in any race. The first one or two runers don't fit one of the straight lines; these are ringers who didn't have to race as hard as possible in order to win - these should be running more competitive races. The next group form a straight line; these are the talented, experienced, but not usually perfectly trained runners (these are the most likely to DNF, by the way). There is then an awkward transition and another, steeper straight line which incorporates the third group, the not very talented but extremely hard working group (these always finish and run a lot of races, but they never seem to actually race as hard as they could). After this, there is a trailing group who really should be running less demanding races (sorry, Les and Phil) until they catch up to the previous group.

The straight line for the third group tends to intersect the time axis at about the finishing time of the last of the second group. This intersect is the fastest time they could expect to run that race and one can compare this group with the same group in other races (one has to keep track of names). The second group's line intersects the axis at a time slower (usually) than the winner's time and these runners also can be compared to each other.

In a very competitive race, those in any of these groups will be in a different group in a less competitive race. Races can thus be compared, but only if one keeps track of names.

So, to predict how one will do, one has to know how one has done before in comparison to other runners and expect that the results from one year of a race will compare to those in another. This isn't always the case; the times in the Voyageur 50 mile have varied by hours because of weather.

Well, it's easier than what I said last time. When I can't run much, I theorize. Let's hope the weather improves!