Racing by time, detecting track changes, and acting on them

[Stan Pope]

I've been tossing about the question of track characteristic changes during the course of a long race-by-time session, how to detect them, and what to do about them if they are detected. Here is some thinking that I'll toss out and see if it "draws fire."

First the concern: During the course of a couple hours of racing, particularly with youngsters staging and retrieving their cars and with the possibility of "track repairs" being required, a track is apt to be jostled about. This is a minor concern with race by points, but possibly a significant concern when racing by time. It is a concern because run times from before a change are compared against run times after a change.

For this discussion, I'd like to not get bogged down in whether such a change can occur or how often or on what kinds of tracks. Rather I'd like to assume the possibility. Then try to construct a minimally invasive way to determine whether a change occurred and if so, what to do about it.

1. Establish a baseline for the lanes: I'm thinking of a calibration set using a simple lane rotation involving 4 cars to provide a profile for the track. The cars are:
A. Strong Left-handed Rail rider
B. Light Left-handed Rail rider
C. Strong Right-handed Rail rider
D. Light Right-handed Rail rider.

Prior to the scheduled race these cars "calibrate the lanes" by making one (is that enough?) run in each lane and having it's time recorded. (Maybe before raceday, we should make several such calibration runs just to establish the typical variance for these cars.)

2. Run the races

3. Certification run: Rerun the calibration set. Compare the before and after times for each car - lane combination. Are they "within tolerance?" If so, the the race results can be certified.

If the certification run results are not within tolerance, compare the magnitude of the time changes to the time difference (using the car's average times) between the top cars. If the certification run variances are less than the differences between adjacent cars in the ranking, then the race results can be certified because the change was not big enough to affect the ranking.

We may be left with (hopefully) at most a few cases in which the ranking are in doubt because the magnitude of the track change was greater than the difference between those cars. That needs to be resolved, but it can be resolved directly between those cars without involving the rankings of any others by head-to-head races.

Suppose, for instance that the cars ranked 1 and 2 had average times closer than the calibration time variances and that the cars ranked 4 and 6 had average times closer. We need to resolve the 1st and 2nd place trophies between those two, and we need to resolve the 4th, 5th, and 6th place trophies. I think that I would not try to resolve placements in question that did not involve recognition, but I'm open to discussion. For instance if the cars that placed 59th and 63rd were closer than the calibration variance, I might recognize those 5 racers (59th through 63rd) as tied for 59th!

Issues ...

1. How to compute the calibration variances? If a simple lane rotation schedule were used, a track change could affect all of the runs of cars! Car A made all of his runs before the change and car B made all of his runs after the change.) If a schedule like PPN were used, a track change would affect one or two runs. This argues that for PPN race schedules, the variances computed should be reduces proportionately.

2. What should be included in computing the calibration variance? I think that the total shift for each lane should be noted: the interval between the greatest time reduction through the greatest time increase for all of the calibration cars. Or, since we don't know when the cars raced in the various lanes relative to the track change, maybe it should be the interval between the greatest time reduction by any car on any lane through the greatest time increase by any car on any lane.

3. How would multiple track changes mess up this plan?

[Stan Pope]

Some whys on choice of cars:

I considered including a "random alignment" and a "dead-on alignment" car into the mix, but these types of alignment exhibit higher time variance according, I think, to RacerX. I believe that he is correct.

I also looked at the types of changes that are likely in a track. These were:
a. horizontal displacement between joints
b. vertical displacement between joints
c. horizontal change of direction between joints
d. vertical change of direction between joints
e. change in height at some point along track

Using both left-handed and right-handed cars should catch horizontal and vertical displacement issues.

Using both strong and light RR should make the cars sensitive to horizontal changes of direction. I think that a strong RR may be less responsive to such changes, though.

Just about any old car will be sensitive to height changes, unless the car is really lossy.

[NealOnWheels]

Stan,

I see what you are trying to do. Here are some considerations that come to mind.

1. This method does not account for events that occur between the start and the end of the race. Perhaps the track gets "disturbed" a third of the way through the race. The "damage" is not noticed until two thirds of the way through the race and is "repaired" at that time. Substitute disturbed, damage, and repaired with anything you like.

2. You would need to know the repeatability of your standards (your four test cars). That repeatability would be part of the overall system accuracy. You would need to run the test cars many times on each lane on an "undisturbed" track to get any meaningful repeatability numbers. With the likelihood of lubricant loss this may not be practical anyway.

3. Perhaps I am mistaken but weren't you a proponent for points systems and multiple elimination systems?

[Stan Pope]

Great comments! Thank you!

Stan,

I see what you are trying to do. Here are some considerations that come to mind.

1. This method does not account for events that occur between the start and the end of the race. Perhaps the track gets "disturbed" a third of the way through the race. The "damage" is not noticed until two thirds of the way through the race and is "repaired" at that time. Substitute disturbed, damage, and repaired with anything you like.

In the situation you describe, there is a known point of intervention (two thirds of the way through) which should trigger a calibration set, perhaps before (to quantify the problem) and after (to prove the correction.)

I spent some time today trying to develop some probabilities regarding multiple unrecognized disturbances. Nothing except a big spreadsheet full of "RAND()" functions so far! My poor old memory niggles me with the thought that random numbers formed from the sum of six uniformly distributed random numbers will have a normal distribution. But, I haven't been able to find anything to substantiate that memory. And normal distribution is a better representation for the kind of track behaviors we're likely to see.

2. You would need to know the repeatability of your standards (your four test cars). That repeatability would be part of the overall system accuracy. You would need to run the test cars many times on each lane on an "undisturbed" track to get any meaningful repeatability numbers. With the likelihood of lubricant loss this may not be practical anyway.

Yes! The question is how many are needed to establish a useful baseline when using well-behaved cars (such as those that I described)? I should go look at a bunch of posted league times to get an idea of repeatability of such designs... just to help decide if the goal is feasible.

I am also thinking that I can add use a simple, quick to use staging guide to eliminate much of the variance due to staging... not the "fork" that Michael L. described, but a simple "align one side of the car to one side of the rail" guide... like a 3" thich try-square with really short legs.

I should probably try to suck some of the "6 Sigma Blackbelts" into the analysis as well... since this whole issue is right in the middle of their sphere of expertise!

3. Perhaps I am mistaken but weren't you a proponent for points systems and multiple elimination systems?

Still am! Well, actually, I am a proponent of using whatever system fits the available equipment and the competition requirements and will provide an absolutely fair and acceptably accurate result. I don't believe that a pack should skip racing this year or accept poor accuracy just because it doesn't have a reliable timing system on the track(s) that they can use!

So, if a group has a track with timing electronics and a manually operated gate, I'm in favor of points or multiple elimination. Definitelyl not time. Replace the manually operated gate with a consistent "snap open" gate and confirm that the track is stable and if I don't have to worry about equipment malfunction, then timed competition is great, too.

The extreme case is a district, council or regional race. If it is to take place it must take place on the scheduled date. If it is called for whatever reason, not only do all the registration fees get refunded, but we are also out most of the expenses in organizing the event. The event has to go, regardless of what goes wrong. The more "stuff" you depend on to work right to make it go, the more chances Murphy has to cut you off at the knees!

Right now I have 5 PCs humming away at some scientific computations on my home network. Three on this desk (a 2GHz P4 and a couple of duallys), one on my brides desk (a 3GHz P4 with HT) and a laptop (believe it or not) in the bathroom, although I can move it around the house as needed! (Avoids separation anxiety, you know!) Mostly they can run for months without attention (although with the frequency of MSoft patches, I end up rebooting 'em monthly.) They only fail when I depend on them to work! Murphy virtually guarantees that! The 6th computer, normally somewhat faster than the one I'm typing on now, is "casters up" waiting for me to figure out how to repair it (or to replace some innards with more modern technology, i. e. "quad-core"!)

[Stan Pope]

My poor old memory niggles me with the thought that random numbers formed from the sum of six uniformly distributed random numbers will have a normal distribution. But, I haven't been able to find anything to substantiate that memory.

This is close: The sum of 12 random numbers, each uniformly distributed between 0 and 1, less 6 approximates random numbers with a normal distribution.

Where was Google last time I tried to substantiate that old memory? Oh ... I remember ... Google hadn't been invented yet!

Now, I wonder how good that approximation is ...

[NealOnWheels]

Another idea, in addition to a test run at the beginning and at the end, perhaps scheduled test runs mid race such at the 25%, 50%, 75% points. That way an unrecognized disturbance can be detected sooner. I guess that would only make sense if the cause of the disturbance is likely to be determined. The middle of the race is not the ideal time to be doing any troubleshooting.

...and a laptop (believe it or not) in the bathroom...

That gave me quite a shocking mental image: Stan enjoying a whirlpool bath while browsing the Derby Talk website... ...Got to get that one out of my head.

[Stan Pope]

Another idea, in addition to a test run at the beginning and at the end, perhaps scheduled test runs mid race such at the 25%, 50%, 75% points. That way an unrecognized disturbance can be detected sooner. I guess that would only make sense if the cause of the disturbance is likely to be determined. The middle of the race is not the ideal time to be doing any troubleshooting.

Good thoughts. Thank you!

A modest disturbance might simply be documented. I think that some "rules of thumb" can be developed to provide guidance in this decision. With the magnitude "known" (at least approximately), its worst case effect can be included in an "uncertainty band". When racing is done, if the uncertainty band is smaller than the difference between the racers, there is no need to do anything more than hand out the trophies!

Even if the cause of the disturbance is identified and corrected, it is likely that there is a residual effect on the competitors times that can not be "corrected."
The trick is to figure out how much of an intrusion is worthwhile, both to the process and to the competitors.

On the remainder of your post, I have no comment. I've said way too much about that already!

[BlackLotus]

I have been rolling this around in my mind for a couple of days and I just can't get past all of the variables. The short answer seems to be, there is no reasonably reliable, and practical, method to monitor changes to the track.

I'm pretty sure we can all agree that every time you run a car down the track, the car is changed, however slightly, by the process of the run. There will be lube lost or moved, the wheels will wear some, the axles will wear some, the wheels will "warm-up" etc.

The same is true of the track. The simple act of measuring the track (by any mechanical means) will cause the track conditions to alter. Even if we take all human factors out, running a car down the track will cause some wear, however slight, and will probably leave graphite and/or other foreign matter on the track. Heck, even the airflow in the room will cause changes.

I think the only reasonably semi-solution, is the do everything reasonable to prevent large changes to the track conditions by controlling access to the area. Other than that lame idea, I'm stumped.

[pack529holycross]

I think that the value of a set of "control" cars is truly a great idea when it comes to the basic step of "track calibration" on its own merits. We really have several specific issues with our track, which is totally brand new this year. "calibrating" the surface of the track is, I feel, the best way to reduce adverse environmental effects on the car's performance. Beyond that, the only thing I think you could do is to run "calibration" runs between Race Groups while awards are announced to watch for problem areas in the track surface. I think that it would be really "cart before the horse" to run races on a (potentially) misaligned track, then try to nullify or asterisk the results by "certifying" the track afterwards.

I KNOW everyone hates the DragRacing analogies I use here, but think about it... governing race bodies do everything possible to "prep" the track for raceday, and in the event of a major incedent, every attempt is made to nullify or lessen the impact of any potential environmental changes that occur during the event itself. I would venture to say that raceday track repairs in the real world is weighted more towards safety than competitive advantage/disadvantage, and perhaps in the event a car jumps lanes or off the track itself, then there could be a cursory examination of the location of the "wreck".

In the end all we can do is all we can do.

Nicholas

[Stan Pope]

Thanks for your thoughtful consideration, guys! Much appreciated. In this case, the "drag racing" analogy is reasonable.

Something for me to work on during the next few snowed-in weeks!

[3 Cub Dad]

From an engineer who manages a manufacturing facility...

The only way to practically implement such a system would be to run the known test cars at routine intervals during the derby.

Call the test cars the "standard", for which you have a predicted result, (time). In the lab, all tests are run bracketed by the standards, (standards are tested before and after the "test" material). If either the pre-test or post-test standards fail, the results are thrown out. On the plant floor, during manufacturing, "standards" are also checked. If any parameter is found out of tolerance, anything produced after the last standard check, but prior to this standard check is suspect and rejected.

Translating this to a derby run: The "standard" cars would be run to verify the track, then racing begins. After a set number of runs, lets say 5, the "standard" cars would be re-run. If they are within tolerance, everything continues. Lets say the check after the 10th race is also good, but the check after the 15th race is out of tolerance. The results for races 11-15 would be discarded, the track fixed, re-verified with the "standard" cars, and the races re-run. Depending on how tight your process control is, (ie the track set up), you could run more races between checks, but you risk more results being throuwn out!

Interesting thought, but may be a little difficult to explain throuwing out blocks of runs.

[Stan Pope]

Thanks, 3 Cub Dad! More thoughtful input!

The relationship to manufacturing floor tolerances and standards is quite apt, I think. A possible difference is that the actual competition results (tightness of some average times) may dictate the allowed tolerances. That is to say, an observed variances from the standard of 0.005 sec matters if the difference between two cars is 0.001 sec, but is immaterial if the difference between them is 0.010 seconds.

And, then, I think that I have two options: "Toss and rerun the heats in question" or "resolve the inconclusive results by head-to-head." The former choice preserves the concept of "track record" (for those so inclined) and an environment necessary for long term track records is more akin to manufacturing product to tolerance.

I might get my arms around this if I'm "snowed in" until mid-May!

Sometimes I wish I were an engineer! But then I think about how much fun mathematicians have! And we aren't expected to wear those pocket protectors,

[davem]

Hmm..

A "what if".

What if you detect a track issue (after heat 15), "fix" the track, and re-run heats 11-15.
And the cars receive slower times...or...SOME of the cars receive slower times.

Now you've got a pickle.

I have re-run heats before (with both times and points methods).
Points is generally no problem, but times...you could be accused (rightly or wrongly) of making the track worse.

Fixing bad joints can make it better...but "what if" it ends up worse...and you can't get back to a repeatable setup (i.e. can't get the standard cars to get back within allowable times)??

[Stan Pope]

A "what if".

Ouch! Well, I didn't claim that this was an easy task!

And, your "what if" is certainly a case to be analyzed in detail with whatever plan is developed.