News from Big White Rally 2011??? Where?

In the events that I have been steward at in Canada, I have seen scoring by tenths produce a different result from scoring by seconds at two other events.

alan

Timing in general has always been a thing we seem to have simple blind trust in its accuracy and consistency.

But I have personally been in several cases where I had caught the car in front--one time with a 2 minute gap on a 18 mile stage----and when the car in front would not pullover, followed them literally less than 4-5 feet off their bumper ( not being an ass but trying to stay close so rocks go under the car and wouldn't go UP and smash my lights, windscreen and radiator, and so dust---or snow in one case---wouldn't billow up and blind us) and then recieve a time showing 5 seconds difference for less than 5 feet difference..
Well at 75mph you go 110 feet/sec so 5 seconds would mean 550 feet..

550 feet is over a 1/10 of a mile... My car is 15 feet long.

Something was obviously screwy....

Inquiries resulted in change to 1 second..

Hmmmmmmm 1 second still was 110 feet....

Did it make a difference? Can't remember but it graphically illustrated a gross error in timing technique.

---

John Vanlandingham
Sleezattle, WA, USA

Vive le Prole-le-ralliat

www.rallyrace.net/jvab
CALL +1 206 431-9696
Remember! Pacific Standard Time
is 3 hours behind Eastern Standard Time.

Quote
Slowwpoke
results page?

http://www.bigwhiterally.com/results/

Quote
Anders Green
In the eight years, I found that not having tenths would have created one tie out of a second and third place. I don't feel that changing one result in a decade is a big difference.

Somewhere in the results of just about every event we've used our ECM timing gear you'll see defined gaps that timing to the tenth enables. It could be seeing Pat and Antoine separated by 2 tenths after a 34km stage or it could be seeing a competitor break the top ten for the first time by 5 seconds after 10 stages.
This is something the competitors of the region wanted and funded and is a direction that has since been followed in Ontario.

Quote
Anders Green
The more stages you have, the less likely the standings with tenths will differ from those without tenths.

I ain't no staimatician - but in this case the effect of rounding is cumilative isn't it? with a 1 second margin of error in each stage then after 10 stages your represented time could be as much as 10 seconds off of your actual time... no?

Quote
tipo158
In the events that I have been steward at in Canada, I have seen scoring by tenths produce a different result from scoring by seconds at two other events.

Yes. This anecdote doesn't change the statistical implications of what I said.

Timing to the tenth increases the resolution of the timing. The impact on the accuracy of the results is not zero, but approaches closer and closer to zero as the number of stages increases.

The bolding is not for emphasis, but to note that the four things:
resolution
accuracy
timing
result
are each a very different thing.

Cheers,
Anders

---

Grassroots rally. It's what I think about.

Quote
john vanlandingham
Did it make a difference? Can't remember but it graphically illustrated a gross error in timing technique.

If you have real, live people involved you will always have 'problems.' The electronic system we us now still has user input data both on the timing lines and moving into the scoring system... so there is room for errors.
Until we employ professionals in timing we need to accept that but I can tell you that the crew of absurdly dedicated volunteers in RallyWest have made the ECM sysytem reliable and trustworthy.
Regardless of how much we trust the times are done right, checking the results and inquiring something that isn't right is part of how the competitors help the organisers ensure the results are accurate. If your result is wrong but doesn't change the results if corrected might be the data-point that shows a cascading error that does change the results elsewhere. (If it doesn't bother you, an informal comment works as well as hanging paper does.)

As a side note, Adam, we looked at Kelly's times on ss1 in detail and it looked legit. when we tried out normal flier adjustments other issues came up.

Quote
Anders Green
The impact on the accuracy of the results is not zero, but approaches closer and closer to zero as the number of stages increases.

Can you explain that to me - it seems counter intuitive.
If every time I put 65 cents in my piggy bank the counter rounds up to a dollar. and I put 60 cents in 10 times, I won't have anywhere near $10 in the piggy bank let alone having $9.60 that would be seen as $10.
But, If my piggy bank sees 65 cents as either 60 cents or 70 cents, when I've put 65 cents in ten times I'll expect either $6.00 or $7.00 and not be majorly bummed when I pull out $6.50

Accuracy is certainly different from consistancy and resolution. For the record, the system we use has the finish triggered by an IR cel and actually resolves to 1/100th but we drop the hundredths as we haven't seen many ties to the tenth on a stage (although there was at least one at Big White)

Quote
Anders Green

Quote
tipo158
In the events that I have been steward at in Canada, I have seen scoring by tenths produce a different result from scoring by seconds at two other events.

Yes. This anecdote doesn't change the statistical implications of what I said.

And statistical implications mean nothing to the individuals involved in any given close race.

alan

Quote
Morison
(timing to the tenth) is something the competitors of the region wanted and funded

It's excellent that they chose what they wanted and then made the effort to make it happen.

Quote

I ain't no staimatician

Grant should be the final arbiter of any real statistics questions... I've forgotten more math than most people ever learn, and I know he's much fresher and closer to it.

Quote

but in this case the effect of rounding is cumilative isn't it? with a 1 second margin of error in each stage then after 10 stages your represented time could be as much as 10 seconds off of your actual time... no?

The answer is not one of yes or no, but of probability of happening.

Timing to the second is technically throwing away data, not rounding it. The data you haven't recorded is unknown. But, with enough stages, it doesn't matter.

Here's your simplified thought experiment to explain it:
The rally runs 100 stages, each of varying and unimportant length. The scoring is to the second, but there's a group of tricksters working the finish controls. For every car at every stage, they flip a coin, and if it's heads, they add a second to the time, telling no one, but marking the cards/logs with the added second. If it's tails, they record the correct time.

After 100 stages, it's very very likely that every car has 50 seconds added to the time. The result, that is, who won, who came in second, third, is entirely unchanged even with this intentional noise added to the data.

Cutting of a fraction of a second (on average, half a second) has an identical end result to the above example, even if you don't know how large or small the amount is.

As I said before, the benefit of timing to the tenths is small but not zero, and the benefit decreases as you add more stages.

If the goal of rally was to set stage records, timing resolution would be much more important. It isn't though, the goal (of scoring) is to rank the racers in order, fastest, second fastest, etc...

Cheers,
Anders

---

Grassroots rally. It's what I think about.



Edited 1 time(s). Last edit at 12/06/2011 01:05PM by Anders Green.

Quote
tipo158
And statistical implications mean nothing to the individuals involved in any given close race.

Aaaahahahaha! Excellent.

Then you should be screaming for timing to the 10,000th or better and decrying all existing results as invalid! One day, somewhere, there will be a tie with tenths! Ignore what math says about the sport, and quick, call up the USBSF! I'll let you borrow my phone for the call Alan.

Yes, statistics mean nothing to the individual. But for the people making decisions about infrastructure, policies, procedures, and technology, if the statistics are not understood, the result is a mess.

Cheers!
Anders

---

Grassroots rally. It's what I think about.

As with most things, you have to find a balance. Here it is precision vs. the cost to get that precision.

I don't understand why this discussion is even happening. Rally West has timing equipment that is reliable to the tenth of a second. The rules allow for events to be timed and scored to tenths of a second. At this point, there is no additional cost for the precision.

Sure, over a large enough number of stages, statistics will eliminate the difference between timing to the second or the tenth. But rallies don't have a really large number of stages. Modern rallies usually only have a couple dozen stages max. Big White had 10.

alan

Quote
tipo158
At this point, there is no additional cost for the precision.

Not entirely true.
There is a $25/car levy for timing gear maintenance
There is a cost in high level volunteers to be in charge of the gear

BUT... these are costs the majority of grassroots competitors support and actively wanted. This wasn't actually led by the top end driver, the idea came out of the middle of the pack.

But yes, the timing discussion has nothing to do with Big White or how awesome it was.

It's a little more complicated probabilistically than I can think about right now.
Each stage has a possibility of being +1, 0, or -1.
Every possible result over 11 stages has a probability of 0.000005. So probability of having 11 seconds one way or the other is zero as there is only one way to get there.
Probability of having 0 for the sum of errors is going to be the N*(0.000005) where N is the number of potential ways to arrive at 0 (IE. 1, -1, 1, -1, 1, -1, 1, -1, 1, -1, 0)
Expected value of the error is zero.
With enough stages, the central limit theorem would jump in and save the day with total amount of error being approximately normal.

But let's just say for ten stages...
The exact probability of being within 2 seconds of your actual time is 0.4795.
Within 3 seconds is 0.626
Within 4 seconds is 0.747
Within 5 seconds is 0.845
Within 6 seconds is 0.918

It gets even more complicated if you were to try to calculate the probability of two drivers having the probability of changing position. Basically there would need to be enough stages and a minimal number of seconds between them.
The easy way would be to assess the number of stages (n).
Then look at how many podium positions were seperated by at least 2*n seconds.
Then you could calculate the probability that the results were incorrect.
From what I recall, there's probably minimal events that this could even be possible. Anyhow, gotta go back to RT-PCR and gene expression analysis. Thanks for the mental distraction...
At the end of the day, it depends on what it costs for that precision. If minimal then go for it. If people are potentially winning money then it's possibly more important.



Edited 1 time(s). Last edit at 12/06/2011 02:42PM by NoCoast.

Quote
Morison
As a side note, Adam, we looked at Kelly's times on ss1 in detail and it looked legit. when we tried out normal flier adjustments other issues came up.

We were taking split times from our location on SS1, we were about 150 yards from the cattle guard, Kelly did have the slowest split time out of any competitor/ I would find it very hard to believe that he made all that time up and then put almost 10 seconds on Max in that very short amount of stage that was left.

Quote
Billyee
We were taking split times from our location on SS1, we were about 150 yards from the cattle guard, Kelly did have the slowest split time out of any competitor/ I would find it very hard to believe that he made all that time up and then put almost 10 seconds on Max in that very short amount of stage that was left.

Either way, it didn't affect the pointy end of the rally, just interesting to note none of us bothered to correct it