Wednesday, September 16, 2015

Final Projected Boston Marathon 2016 Cutoff at 91 seconds*


With all of the top 25 feeder marathons complete, the current trend for the 2016 Boston Marathon Cutoff is 91 seconds (90 if you're an optimist).

*Or, it could be higher (see the heading "Dirty Little Secret")

The Caveats:
The analysis assumes a lot of assumptions. I have outlined the objective and methodology on this page. Basically, I look at the number of people who qualified for the 2015 race, using the "feeder marathons". To qualify for the 2015 race, participants needed a BQ minus 1m2s. These races were all run last year. I then look at the same races for this year, and see how many people would qualify using the same cutoff.  If fewer people qualify, then I decrease the cutoff (make it easier to qualify), until the same number of people qualify as did last year. Conversely, if more people make the 1m2s cutoff than last year, I increase the cutoff (making it harder to qualify) until the same number of people qualify.


Table 1: Year vs Year, By Marathon
2015 Qualifying Year2016 Qualifying Year
MarathonFinishersBQBQ2015BQ2015%FinishersBQBQ2015BQ2015%
Big Cottonwood1,30826224419 %1,52728425217 %
Erie 97831228029 %95931728930 %
Berlin 36,1225,7695,45515 %29,0274,5304,30815 %
St. George 5,8181,1981,13720 %5,80699793916 %
Twin Cities 8,8551,01095311 %8,8531,04196711 %
Portland 6,9455044707 %6,4294293956 %
Chicago 38,8543,5273,3149 %40,5814,0043,7629 %
Steamtown 2,16537735917 %2,18443140819 %
Columbus 5,52361156110 %5,44562058111 %
Toronto Waterfront 3,60448745313 %3,97165460115 %
Baystate 1,30430127121 %1,53137835023 %
Marine Corps 23,3851,1701,0735 %19,6877186623 %
New York City 50,1343,9803,7257 %50,4333,6293,4007 %
Indianapolis Mon'tal 3,52151848314 %3,72559455215 %
Richmond 4,8424414189 %5,09552048710 %
Philadelphia 10,9141,3411,25211 %10,3331,3371,23512 %
California International 6,2381,2921,19519 %5,7931,2911,18520 %
Houston 6,9456986519 %7,1336986489 %
Boston 31,21010,56410,04632 %26,63512,79212,21846 %
Bayshore2,01731430015 %2,00834132916 %
Ottawa5,32481274514 %4,50180675017 %
Mountains 2 Beach1,61739937023 %1,58645241926 %
Grandma's 6,1731,1081,04717 %6,0751,1341,06618 %
Santa Rosa1,23125822218 %1,43634131822 %
Lehigh Valley1,13923521519 %91818616818 %
TOTAL266,16637,48835,23913 %251,67138,52436,28914 %
Notes:
1. BQ = Number of Runners who met the minimum Boston Qualifying Standard
2. BQ2015 = Number of Runners who met the Cutoff for the 2015 Boston Marathon (BQ minus 1m 2s)
3. BQ2015% = Percentage of finishers meeting the BQ 20015 cutoff time.
4. AG = Age Group on the day of the feeder race, not the subsequent year's Boston Marathon.
   This is a small source of error, as a person may "age-up" for the subsequent year's Boston Marathon.

Analysis
Looking at the table above: 266,166 runners finished a 2015 "feeder" qualifying race, whereas 251,671 finished a 2016 "feeder" race: 14,495 fewer people. However, more people met the minimum BQ standard 38,524 for the 2016 qualifying year versus 37,488 for the 2015 qualifying year. Moreover, 1050 more people met the 1 minute and 2 second cutoff in the 2016 qualifying year so far than did in 2015.

Now that the Boston Marathon registration is open - assuming the same proportion of qualified runners apply for the race (see objective and methodology) - there wouldn't be enough spots for everyone. We have to increase the cutoff so that we whittle the 36,289 people (who met the 2015 cutoff in a 2016 qualifying race) down to 35,239 the number of people who met the Cutoff for 2015.

Through a numerical modelling approach, the answer to the solution, of what the cutoff should be becomes 91 seconds.

At 91 seconds, 35,230 people will make it into the marathon. (only 9 fewer than this year). For the optimists, at 90 seconds: 35,266 people will make it (27 more than last year).

Table 2: Projected qualifiers, from this years races, by AG based on an 91 second cutoff.
AGFeeder Race FinishersFinishers who met minimum BQBQ minus 91s% of AG meeting BQ minus 91s% of Total Boston 2016 Field
F18-34 42,914 5,437 4,975 12% 14%
F35-39 17,300 2,551 2,348 14% 7%
F40-44 16,693 2,578 2,320 14% 7%
F45-49 12,459 2,618 2,375 19% 7%
F50-54 8,244 1,644 1,510 18% 4%
F55-59 4,127 888 823 20% 2%
F60-64 1,690 400 379 22% 1%
F65-69 542 116 112 21% 0%
F70-74 158 32 30 19% 0%
F75-79 30 5 5 17% 0%
F80+ 6 3 3 50% 0%
M18-34 41,531 5,229 4,832 12% 14%
M35-39 22,569 2,760 2,510 11% 7%
M40-44 25,197 3,303 2,991 12% 8%
M45-49 21,937 4,006 3,653 17% 10%
M50-54 17,314 3,066 2,765 16% 8%
M55-59 10,266 1,965 1,801 18% 5%
M60-64 5,396 1,162 1,080 20% 3%
M65-69 2,253 543 513 23% 1%
M70-74 788 173 163 21% 0%
M75-79 212 35 32 15% 0%
M80+ 45 10 10 22% 0%
Totals 251,671 38,524 35,230 14% 100%

* And now for a dirty little secret...

Dirty Little Secret

As was noted earlier in this analysis, the Berlin Marathon for this year appears to be a large outlier, with a massive influence on the cutoff time. It is keeping the cutoff lower than it probably should be.

2015 Qualifying Year2016 Qualifying Year
MarathonFinishersBQBQ2015BQ2015%FinishersBQBQ2015BQ2015%
All 25 Feeder Races266,16637,48835,23913 %251,67138,52436,28914 %
(Minus) Berlin 36,1225,7695,45515 %29,0274,5304,30815 %
New TOTAL230,04431,71929,78413 %222,64433,99431,98114 %

If I run the numerical model again without the Berlin numbers and a new cutoff such that the number of qualifiers is equal to 29,784... the result is ....

... Pause for effect ...

131 seconds (2 minutes and 11 seconds!!!)


Lehigh Valley Marathon Results Comparison 2014 vs 2015

AG2014 Event2015 Event
FinishersBQBQ 2015BQ%FinishersBQBQ 2015BQ%
F18-34179413723 %144252317 %
F35-3993242326 %73111515 %
F40-4467141421%6516625 %
F45-4959101017 %488617 %
F50-54385513 %348524 %
F55-59245521 %175829 %
F60-6452240 %101110 %
F65-6954480 %62233 %
F70-74
0 %0 %
F75-790 %0 %
F80+0 %0 %
M18-34194393320 %134302622 %
M35-3996211922 %96181619 %
M40-4492151516 %80131216 %
M45-4992181620 %77222229 %
M50-5488121014 %649714 %
M55-5959131222 %37121132 %
M60-64358623 %234417 %
M65-69134431 %72229 %
M70-74
0 %3000 %
M75-79


0 %
0 %
M80+


0 %


0 %
Total1,13923521521 %91818616820 %
Notes:
1. BQ = Number of Runners who met the minimum Boston Qualifying Standard
2. BQ2015 = Number of Runners who met the Cutoff for the 2015 Boston Marathon (BQ minus 1m 2s)
3. BQ% = Number of Runners who met the minimum Boston Marathon Standard.
   This is a small source of error, as a person may "age-up" for the subsequent year's Boston Marathon.

46 comments:

  1. Could Boston Marathon also be considered an anomaly like Berlin, but in the opposite direction? Do you know the percentage of qualifiers at Boston who return to run it again the following year?

    ReplyDelete
    Replies
    1. I also had thought of this scenario too on Boston. Not sure what to make of it. Could there be lots of double qualifiers in the sample (Boston and another race), could there be a 'been there done, that, don't need another expensive running race'. Don't know, it is predicting human behavior. I am not sure if we can get at registration numbers per wave as we go, think I've seen those in the past, comparing that to last year may be helpful too.

      Delete
  2. Wow! Now we all just wait and see. I'm sitting at BQ-143 seconds (2:23) and going crazy as this is my first BQ. I'm really hoping to get in. Only time will tell now. Good luck to everyone!!!

    ReplyDelete
  3. It'll be interesting to see how close you come to the actual cutoff. Thanks for feeding our squeaker frenzy. Here's hoping that you and everyone else who wants to run gets in!

    ReplyDelete
  4. Thanks! Good luck everyone. Too bad you can't "borrow" extra seconds from a friend.

    ReplyDelete
    Replies
    1. I am gonna think positive thinking to believe that you and I will squeak in with under a minute!!! Positive talk=Positive results!

      Delete
  5. In BQs May 1 post under The Big Wrench he said there were a 1000 fewer Berlin BQers this year. So why does his projected cutoff go up under September Dirty Little Secret section in today's post to -131? What logic am I missing?

    ReplyDelete
    Replies
    1. It's because when you remove the Berlin results - the total number of people who meet BQ minus 62 would be 13% last year, but 14% this year. That extra one percent of the total field is driving the BQ cutoff upwards. And as noted below - the Berlin representation at Boston is presumably statistically lower than the ~5000 qualifiers it produces.

      Delete
    2. So since you consider Berlin an outlier, in the sense that Boston representation is lower, but it is still considered a top 25 feeder, would it not be more accurate to assume that the cutoff is between 91 and 131 seconds? Put differently, you probably should throw some kind of factor at Berlin, such that if that factor is 1, cutoff=91. If 0, cutoff=131.

      Delete
    3. Yes. An outlier in that it has lower representation, but also in that the number of participants and qualifiers varied Significantly from last year and this year.

      Delete
    4. There might be an externality here that works in our favor. The strong US dollar means that fewer foreigners will have the incentive/means to fly to Boston and stay there this year.

      Delete
  6. I would that it's guess because even though Berlin produces a high number of BQs, very few of those people actually run Boston. So their representation in Boston is low. So low, that the BQs from Berlin shouldn'the be in his analysis. I believe that the overall finisher totals from Berlin were down last year when compared to 2013. That would explain why the number of Berlin BQs are down.

    ReplyDelete
  7. Thanks for doing this! Could you re-run the numbers for the new 30k field size that was announced? I suspect the time will be much harder with 5k less spots used in the calculations.

    ReplyDelete
    Replies
    1. Using my methodology, it doesn't matter what the actual field size is as long as it is equal to last years - which it is. My analysis sums up to > 35,000 qualifiers, but this is just based on the feeder races. Not everyone from the feeder race who qualifies actually registers (and has been pointed out, some people qualify more than once). The analysis assumes that the same percentage of people from those 35000 feeder race qualifiers apply for next years race as did this years.

      Delete
    2. Ah I see. I thought the field size last year was 5k higher, which is why I asked (but maybe that was including the charity entries). Thanks!

      Delete
    3. Audrey, that was 2 years ago. And most of those extra 5,000 runners were "invited guests" and such. So 2014 had pretty close to the same number of "qualifying" spots available with 35,000 field size as 2015 did with a 30,000 field size and as 2016 will have with a 30,000 field size. In all 3 years, the actual space reserved for "qualifiers" has been running pretty consistent with ~25,000 spots.

      Delete
  8. The heck with it, let's start the guessing contest. I will predict 84 seconds.

    ReplyDelete
  9. Im betting cutoff time = BQ; that the appetite to run Boston will subside slightly to where a cutdown will be unnecessary.

    ReplyDelete
    Replies
    1. I think (and hope) you're right. There was obviously a surge in demand in 2014, and I'm sure some of that spilled into 2015 as people that weren't in great shape in 2013 couldn't get their times down fast enough. Hopefully that demand surge subsides more this year, and the number of qualifiers (and therefore this website's analysis) is irrelevant. We'll know pretty soon.

      Delete
    2. Unfortunately, this does not appear to be the case. It has been reported by BAA that there was an overall increase in applications during the BQ-20mins, BQ-10mins and BQ-5mins period. We know that overall marathon participation is down this year over last year, but it appears Boston still has that magic allure. Add to that the fact that more people ran faster than last year, and it looks like we're heading to a cutoff in excess of last year.

      Delete
  10. Don't know if y'all saw this mornings update, but apparently there's been an increase in Week 1 registration over last year. Doesn't bode well. Only about 5,000 spots left for mortals.

    ReplyDelete
  11. No one to blame but myself for not running faster - :58 seems like it wont be enough based on that update.
    http://www.baa.org/races/boston-marathon/participant-information/athlete-registration/registration-updates.aspx

    ReplyDelete
  12. Can we get an update after they just released that there are only 5,000 spots remaining? 3,000 less than last year. Yikes!

    ReplyDelete
    Replies
    1. My analysis does not take that into account. I make only three assumptions: 1) The feeder races are a fair representation of overall qualifiers 2) The number of slots allocated to qualifiers is the same as last year (which from everything I have read is accurate) and 3) The same Proportion of people apply for Boston as did last year. <- This is the key thing, if 20% of all finishers last year applied, but only 18% of all finishers did the year before, then that shows an increase in demand, and that will skew the final cutoff more than anything.

      Delete
  13. So last year there were 8000 entries left and basically all but 1900 or so got in. that would mean roughly 9900 or say 10,000 people met the standard less 4:59. if you then assume roughly the same number sign up over the next three days, what does the cut off need to be to eliminate 5,000 (from 10,000 down to the 5,000 left) entries from the pool??

    if the 10,000 is ratable over the 5 minutes, than the cutoff is close to 2:30...

    ReplyDelete
    Replies
    1. The press release from last year indicated that only about 6500 runners were accepted from the second week of registration. I'm not sure what happened to the other 1500 slots. So really there were about 8500 people that registered the second week, not 10,000. Of course, is 5000 really the number of slots up for grabs this year, or will it be less
      like last year?

      Delete
    2. Yeah, 5000 spots left is not a good sign. The 2:30 number makes sense assuming 10,000 are left to register. It also assumes all the people left are 5 minutes or less. My wife registered today, and she is BQ - 14 min. Hopefully she is an outlier. I am sitting at BQ - 2:25.

      Delete
    3. Yeah, there were 6,447 spots for 8,394 applicants. If the same number of applicants applied this year and if they're evenly distributed over 299 seconds and if there are exactly 5,000 spots left this week, The cutoff would be 121 seconds. If you were to run the same calculation assuming the 6,447 runners were evenly distributed over 238 seconds (4:59 to 1:02), then the cutoff would be 114 seconds. That's a whole lot of ifs and ridiculous assumptions. Also my time is BQ - 113, so I'm all worked up. I wish I could just go into suspended animation for several days.

      Delete
    4. Completely agree, Dick Sullivan! I just want to put my brain on pause until it's announced. So many ifs, buts and other possibilities. BQ-2:32. "Squeaker frenzy" made for a solid run tonight as I ran off some of this extreme anxiousness. All of my running pals are in (they're all in a higher age bracket). I'll be absolutely heartbroken if I miss the cut-off.
      Sending positive vibes to you all :)

      Delete
    5. I've been hatching a theory. We all now that the application rate has been higher this year than last year. What if all of that is just attributed to a better understanding of the "new" process compounded by better social media messaging from BAA. In other words, what if all the fast qualifiers applied in the first week as opposed to previous years where they may have only applied in the second or third wave. Perhaps there really isn't more people applying this year afterall. Just that the fast ones applied as soon as they could, and perhaps that pool of qualifiers is exhausted leaving only the true squeakers to qualify now.

      (Grasping at straws obviously)

      Delete
    6. BQ, first, your site has been amazing. I greatly appreciate your input. Knowing some numbers people who always say, "trust the stats,"....are you still relatively confident in your final analysis, and much of this is speculative conjecture?

      Delete
    7. This comment has been removed by the author.

      Delete
    8. I saw a comment on Facebook where someone suggested there were some feeder marathons last fall that took place between week 1 and 2 of registration that would have added a bunch of people to the week 2 registration process, and these races occurred (or will occur) on different dates this year. I have looked to see what those races would have been or what impact they might have. Instead, I'm going to cling to the belief that we are more or less down to just the 0 - 5 minute people, and most of us will get in.

      Delete
    9. I have used Dick Sullivan's numbers, but assumed there will be 15% more registrations this year for the 0-5min people (like for the 5+min people), which equates to 5000 entries out of 9653 applicants. Furthermore, instead of evenly distributing the 5000 entries over the 300 seconds, I have put more people in the first cut-off minute based on the 2015 numbers (8394-6447=1947/8394= 23.2% for the first 62secs instead of about 20% if evenly distributed), likely because a lot of people follow rabbits and outsprint them towards the end. Using the same logic, I have hence put 22.5% of people in minute 0 to 1, 20.5% in minute 1 to 2, and 19.0% in minutes 2-3, 3-4 and 4-5. This gives me a cut-off of 2min17sec. My time is BQ - 2min18sec.....................

      Delete
    10. Nice analysis using the new numbers. I hope your projection of 2:17 is close to accurate. I'm sitting at BQ minus 2:23 and feeling very doubtful about my chances of getting into my first Boston. Good luck EVERYONE! Registration closes in just a few short hours! I, for one, would love to see all of us make it in! :-)

      Delete
    11. I like the analysis giving a cut-off of 2:17 as I am at 2:55 and would take that. However, if you look back at 2015, BAA announced 8,000 entries remaining for week 2 but only 6,447 registered with a time 0-5 min. The remaining 1,500+ must have registered during week 2 with a 5+ qualifying time. If the same holds true for 2016, perhaps there are only 3,500 or less spots actually left for 0-5 min qualifiers therefore driving the cut-off a bit higher.

      Delete
  14. Boston Qualifier, Does the news of increased early registration change any of your analytics? Thanks as always

    ReplyDelete
    Replies
    1. No, the analytics that I run, make only three assumptions: 1) The feeder races are a fair representation of overall qualifiers 2) The number of slots allocated to qualifiers is the same as last year (which from everything I have read is accurate) and 3) The same Proportion of people apply for Boston as did last year. <- This is the key thing, if 20% of all finishers last year applied, but only 18% of all finishers did the year before, then that shows an increase in demand, and that will skew the final cutoff more than anything.

      Delete
  15. Is there a way to verify if your analytics are correct thus far? In other words, could you have predicted so few spots left for the <5min qualifiers? Thanks for your hard work! I applied with a BQ-2:12, I'm hoping you are correct!

    ReplyDelete
    Replies
    1. He doesn't predict application rate (percent of qualifiers who applied) and the distribution of qualifying times (percent of qualifiers in each time buckets). Those are actually underlying assumptions for his analysis.

      Delete
  16. One reason to ignore/deemphasize Berlin:90%, more or less, of Boston entrants come from the US.

    ReplyDelete
  17. I hope you are right! I am one of the people who qualified with a marathon that isn't on your list, and I imagine there are a lot more of us out there. I have 210 seconds in the bank, and according to your stats there are about 2500 people ahead of me..... and I know people (like my husband) who didn't register til the second week who had much more time in the bank (he had 23:30 minutes) because they were waiting to see if squeaker spouses could even apply. 12 more hours.....

    ReplyDelete
  18. Too much fuzzy logic. The number of finishers in a race as no bearing on how many qualify. Also, adding the 62 to the analysis is pointless. Better to see what percentage of the field was filled last year in the first week and do some simple calculations based on the numbers of slots left.

    ReplyDelete