I asked on another thread for the methodology behind a statistical study that many are quoting. Thus far I have not seen any answers provided.
Here is my own study, with methodology provided.
What was looked at:
- Race results for all five major marathons (most recent runnings) plus Boston '09 and '10
- Looked exclusively at top dozen competitors. No offense to those running 2:25 as this is a great individual achievement for most runners (well, maybe not the Letsrun variety but for normal humans). It is just that beyond the top dozen you start to get into folks who just are not in the same galaxy as a Ryan Hall so I did not think that including their results would be particularly enlightening.
- Compared the time run for a particular race to that person's marathon PR (set on any course) coming into that race.
- Threw out all debuts since there is no PR to compare.
What is shown here is the average of these time differences (difference between day's result and former PR).
First, Boston to Boston:
Boston 2009: +2:42
Boston 2010: +0:38
Boston 2011: -1:06
That is, in 2009, the average of the top dozen finishers came in at 2 minutes and 42 seconds slower than their previous PR. Similarly for the other years. This is not surprising. Most races do not yield PRs for most runners.
First thing to note is that for 2011 the average of these elites beat their own PR by over a minute. Even on the much faster courses I would guess that this is a rarity (although I only crunched the numbers as described above so that is all I can say for certain).
I think that it is well understood that this year's race was noticeably faster than the average Boston. What some may not realize is that last year's race (2010) also had favorable conditions (just not as favorable).
Compared to 2010, this year's race was 1:44 faster for the elites (top dozen). That is, relative to their PRs coming in, these folks ran 1:44 faster in 2011 than in 2010.
Making no assumptions about 2009 being representative, 2011 was 3:48 faster than 2009 for the elites (top dozen). I don't know how 2009 is considered relative to Bostons past. If it was a pretty average year then this suggests, (note that I did not say "proves"), that the 2011 version was somewhere around 3:48 faster than average (for the top dozen finishers).
Now on to the other majors:
The average performance relative to their then-current PR for the top dozen finishers was as follows:
Berlin 2010: +1:06
New York 2010: +4:13
Chicago 2010: +1:08
London 2011: +0:32
Boston 2011: -1:06
What we see here is not surprising. London, Berlin and Chicago are all reasonably fast courses while this year's London race conditions may have been a bit more conducive to fast times than usual. New York is well known as the slowest of these courses. I don't think that a 3 minute differential between New York and Berlin would surprise a whole lot of folks.
The one anomaly that stand out is, of course, Boston. Boston is considered by many as being 1:30 - 2:00 slower than Berlin or London and yet it ran 1:38 - 2:12 faster this year. The difference between the normal positive handicap of Boston to this year's advantage at Boston could be anywhere from 3:08 to 4:12 or an average of 3:40.
If we use the difference between the 2011 Boston and the 2011 London numbers as an indication one could conclude that Geoffrey Mutai might have been able to pull off something like a 2:03:02 + 1:38 (= diff between +32 and -1:06) = 2:04:40. The fact that this number exactly matches Emmanuel Mutai's London effort is purely coincidental but to me it seems fitting to have Mutai 2:04:40 and Mutai 2:04:40.
Similarly, looking at Ryan Hall's performance at Boston (since this is an American-centric crowd) one could reasonably say that Hall may have been able to run something like 2:04:58 + 1:38 = 2:06:36 at London this year. Certainly an excellent result and very much in line with his spectacular London performance of a few years back.
Now, I make no claim that this "proves" anything. Indeed, while there has been much dispute on these boards over the Boston results I am not certain if anyone has every actually stopped to define precisely what the question is. At any rate, I do think that if one is to make any sort of mathematical claim then one is obligated to provide the details of their methodology in order that others may critique the method and/or reproduce or dispute the results.
Please note that these results would appear to be in line with the expressed opinions of some on these boards and very much out of line with others. For my part, they at least pass the laugh test and are open for your consideration.