Allowance for Error in Validation Measurements

I have a question:

When doing a validation measurement is the SCPF used? In other words, if the measurement was perfect, would the acceptable minimum be 10,005 or 9,995?

No SCPF is used in a validation measurement, and the average constant is used. If the initial layout measurements (done for initial certification) and the validation measurement are all done perfectly, a 10 km course will measure out at 10010 meters.

The goal is to ascertain the actual length of the course, in real meters.

As I have said before on this board,the SCPF is illogical and confusing. It would be better if courses were measured to their true distance, and a slightly greater allowance than 0.05% be made for normal measurement variation during validation. Of course now that the SCPF is so firmly established, this is going to be difficult to do. Perhaps one way would be to adjust old records by 0.1%.

I now understand as to how the validation process was started based on Pete's statement.

However, I feel that we only accept records when the validation comes out to at least the desired distance. Does anybody share this view?

I would share Gene's view if validation measurements were without error. Since they are not, some allowance for error is appropriate.

As an extreme example, let's look at a 10k that validates out to 9999.9 meters. What's to be done? Tell the runner that we have shown his course to be short, when all we have established is that is very close to the mark and we can't really say either way? Could be over, could be under - we can't tell.

I don't want to be the one to enforce a rule that ignores the reality of measurement accuracy. Rule 185 wisely requires a degree of proof. Zero negative tolerance does not provide this.

THE GRAY AREA
There is a gray area, or area of uncertainty, associated with every measurement. The center of this area is the exact value obtained by the measurement.

Using 10 km as an example, let’s say a validator measures the course and obtains a value of exactly 10000 meters. It is highly unlikely that the course is actually exactly 10000.000 meters in length. It is either somewhat longer or it is somewhat shorter.

We would like to know for sure. What do we need for a reasonable degree of certainty?

After extended discussion we concluded that it was reasonable to say that the gray area is ± 5 m in 10 km. In other words:

If we remeasure and obtain 10005 meters we are reasonably certain that the course exceeds 10 km in length.

If we remeasure and obtain 9995 meters we are reasonably certain that the course is less than 10 km in length.

If we remeasure and obtain 9995 to 10005 we are uncertain as to whether the course is longer or shorter than 10 km.

What we do with these numbers is not ours alone to say. The Records Committee has an interest in this, and their interests, and those of the athletes, also have weight.

I believe that what we have is about as good as we are likely to get.

I side w/Pete in this discussion, feel he's explained things well and don't agree w/anything Neville has to say on the subject. There's nothing illogical or confusing about the SCPF; it's there to help ensure courses are long enough. On all points, we acknowledge our tools aren't perfect and Pete illustrates how to handle the 'gray area' succinctly.

So the SCPF ensures that the original measurement won't be short. The Hypothetical 10K, if it was measured perfectly without SCPF, would be certified at 10,010 meters once the SCPF is factored in. In other words, you'd move the finish line to compensate for any error you might have made during the ride.

A perfectly ridden validation ride of the Hypothetical 10K would show 10,010 meters on the nose, assuming the course was laid out using the SCPF.

However, in order to pass validation, the course only has to measure 10,000 meters, since this is the advertised distance.

Have I got it right?

Almost right. In a perfect world the course validation would pass at 10000 meters. To allow for error in the validation measurement we grudgingly allow 5 meters for potential error. Thus the course would pass at 9995 meters.

One of the reasons as I see it for giving the 5 meters for potential error is how close to the kerb or edge of the road a measurer rides the corners.

In the extreme(?) case of a completely circular 10,000 meter course, the radius of this circle would be 1,591.55 meters. If the validation measurer did not ride 30 cm from the edge of the (kerbless) road but right on the edge, then the radius would become 1591.25 meters. This would result in a measured course length of 9,998.12 meters.

Scott:
You only have to look at the above posts to see what confusion the SCPF generates even with very experienced people. What we illogically call 10-km records are really 10.01-km records with a shortage allowance of 0.015 km.(IAAU allows only 0.01 km, but will consider allowing 0.015 km next March.) Far clearer it would be to measure 10-km courses to 10 km and have a shortage allowance of say 0.01 km (0.1%)

How the SCPF Came to Be

In 1981 I was just starting to measure. At that time Ted Corbitt was the certifier. Ted was working closely with Ken Young, who was keeping the road running records at that time. The records had not yet been absorbed into USATF (then TAC), and were unofficial. A few validations had been performed and it was found that about half of the records had been set on short courses. A solution was sought. Ted decreed that in future all courses should be laid out 1 m/km longer than the nominal distance, so that records would be set on courses which were at least as long as the nominal length.

At that time all courses previously certified were decertified.

In 1982 I was appointed Ohio certifier. I designed forms which were much like what we have today. They were based on the forms which Ted used, but had separate sheets for calibration, measurement, and course and race general information.

Although it was not difficult to add the extra 1 m/km to the measured course, the splits also needed adjustment, especially in the longer courses. In my calibration data sheets I asked the measurers to add 1 m/km to their constant for the day, just as is done today. This put the splits, as well as the overall length, in their proper places.

Course measurement does not exist on its own. As far as USATF is concerned, course certification exists to serve the record process. A guiding principle, oft quoted by Ken and Ted, is “A course shall not be short.”

This is why we use the SCPF. Some may be confused, but I have not found that measurers have a difficult time using the present process.

Those wishing to adopt Neville’s suggestion will have to confer with the Records Committee and see whether they are ready to see a large number of records set on demonstrably short courses. I suspect they will greet the suggestion with reluctance.

When Ted declared that all courses be measured with a SCPF of 0.1%, he did not change the error of measurement, but instead shifted it from a course of say 10.00 km to a course of 10.01 km. By pretending that this longer course was only 10.00 km, the measurement error was effectively concealed from most people.

I am not clear as to why an additional shortage allowance of 0.05 % is necessary, but perhaps despite the shortage allowance of 0.1%, some did not like to see a course fail validation when only a meter or two short of the nominal distance.

The guiding principle should be more accurately stated as “A course shall not be short within the normal error of measurement”. My proposal would be to have courses measured to their nominal distances, and pass validation if they proved to be within the normal error. If this normal error is accepted at the present 0.15 %, failure rate would be no different from now.

Note that I am not proposing anyone try to change the present system – it is too firmly entreched!

Neville, I think you are fighting the problem. We are tasked with providing record-quality courses, as defined by the Records Committee. They want, and have always wanted, courses that are at least the nominal distance. Ted's approach neatly solved the problem.

If it gives us mild misunderstanding on validations, which amount to less than one percent of our courses, it's something we can cope with.

Pete:
I am not fighting the present system because as I say it is too firmly entrenched, and my discourse was just to explain to Scott as to why I think it is confusing. (It might be nice though if the IAAF and the US could get into agreement on the 0.05 % factor – either the IAAF should adopt it or the US should abolish it. Support for the latter could come from the abhorrence I see for validation of courses at less than their nominal value.)

Probability

Consider a 10 km course which, because of mistakes in layout procedure, is exactly 10000 meters in length. A record time is run on the course.

If we set ten expert measurers to check it, about half will find it to be less than 10000 meters, and half will find it greater. Who do we believe?

Since we use only one validator, it’s then a coin toss as far as recognition of the above record is concerned. The USATF rule requires us to show that the course is short, or accept it as valid. The result of a coin toss is not persuasive evidence.

Any single measurement lying between 9995 and 10005 meters is inconclusive as far as answering a question of longness or shortness of the course. We could wish it otherwise, but there it is. The exact value of any length is always unknown. All we have is probability. The negative allowance allows us to give a strongly and probably correct answer to the question of shortness.

I would hate to explain to informed questioning the reasoning behind disallowing a record when a tiny shortness is indicated by a single measurement.

This was on the agenda for the recent IAAF Technical Committee Meeting.
The IAAF Tech Committee STRONGLY was against adopting this. Their policy is not allowing any measurement of a track or road distance being anything less then the required distance.
Just as an added note: The Committee voted 15-1 against adding the 5k for road race records. Guess who was the only one who supported it?

IAAF can deal with the fallout when a marathon remeasures at 42193 meters and is shot down for shortness. They are thinking with their hearts, not with their heads. They require education in measurement accuracy.

As far as I know, IAAF does not yet have a rule similar to our USATF Rule 185.3, which contains specific language which governs how measurements should be interpreted.

As IAAF road course measurements are generally done before the race, by an approved IAAF measurer, their stance may make sense. If applied to a post-race validation, it does not.

Pete
What is the opinion of the other top measurers from around the world....Aims world coordinators? Hugh Jones and company are the ones that have contact with the IAAF at their annual meeting (after the World Road Racing Championship).

I haven’t a clue what others may think, aside from those who participated in the US exercise two decades ago, and the opinions expressed on this bulletin board.

USATF began with a rule which required us to show that the course was short, or accept it. The relevant single measurement was the one done after the race was run. I believe we got it right. This was done long before IAAF got into the road running game.

In the IAAF case, they take the result of a measurement by an accredited expert. He will lay out the course 1.001 oversize, just as we do. From this point on the course is accepted as not being short. As far as I know only a single measurement of a road course is required.

What IAAF will do when an after-the-race validation measurement is done I don’t know. That is up to them.

One would hope that when or if IAAF asks the opinions of people on the subject, they pick people who know something about measurement error, and not just take a general poll.

My view is that Pete and those who formulated the USA procedures some two decades ago defined a good approach to validation procedures and the 0.05% allowance. Just as the SCPF is a very simple rule for all measurers to use and which helps ensure that distance run is not less than the advertised distance, so the 0.05% validation allowance makes provision for variations in a validator's measurement (as described in Pete's postings in this thread).

However, I note that not all experienced measurers are entirely comfortable with the 0.05% allowance for the validation. Three instructors discussed the topic with UK Athletics grade 2 measurers at a seminar in 2006 for grade 2 measurers who were undergo a measuring test to demonstrate their capabilities as part of the process of being upgraded to grade 1 (which can be considered as approximately equivalent to IAAF grade B in measuring experience) Just as on this message board, there were varying views about whether the 0.05% validation allowance should be allowed. After the discussion measurers then did a practical exercise with the following result:



My view is that these results show that the 0.05% validation allowance and 0.1% SCPF are well chosen. If the 0.05% validation allowance were ever to be reduced to zero then I would want to see the the SCPF being increased to 0.15% (and perhaps even a bit more) in order to avoid penalising record runs purely on account of typical measurement errors. (I speak of course about unavoidable measurement errors rather than measurement mistakes.)

The measurement exercise was held on a 4533.2 m course near my home which has been measured many times over the last ten years. The course was deliberately chosen to provide some measuring challenges, e.g. plenty of corners requiring careful riding exactly 30 cm from the kerb and a few vehicles parked on the SPR. Over the years IAAF grade A, UKA grade 1, and experienced grade 2 measurers have obtained results varying from 4528.8 to 4538.8.

Probably uniquely amongst all such courses used for such trials, I have obtained the exact length of the SPR by carefully measuring the whole course with a steel tape. (Yes, techniques were used to ensure the measurement was made exactly 30cm from the kerb on all corners. Over 100 taped lengths were needed, each reading being double checked.) So we know the true length of the course to be 4533.2m. The measurer's bicycling results thus all fall with in a range which extends from -4.4 m to + 5.6 m relative to the steel tape result, i.e -0.10% to +0.13%.

This range just about meets the requirements of IAAF rule 240 (3) "The uncertainty in the measurement shall not exceed 0.1%". The range observed may seem surprisingly large. I think the main contributors are:

1. Not following the SPR really accurately.
2. Variations in surface roughness causing changes in calibration constant.

There is a fuller discussion of all these aspects in the .pdf file containing the SEMINAR REPORT

All the figures quoted above do not include the SCPF, but we can see that with the application of the 0.1% SCPF then just about all the measurers would have laid out a course which was at least 4533.2 m long and would have passed validation by my steel tape (without making any allowance for validation measurement errors). Of course once I use a bike for validation then I need an additional allowance for the inherent errors - or a larger SCPF. Let us stick to Pete's formulation which he has explained very clearly.

As far as the IAAF rule book is concerned my view is the rule book does not prescribe a validation allowance-- it merely says that for a World Record (Rule 260.8 e) The course must be verified on site (i.e. within two weeks before, on the day of the race or as soon as practical after the race), preferably by a different A or B measurer from the one that did the original measurement. The IAAF rules are silent on any allowance the verifier should make for his own errors. So at present it seems up to the the verifier to decide how to proceed and there is scope for disagreement between measurers! This could usefully be clarifed in the IAAF rules.

I am not clear as to why an additional shortage allowance of 0.05 % is necessary.[/QUOTE]

Neville, as I said a few days ago and others (Mike & Pete) have provided explanations & anecdotes to illustrate, we allow the shortage because our tools are imperfect and the measurer/validator is similarly challenged. In a perfect scenario, there would be no SCPF or .05% allowance. We acknowledge our limitations and, over the years, the little we add and 'allow' has held up satisfactorily.

If there has been confusion on these points, I hope this thread has helped clear things up.

I've done a few validations and none of them has gone 'smoothly'. There's always some wrinkle that needs addressing, potentially affecting our outcome. Without going into detail, we size up the situations and make our best call to handle them. Ah, if only all our roads were perfectly smooth, traffic free & no deviations from the certified path.

Advocates for the 0.05 % shortage allowance claim that we need this to cover normal validator error. Of course this view loses sight of the fact that we already have the 0.1 % SCPF for this.

Mikes view that, if the 0.05 % allowance was eliminated, that the SCPF should be correspondingly increased to 0.15 % has merit. Complexity would be reduced, validation pass rate would be exactly the same as now, and those who abhor passing a course at less than the nominal distance would be made very happy. Unfortunately it would condemn runners to run longer distances. On a so-called 10-km course, distance run would increase from the present average of 10.010 to 10.015 km.

The best option would be to eliminate both the 0.05 % and the SCPF of 0.1 % and replace them with a shortage allowance of 0.15 % on courses measured to their true distance. Confusion, pretense, and complexity would be totally eliminated, validation pass rate would be exactly as now, and runners would on average run a true distance.

Of course, jeopardizing a record through validation can always be avoided by doing the validation beforehand.

(Mike: I note that your tape measure is shown as 4533.5 m in your chart but as only 4533.2 m in your text. Also, your calibration course is ideally located as a large part of your demonstration course. If the calibration course had been at a distant location as frequently happens in practice, I anticipate that your range of -0.10 to +0.13 % would have been greater; and, because precalibration courses are usually measured at lower temperatures, favored the finding that the course was shorter than it actually was.)

While your proposal to raise the SCPF to .15% is well-meaning & has some merit Neville, it's unnecessary. Our mandate & goal as measurers is to get the distance as close as we can to what's advertised and we get satisfactory agreement & acceptable validation numbers using .1% as the SCPF (along w/the allowed .05% shortage). We add and allow, as I mentioned above, because our tools & methods are imperfect.

Your proposal would further ensure courses would pass validation but I don't feel it's needed.

While Neville’s proposal concerning how courses should be measured and checked may have merit, it does not address the fundamental philosophy of record-keeping. Ken Young, who more than anyone created the US system of record-keeping, has stated the fundamental definition of a record as follows:

A record is simply a statement that the runner ran AT LEAST the specified distance in a time NOT SLOWER THAN the specified time.

Our measurement philosophy is governed by the needs and desires of USATF’s Records Committee, and not the other way around. How we treat layout and validation must fit the needs of the record-keeping system. As we presently do things, it does.

As for a 10k course not being exactly 10,000 meters, this is a matter of definition. On a curbed track the measure line is 30 cm out from the inner curb. This is an arbitrary figure which is deemed to fit the way a runner will run. It could as well be 40 cm or 5 cm. It is a rare track runner who will complete a 10,000 meter track race without running a course longer than the defined path. On roads we have defined the path, and we have set the procedure so that we assure that the course will not be less than the nominal length.

Until such time as the records-keepers change their philosophy we are constrained to use methods which fit their needs. I believe the present record-keeping philosophy is a good one.

Scott: my preference is not to raise the SCPF to 0.15% but to eliminated it altogether and increase the present 0.05% allowance to 0.15%.

Logically the statement “AT LEAST the specified distance” should not be taken literally and should imply “within measurement error”. In fact, I think that this has been acknowledged with acceptance of the 0.05% shortage – so called 10-km courses are accepted for records if they validate at 9.995 km.

Neville, again, our mandate is to try to provide courses that are at least the nominal/advertised distance and we're far more likely to achieve that goal w/the time-proven .1% SCPF than by eliminating it. It doesn't take much measuring experience to realize my version of the SPR may differ and/or my attempt to ride same may differ from the next measurer, so adding a tiny built-in cushion
helps ensure we're long enough.

Our current standards are easy for measurers & the public to understand & I'd suggest your proposal would be less defensible to both parties. How to explain a 10000m course found to be 99985 meters on validation is okay for records?

Scott: I don’t know how you can say that our present standards are easy to understand when experienced measurers like Gene Newman and Paul Adams above are confused by them. There is no difference in quality control between a course that is supposed to be measured to 10,010 m but passed at 9995 m, than one measured to 10,000 m and passed at 9985 m. However by pretending 10,010 m is only 10,000 m, we deceive the unsophisticated and those unaware of the SCPF into thinking that our tolerance for error is far less than it really is.

There are so many that have commented on this issue and I hope it has helped many to understand how the SCPF and the validation process was born.

I am not confused as Neville states. I did feel that records set on a course where a post validation came up short of the distance should not count. Seeing some of the past history as to how all this has come about has made me feel the process we are using is fair to the runners.

When a course is post validated short of the distance, but within the tolerence then the record is accepted. This course must be remeasured for any future records. This seems fair to me at this point in time.