Return to Olympics Marathon index page

Analysis of the Measurement Data for the Olympic Marathon taken on 13 June 2012

I find I have 184 emails relating to the measurement and the subsequent analysis. This must surely be a record of some sort - it certainly is for any measurement I have made. It looks like I can extract bits from those emails which will tell the story of the post measurement data-analysis. The twists and turns during this analysis phase may make interesting reading for the dedicated measurer so I have selected here extracts from these emails which describe the key elements.

Warning this page is rather long. You may wish to view a summary before deciding to look further at the blow-by-blow detail below.

First calculation results

13 June: On reaching home I had a short sleep then calculated my results and at 15.31 sent them to Hugh and David

Dear All, Enjoyed the morning. Got back at 8 am, then to bed for 90 mins. Have now copied my data to a spread sheet (attached) My small lap was: 3.5679 km My large lap was: 12.8232 km My laser measurement of the cal course came out to 328.621 metres which is 61mm longer than the value which I wrote down from Hugh [i.e. 328.56 m]. Did that include the NIST calibration correction for the tape? I need to go through the small geometry corrections for the laser measurement again, but I will wait until tomorrow when I hope to have a clearer head after a good rest.

Hugh replied at 18.50

"I'm afraid these may need recalculation. David said the correction on his 100m tape was 16mm (so 3.28 x 16mm total). However, it seems it should have been 13mm, so the correction should have been 3mm x 3.28 less - that is, the calibration course should be 10mm longer than was given (ie 328.57m instead of 328.56m)

My figures, based on the 328.56 length of the calibration course, were: short lap: 3568.04m long lap: 12826.55m

David's were: short lap: 3568.58m long lap: 12826.27m

All extremely close"

Note: All the above distances were calculated using the two turn around references ( the expansion joint on Westminster bridge for the small lap and the East end of the central divide at the Tower of London for the large lap.) So these distances do not include the length of the 4.3 metre radius turning sem-circles.

Measurement done, David caught a flight back to back to New York and on 15 June I wrote to David and Hugh:

Hope you had a good flight, and that the prognosis on the knee is OK. I have updated the spread sheet I sent your previously, and added more cal data from Long Tow, Abingdon, in the 24 hours after the London measurements. The sheets are:

  1. My Cal data Long Tow and the Mall - also my laser check on the Mall Call course which makes it 5.1cm longer than the length which Hugh reported to me.
  2. My Olympic Course measurements of the two laps.
  3. Correction of My cal data for temperature and deflation using the coefficients for this tyre which I determined in June 2011. These are consistent with Long Tow being rougher than the Mall, But I am surprised the change is as large as 6 counts in 11200. I think I would have expected about 3 counts.
  4. Summary of the results - Hugh's and David's values for the two laps were sent to me in an email from Hugh - I dont know if these are the final figures. I am about 9m shorter for the marathon. However it is an extremely twisty course.

Here is what I wrote in May: "Looking at the map on the web (version 9 -30/6/11) I have added up the equivalent of roughly 30 complete 360 degree turns in 42 km. So it will be equivalent to 30 laps of a simple oval loop of length 1.4 km. If the feet of the crowd control barriers are placed on the road way sticking out say 30 cm into the road beyond the kerb, then this will add 54 m to the length when compared with what would be measured if the measurers were to use the kerbs as the edge of the allowed route."

Using that estimate for the effect of mal placement of crowd barriers, then if my average line for the measurement was 9/54*30 = 5 cm closer to the kerbs than that of Hugh and David it would explain why I got a low value. I did try and not measure loosely but I would not have thought that I was more than 3 cm closer to the kerb on average than the nominal 30 cm. Of course the course surface was less smooth in places, and there were some flattish cobbles - this could produce slightly different effects on our different tyres.

Question raised over the steel tape tension and the correct NIST calibration to use

Later on the 15 June, after receiving my email, I got a phone call from David. He explained to me that he wanted no public discussion of the results. He said that the debates that had occurred in previous group rides of Olympic Courses were not helpful. Having followed some of the previous debates especially that on the measurement of the 1996 Atlanta Olympics marathon I understood what he meant. However there were a large number of measurers involved at Atlanta, all getting somewhat different results. So there was plenty of scope for debate as what to choose for the final distance. I agreed with David.For London 2012 we had just 3 measurers involved and our results were closely grouped.

I asked about the taping of the calibration course. David suggested that I be put in touch with NIST concerning the tape calibration. This I declined since I did not see any need - the NIST report appeared perfectly clear to me.

After reflecting on what he had told me I emailed back to David,

Have I got this right? On the phone did you say that your tape tension was an estimated 10 lbs? The NIST tape calibration was at 10 kg. So if you used the NIST table of tape length data at 10 kg, your real course length would be shorter than you worked out. In the attached spread sheet I calculate it would change 328.570 to 328.517. Please check if I have got the sign correct.

I am a bit concerned that it makes the tape to laser difference larger:10.4 cm or 0.04% which is more than I would expect. Perhaps I made a mistake in the laser setup. For example if I had the laser zero reference set to the wrong place on the device it could give a larger reading. The laser comes up by default with the reference as the back edge of the laser. But had it somehow got changed to use as the reference the end of the backwards extension arm, it would have added 45 mm to each laser reading, giving a total extra of 9 cm. I would be surprised if this happened since as I say I had planned to use the default reference, but I did not actually check the reference setting which is shown on the reading display. My raw laser readings without any corrections were West half:159.539, 159.539, 159.539; and for the East half: 169.058, 169.055, 169.059, 169.054, 169.058.

16 June I wote to David

My 50 m tape weighs 0.8 kg including its light metal case, so guessing 0.1 kg for the case, the tape weighs 0.7/50= 0.014 kg/metre. The NIST reports shows David's tape is 0.152 kg/metre so the two tapes must have nearly the same cross-section, and assuming steel of the same Young's Modulus they will stretch by the same amount under a given tape tension. Therefore my conclusion that David's tape has similar elasticity to mine is confirmed. This is actually not directly relevant to the question of the size of the correction. I was just interested to see how your tape's cross-section and therefore elasticity compared with mine.

Of somewhat greater importance and to reinforce what I sent in my spreadsheet yesterday yesterday evening about the length of the Mall Cal course. If David's tape was pulled at only 10 lb (4.5 kg), which is what I understood you tried to aim for, then the Mall Calibration length is 328.517 ( not 328.570) and the marathon course measurement results should be reduced by 0.16% - just 7m in 42k. If of course you pulled twice as hard then you are correct in using the correction from the NIST table of 6.61 mm for 50 m. I am not sure why NIST pulled the tape at 10kg. The standard in Europe for Class II tapes is a pull of 50 N = 5.1 kg = 11.2 lbs. This is what I teach measurers to pull class II tapes at.

Not getting any response, on the 19 June I followed upto David with:

Do you have any thoughts about the tape tension? See my email.... It would make 7 m difference to the marathon length if your tape was pulled at about 10lb as you said to me, rather than the 22lb which NIST used for their calibration readings. Do you have a copy of the raw readings for yourself and Hugh? I would like to put them into my spreadsheet which I sent you last Wednesday because I am keen to see if any of my intervals were badly out, perhaps because of a wrongly written down number. Alternatively I would be interested to see if the small difference in our results built up gradually over the whole course, or was perhaps worse in specific areas perhaps with certain types of surfaces, e.g. the cobbles.

Hugh who had been copied into the correspondence sent me his results and I replied to him

Hugh, Many thanks. I will do the comparison tomorrow. Did you make any notes on the calibration course layout? Particularly what tape tension do you think you used?

David was busy he replied

Dear Mike, I have been non stop traveling, working events over the past week and hope to be able to send you the data over the next few days. David

Next morning (20 June) Hugh emailed

Mike, I nowadays use just the "firm pull", but it was David who tensioned the tape from his end. We measured 6 tape lengths westbound and then 28+ metres. Going back we found 3 mm of difference. We did the temperature correction for 11C and also compensated for the laboratory verification of David's tape (he at first said this was 16 mm over the 100 m tape length, but later corrected that to 13 mm, causing the calibration course length to increase from 328.56 m to 328.57 m (3 mm /100 m x 328.5 m) Hugh

I replied

Hugh, Thanks for that. The exact tension wont change the overall conclusion that the course was well measured, but it does make me slightly worried about my laser technique because if David pulled at only 10 lb then the difference is a little larger than I can account for from what I presently understand of the errors in the laser technique. I will do more study of my laser errors.

I have compared my data with yours. I will send a spreadsheet later to you and David when it is complete, but here is an image of the current status showing one point, marked in yellow, where I may have stopped 0.5 m out of position, and 3 intervals, marked in red, where I seem to have ridden a bit tight through the very twisty parts. Perhaps the cobbles also had an effect on two of these intervals. It will be interesting to see in due course how David's intervals compare.

Hugh-Mike Interval Comparisons

Later on 20 June I sent out a draft assessment of the interval differences between Hugh and my counts for the large lap (5 MB pdf).The final version of the comparison of intervals, including the data from David which I received later is shown in the graphs on this page .

Wall Street Journal Article

21 June

We heard from David that the Wall Street Journal article had appeared. David said that much had be written from an interview he gave before he headed to London for the measurement with updates from a post measurement interview.

I replied,

David, I was delighted to see the smashing article which is excellent PR for measuring. Obviously we can link to that article. Can you get a contact for the photographer Gareth Phillips. I would like to seek permission for reproduction of his photographs. I would particularly like to ask him for a photo showing he whole measurement procession complete with the police blue lights. I think it might provide a good general impression of the procedure, which I might be able to use in our course measurement website over here.

Geoffrey had been supposed to get such a picture for me, but he was not able to get his new camera into the right mode to get really good pictures in the gloom. We clearly should have practised properly beforehand but it was all rather a last minute plan.

I was amused to see Geoffrey and myself in the background of this picture of yourself. We are engaged on my attempt at laser measuring the calibration course. It nicely shows my reflector box. I am still keen on ironing out the practical procedures for this laser measuring, despite the comment in the article implying that lasers at present have no place in course measurement.

calibration course David poses for the Photgrapher. Mike and Geoffrey at work in the background

Months later I did a web search to find the phographer for the WSJ who had accompanied our measurement ride in order to ask for a photo of us riding surrounded by police motor bikes with blue lights flashing. I thought it might be nice to use it on the strictly non-commercial UK course measurement web site. The photographer replied offering a "discounted rate" of £50 for use of one of his photos. Now I understand that phtographer has to make a living, but I was surprised that an unpaid volunteer who had been featured in his photos was being asked to pay. Presumably he had already been paid handsomely by the WSJ for the photos which they had used. Did I have any privacy right to ask for photographs of myself not to be used in his commercial/jounalistic activities? Probably not.

Effects of electric assistance on wobbles during my measurement

I was asked about my use of electric asistance on my bike:

I don't know how much electrical assist you used, but I suspect it produces less wobble than does pedaling. If you didn't use it when calibrating, but did use it while measuring, the front wheel would not have to compensate for the back-and-forth forces caused by pedaling.

22 June - I replied:

I used medium electrical assist although the ride was not challenging and minimum assist (an extra 50% on top of what I provided into the pedals) would have been sufficient to keep me close. When at the back you tend to need to accelerate to catch up from a stop hence I chose medium (100% addition to what I provided)

You could well be correct that I wobble less with the benefit of extra power. It is on steeper hills where I really notice the reduction in wobbling. But of course Hugh is an excellent rider and I expect he wobbles very little even when accelerating hard or going up slopes.

I have gone over the map adding the angles turned through. The large lap has about 9.1 X 360 degrees of turning. So if on the large lap the 3 m, which Hugh and I differ by, is all due to me riding a tigher line on bends, then I rode 5.2 cm closer to the kerb on average than Hugh.  So if Hugh rode the perfect 12 inches, then I would have had to ride at 10 inches - just about possible I suppose since Hugh is very  familiar with the corners whereas I was seeing them for the first time. To be sure of a perfect 12 inch ride I would have wanted to stop an check the distance in a few places to help judge it right.

Probably the difference is a mix of many effects including this one. We shall never know. The great number of turns help to made it a challenging ride which would test a poor rider.

The laser ranger method for measuring calibration courses

Because of the unexplained difference between my laser ranger measurement of the Mall calibration course and that reported by Dave and Hugh from their steel taping, I carefully looked at my laser calculations and procedure.

Firstly I found that my initial calculation of the laser distance was in error by 10 mm. I had not allowed for the fact on the second step the plumb bob was aligned 12 mm SHORT of the calibration course marker position. I also made some minor corrections for the fact that the effective reflection point of the retro-reflector was not at its surface. This changed my laser measurement from 328.621 to 328.631.

It was now 114 mm longer than the figure we had at this point in time for David's steel tape result.The difference was now 0.035% far more than the specified accuracy of the Bosch laser which is given by Bosch as 0.013% in unfavourable conditions and 0.005% in favourable conditions.

I decided to document fully the method which I had used. When on the Mall in London I had not made a photographic record of my method, I decided to repeat the method on another calibration course and photograph each step. I chose Park Road in Abingdon to lay out a 286 m course. My procedure and the calculations are fully described here.

At this point , on 23 June2012, it appeared from the Mall data that I had proved that the laser was not as accurate as specified. However, I could not square that with earlier comparisons which I had done on 29 March 2011 between my own "Silverline" steel tape and the laser on a 249.906 m calibration course. I than found that the laser gave 249.955 m So the laser gives a result 49 mm longer than the steel tape. So in this test my laser gave about 0.02% more than my "Silverline" steel tape

I know my own Silverline tape reads very slightly less than the true distance because on 23 April 1998 I had been able to compare it against Pete Riegel's Brazillian made "Stanley" tape which he had had calibrated by Stanley in the USA. In fact the result from this 1998 check was that my Silverline tape read low by 2.2mm in 30m relative to the Stanley calibration. Thus my Silverline was then reading 0.007% low. If the laser reads 0.013% high - the limit of the Bosch specification - this would be exactly predict my March 2011 result. However, I could not understand 0.035% difference found in the Mall comparison with David's NIST calibrated tape.

The solution to this conundrum was revealed on 2 July when Pete Riegel found an error in the sign of the NIST calibration when it was applied to David's tape. This discovery is described in the following section. Once we had corrected the tape calibration error for David's tape, the laser result on the Mall was only 0.01% high. This is a highly satisfactory result which leads me to conclude that my laser ranger properly used is entirely acceptible method for measuring calibration courses as an alternative to the recognised steel tape method.

Pete Riegel reviews David's data and reports finding an error in the calibration course length calculation.

On 29 June David sent Pete Riegel a copy of his and Hugh's data and asked,

" Please see attached data for myself and Hugh. The final report will contain more including photos of split locations, drop & separation data, etc. Please look over it with your microscope and let me know of any typos or errors. I am having a few others look this over for accuracy. Obviously this draft is not ready for publication."

On 2 July Pete replied to David, copying the correspondence to me,

I'm operating on generator power for the last few days. I hope for restoration in the next 2 to 4 days. Meanwhile, we roast.

I think there is an error in your calculation of the length of the calibration course. Also, I see only one measurement of the cal course. If two were taken, the other one should be included in the report.

I suggest you contact Mike Sandford, a world-class measurer and calculator, and see what he says. I am buried in dealing with the present power outage and cannot focus properly. He may also be helpful in integrating whatever data he has with that of you and Hugh. His data should appear in the report, whatever choice you make as to whose is "official."

I emailed David

2 July - Dear David, I am happy to help look at the calibration course length calculations. I remember you did suggest on the phone that I might make contact with NIST, but I said at that time that I thought fully understood the NIST report and no contact with NIST was needed.

As you know I have been studying the cal course length question rather closely when I found a somewhat larger than expected difference between your value of 328.570 and my laser value of 328.631. What disturbed me about my laser result was that the difference would increase still further if you pulled the tape at 10 lbs rather than 10 kg which the tape was calibrated at.

I have since then checked my laser technique very carefully - I have laid out another calibration course in Abingdon [see my report on the Park Road Calibration Course] using the exactly the same laser method as for The Mall. I have not found anything wrong with my laser methodology, although I have found out something unexpected about my bike tyre : it behaves like a solid tyre on rough/smooth surface even though it is a pneumatic tyre, ie on smoother surfaces it gives a smaller constant. This easily explain the difference between Hugh and me over the cobbled sections.

I did hear from Hugh that you and he made two tape measurements of The Mall and that these differed by 3 mm, which sounds fine. But Hugh did not mention the raw values, so if you send me these over I will be very happy to carefully go through the cal course length calculation.

Please call me on Skype if you want to discuss. I hope you have not suffered in the storms as Pete has in Ohio.

David sent me his data:

Mike, Please see attached data. Please respond to me only. I had another individual check my data as well, but it's easy to make errors. Thank you David

I responded:

David, I have quickly looked at the calibration course calculation You wrote:

June 13, 2012 00:45-01:15 Calibration course- On the Mall east face of lp -18 & 30 raw measurement: 328.645 meters (6 tape length x 50 + 28.645 meters)
NIST correction: @0.013m/100m 0.013m x 3.28645=0.0427238m 328.645-0.04272238= 328.60228m
Temperature Correction 11C @0.0000115* = 0.034m 328.60228m- 0.034m=328.56828 (*AIMS uses 0.0000115)
Calibration Course Length=328.57 meters

The NIST correction is in the wrong direction. The NIST report (table on page 1) says that under 10kg tension the 0 to 50 metre interval was measured by NIST to be 50.00661m, with a tiny uncertainty (negligible for our purposes). So the tape is actually longer than the scale on the tape says. You have to add the 0.0427metres which you calculated. So the length (before temp correction) ought to be 328.6877.

Now when you apply the temp correction by subtracting 0.034 m you get 328.6537 m round to 328.654 (the sign of the temperature correction is right because with a tape colder than 20C it will contract and so give a larger reading than it should.

Now when I compare 328.654 with my laser value of 326.631, my laser is now in much better agreement, Laser - Tape = - 23 mm. which I am now very pleased with since the laser error is now only 0.007%.

However caution is still required over the actual tension which you used. If the tension was only 10lb rather than 10kg, there will be hardly any tension correction required, just the temperature correction. Here is the actual calculation which I sent previously.

Tension Tape length 40m-100m  
NIST report p2
NIST report p2
4.54 kg (10lb)

I see I made a small mistake in that calculation I think it should be at 4.54 kg true length tape length= 60.00776 - (60.02946-60.0076)*(10-4.54)/(20-10)= 59.99569 So the 60 m of the tape is 4.3 mm short at 10 lb pull. 328 m of the tape will be short by 23 mm. At somewhere close to 14 lbs pull the tape will be exactly right!

When you come over again can you bring your tape so that the we can get together and compare the laser and the tape over 100 m and 10 kg pull? I think it would enable me to get a really good check on my laser - traced back to NIST standards! Since I started writing the above Pete copied to me his calculation which is for a 10 lb pull, not 10 kg.

328.645 minus the 23 mm which I calculate above= 328.622 this is close enough to the value Pete gives of 328.6227 - So Pete and I agree!

Final Agreement on Measurement Result

I then updated my summary of the three riders' measurements including calculation of the turn round adjustments and I reccommended:

Since Hugh was most familiar with the course having ridden it about 20 times take Hugh's values for the adjustment. Hugh's values are also the median values of the 3 measurements.

David agreed, in an email to Pete:

As discussed with you and Mike ( and suggested in Mike's email), I will be going with Hugh's landmark notations.

Throughout much of the period since our measurement on 13 June, David had been on duty at other events including the US Olympic Trials which were held in in Eugene between 21 June and 1 July, and therefore he had to rely on others to spend time checking the results. However, he kept in touch even while working with his Blackberry in airports and while on the plane home on 2 July:

Thank you for the very prompt work. I'm flying back from Eugene, Oregon (yes, inflight internet). There's a ton of data!

3-5 July,
Over he next three days I had a round of discussion mainly with Hugh who regretted the fact that he had not checked the cal course with a certified Japanese tape which he had. Hugh also made the interesting point that it might better to go with the manufacturer's calibration of the tape scale rather than try to apply corrections for tension. During this exchange Hugh gave me the actual raw lengths that he and David obtained, and the average of the two. I replied:

Hugh, Thanks for this, with the two actual measurements values for the cal course - first time I have seen the 328.648 figure - you told me your reading was 3 mm different but I did not know which way! The normal protocol is to take the average, as you have done 328.6465, as the base to which the corrections must be applied...... David asked me to write a one or two page summary of the tape calculations for his report, and to include mention of the laser check, which of course now confirms the properly corrected value which was obtained from the taping.

I sent my final summary of the calibration course length calculations which appears here

Hugh, Yes, except I have now recalculated using the average tape value which you gave of 328. 6465 so it gives (after rounding) a value 2 mm greater 328.597. In the next email I am sending a draft for David's measurement report as he requested. Mike

Then I updated my overall measurement spreadsheet with this final value 328.597 (see here)

Dear David, I have updated my spreadsheet with the latest value for the cal course 328.597 (2 mm more after taking the average of your and Hugh's taping, which I think will give the best estimate of the cal course length). I have left out the sheet where I did temperature corrections for my calibrations taken over two weeks, I did that when I was still trying to understand how all the numbers fitted together. But now that we have a clear, clean story for the taping and the laser check of it, I don't want to go into the complexities of my tyre's changes on different cal courses. It would not in anyway assist your report, which should obviously focus tightly on the job in hand. I think I have none done everything you asked for. Let me know if there is anything further.

Hugh revised his report using the finally settled cabration course length 328.597.

Mike, David, I attach my amended report, based on 328.597 m being the calibration course length. This is the third version (previous lengths having been 328.56 m and 328.57 m) Hugh

On 6 July Hugh sent to The Race Organisers the final version of his calculations

Lisa, I attach what I have done myself. there was some discussion over the length of the calibrations course (depending upon corrections for the laboratory test of the tape we used, and this added a few centimetres to our 328.56 m first used. The agreed figure is 328.597 m. I re-calculated using this as the basis yesterday (as in the attached report). David Katz will be submitting his own report, using these same figures (and therefore specifying the same course). I think he is near completion but was delayed by his duties at the US Trials over the last weeks. Hugh

And David to told the Race Organisers to use Hugh's numbers.

Lisa, Sorry for the delay. The "nerds" of the world (including myself) have been fine tuning the numbers! I will be preparing the "official" report over the next week which will recommend that you use Hugh's landmark notations for all required marks along the course (miles, 5k's, etc.). This weekend I will review the report that Hugh emailed you today for any inconsistencies (typos) so your team can move forward with their preparations. David

Another Newspaper Story

On 9 July ajournalist called Nicholas Spangler emailed me:

Dear Mr. Sandford, I'm a reporter for Newsday, a U.S. newspaper covering Long Island. I'm doing a story on your colleague, David Katz. Do you have time for a quick phone interview this week, or the patience to answer questions over email? Regards, Nicholas Spangler

We later spoke on the phone and I reported back to David:

David, This reporter phoned me this afternoon, asked some of the usual sort of questions, e.g. how many times does the bike wheel go round for a marathon, etc. But there was also a wider range of questions, some unexpected.

Do measurers have different styles? Answer: For good measurers the important thing is consistent riding and calibrating, but style may differ in how they deal with race directors.

What is it like being an anonymous measurer giving interviews to the media. Answer: we have experience of media enquiries when runners complain a course is short and we have to work out whether the race director sent the runners along the measure route.

I referred a question about how our 3 measurements compared to David, saying he was in charge of the official report on the measurement. Regards Mike

By means of a web search I found out that Spangler's article appeared on 4 August but unfortunately it is hidden behind a paywall. I am not going to subscribe, so I don't know if anything I said got published. After my run in with the WSJ photographer described above I decided not to plead for free access. Perhaps I had better draft an access and fair non-commercial use contract ready in case I ever encounter these journalist types again.

A further web seach revealed a facebook page containing what appears to be a cut and paste of the Long Island Newsday article.

Finally ..... the web search brought up an article on The Runners World Website about Hugh Jones's preparatory work for the official measurement.

Return to Olympics Marathon index page