# Interesting article on the (in)accuracy of race time predictors (Read 525 times)

Running any race at "the line" is going to hurt -- if it didn't, you wouldn't be at the line. But the hurt is different in different kinds of races. I much prefer the marathon kind of hurt to the 10K kind of hurt. Paced well, it doesn't really get bad until 22-23 for me, and even then, it's different. But your first marathon is probably not going to be at this line anyway.

I get annoyed when I hear people say, a certain race distance is the hardest."  I usually respond (and it shuts them up) with "Every race from 400m and beyond hurts if you give it your all... If a race doesn't hurt, you aren't trying very hard."

The other thing that annoys me when I hear people say, "It's so easy for those fast guys." I usually respond with, "those fast guys are hurting just as much if not more than the slower guys. They just know how to look more relaxed when running fast."

There's an article in the most recent Running Times magazine about which race distance hurts the most.  By and large the opinion is what I and bhearn just said (they all hurt if you push yourself), but the results more favored the marathon and ultras. I gotta believe that's simply the results of the readership and who voted.

scappodaqui

rather be sprinting

I inferred that he compared half and full times for individuals, not medians or means:

" I found a core of a thousand runners who (you'd hope) would have recorded a representative time at both distances. They range in ability from five-minute milers all the way down to twelve-minute milers."

"I fed all of my 1071 runners through that formula, and found that only 49 of them managed to hold on to the tails of 1.06 - it was far more common to see a score of 1.15."

So he did calculate an exponent for each person.  he goes on to discuss the difference between men and women.

1. Do 1,000 runners count as enough data?

2. Controlled for 'fun runs' vs. 'hard runs'?

PRs: 5k 19:25, mile 5:38, HM 1:30:56

Lifting PRs: back squat 176 lb

DoppleBock

The last 6 miles are tough because of fatigue, dehydration, glycogen depletion.

It can be more than just pacing too fast.

A certain amount of desire / mental toughness can overcome some issues, but not all.

To me the answer is to pace correctly and not conservatively.

Retired 1/1/13 ...  12/1/16 return to Gallowalking ... Running is beautiful and forgiving, it will always take you back with no questions asked.

GC100k

1. Do 1,000 runners count as enough data?

2. Controlled for 'fun runs' vs. 'hard runs'?

1,  1000 runners are a lot of data.

2.  no.

Y'all are making this too hard.  He took half-marathon and marathon times for 1000 runners.  Very few fit the old 1.06 exponent.  1.15 was more representative of the data.  So, if you're going to predict what someone may run, not what they should or what they're capable of but what they are actually likely to do, you'd be better off using 1.15.  That's it.  It's not complicated.

Some years back airlines discovered that their estimates of weight based on the number of passengers were low from using old data. So they updated the data to reflect actual averages.  Nothing about what people should weigh, BMIs, weight vs height, fit vs unfit people.  The data just said the average person weighs (say) 180 lbs so if you have 100 passengers you should estimate 18000 lbs.  You might be off for different groups, but that's what the averages would predict..

It's kinda like that.  It is what it is.

DoppleBock

There is some difficulty in what the purpose of the race - The size of the population is OK, but the control of the experiment (People putting max effort into each) would seem to where the question is at.

I have never tapered for a 1/2 marathon or considered it a goal race.

For it to be an accurate predictor, it would have to be done between 3-4 weeks prior to the marathon and at a race type effort.

Retired 1/1/13 ...  12/1/16 return to Gallowalking ... Running is beautiful and forgiving, it will always take you back with no questions asked.

No more marathons

1,  1000 runners are a lot of data.

2.  no.

Y'all are making this too hard.  He took half-marathon and marathon times for 1000 runners.  Very few fit the old 1.06 exponent.  1.15 was more representative of the data.  So, if you're going to predict what someone may run, not what they should or what they're capable of but what they are actually likely to do, you'd be better off using 1.15.  That's it.  It's not complicated.

.

Gotta agree with GC here.  All the couda, shouda, wouldas don't mean squat.  This isn't  talking about what these 1,000 could have done if they had or had not done or not done lots of other stuff.  This is simply reporting the facts ma'am.  This is saying given a set a data, the relationship between set A and set B is 1.15.  You can argue all day about whether or not this applies to you, but it is what it is.  (And by the way, I like it because it helps justify my lazy ass approach to marathoning.)

Boston 2014 - a 33 year journey

Lordy,  I hope there are tapes.

He's a leaker!

Gotta agree with GC here.  All the couda, shouda, wouldas don't mean squat.  This isn't  talking about what these 1,000 could have done if they had or had not done or not done lots of other stuff.  This is simply reporting the facts ma'am.  This is saying given a set a data, the relationship between set A and set B is 1.15.  You can argue all day about whether or not this applies to you, but it is what it is.  (And by the way, I like it because it helps justify my lazy ass approach to marathoning.)

The quantity of data in a data set is only one marker among many in reliability. The relevancy of a sample of data is much more important than the quantity of the data.

What's being argued over here is the relevancy of the data. Most race predictors only consider trained runners to be relevant data, and this makes sense to me because why would someone want to know how fast they could run if poorly trained?

+1.  I don't want to know what I could do if I have a bad training cycle.  I want to know what to shoot for if I can get all the stars to align.

The quantity of data in a data set is only one marker among many in reliability. The relevancy of a sample of data is much more important than the quantity of the data.

What's being argued over here is the relevancy of the data. Most race predictors only consider trained runners to be relevant data, and this makes sense to me because why would someone want to know how fast they could run if poorly trained?

"When a person trains once, nothing happens. When a person forces himself to do a thing a hundred or a thousand times, then he certainly has developed in more ways than physical. Is it raining? That doesn't matter. Am I tired? That doesn't matter, either. Then willpower will be no problem."
Emil Zatopek

No more marathons

+1.  I don't want to know what I could do if I have a bad training cycle.  I want to know what to shoot for if I can get all the stars to align.

Well, there's bad and then there's meh!  I know what I need to do to get that 1.06 marathon performance result.  I need to run 70 to 80 miles per week and include a good mix of endurance and speed workouts.  I also need to lose about 25 pounds.  Neither of those is likely to happen - 45 to 50 mpw and pizza and beer.  And so the 1.15 becomes "relevant".  I can live with that, especially since it gives me a realistic expectation - less likelyhood of the death march at the end.

Of course, I'm done with marathons with the exception of Boston 2014.  Hate the damn things.  Too unpredictable.

Boston 2014 - a 33 year journey

Lordy,  I hope there are tapes.

He's a leaker!

100K or Bust

Some interesting observations here between the 1.06 fitting the WRs but the WRs aren't all held by the same person to the 1.15 being more pragmatic as it fits the data used and is probably more accurate for the recreational runner. Many years ago before Riegel ever published his formula my coach introduced me to a similar formula based on the log of the distances. What was important was that it also had a coefficient that could be determined by solving the equation using two equivalent performances at different distances. It could also be determined by plotting race pace vs log(d) on semi-log graph paper. The coefficient was specific for the person. A middle distance runner was more likely to have a higher value than a long distance runner or marathoner assuming each had gravitated to their distance because that's where they raced best. The equivalent would be to determine what value the exponent in Riegel's formula fits your best effort data. I did some quick checking with Eric's calculator and 1.09 seems to fit my historic data from 3K (yes 3K, not 5K) to marathon. My 800m and mile times are much faster than what the calculator would predict, so it's no surprise to me that my personal exponent is higher than 1.06.

2017 Goals: for races not to be exercises in futility

Yeah using Eric's calculator I get a 1.08 coefficient from 3k to marathon based on my best times as an adult road racer (going back to high school 800 and mile times would not make much sense for me.)

My best distance appears to be the half marathon, both in terms of my all time PR and my best race from the last year or two. I suspect this has much more to do with the kind of training I do than natural ability. I train to be a generalist, not a specialist.

Runners run.

Maybe if Eric wanted to get really fancy with his calculator, it could derive a coefficient based on the last six months of training from your running log. (?)  Analyze all the existing data do derive the formula, then apply it for future calculations, perhaps?

Well at least someone here is making relevance to the subject. - S.J.

Maybe if Eric wanted to get really fancy with his calculator, it could derive a coefficient based on the last six months of training from your running log. (?)  Analyze all the existing data do derive the formula, then apply it for future calculations, perhaps?

Why 6 months? Why not 2 years?

And in my experience there is only one data point Eric would need to include, in addition to whatever race times are available: mileage.

Runners run.

Why 6 months? Why not 2 years?

And in my experience there is only one data point Eric would need to include, in addition to whatever race times are available: mileage.

Sure, 2 years.  Maybe only races that you actually raced.

Well at least someone here is making relevance to the subject. - S.J.

I put my 7k and 10k into the calculator, it predicts my marathon is 3:43:14. If I use 5k and 7k, it gives 3:12:42, but if I just use 10K alone, it gives 3:23:12. Why 7 and 10K produces slower?

5k - 20:56 (09/12), 7k - 28:40 (11/12), 10k trial - 43:08  (03/13), 42:05 (05/13), FM - 3:09:28 (05/13), HM - 1:28:20 (05/14), Failed 10K trial - 6:10/mi for 4mi (08/14), FM - 3:03 (09/14)