Page 1 of 2

How to tell if you have a "Slow" Lane

Posted: Thu Feb 15, 2007 7:32 am
by gpraceman
Ever wonder if you truly have a "slow" or "fast" lane? Just looking at average times and standard deviations for each lane may not really indicate if there is something that needs to be addressed with the track. If you really want to know if the difference between two sets of data (e.g. Fast Lane vs. Slow Lane) has a "statistically significant" difference, then there are a couple of statistical tests that can be performed.

t-Test - Comparison of averages of two data sets
http://en.wikipedia.org/wiki/T-test

F-Test - Comparison of standard deviations of two data sets
http://en.wikipedia.org/wiki/F-test

For those of you with your heads spinning already, you can use this simple spreadsheet to perform the two tests.
http://www.asq.org/sixsigma/tools-excha ... f-test.xls

For best results, use data on the faster cars, since slower cars can have a great deal of variation due to car issues (as opposed to track issues) and will skew the data.

Re: How to tell if you have a "Slow" Lane

Posted: Thu Feb 15, 2007 7:47 am
by DerbyAddict
I knew I shouldn't have gotten into this thread.... Now my head is starting to leak.




Must... find... more..... duct..... tape.........

Re: How to tell if you have a "Slow" Lane

Posted: Thu Feb 15, 2007 9:12 am
by BigDozer66
I am just 'tagging' this for the great ideas once the track gets here! :mrgreen:

Lynn

Re: How to tell if you have a "Slow" Lane

Posted: Thu Feb 15, 2007 9:36 am
by gpraceman
DerbyAddict wrote:I knew I shouldn't have gotten into this thread.... Now my head is starting to leak.

Must... find... more..... duct..... tape.........
That is why I like the spreadsheet. Just plug in your raw data and it tells you if there is truly a significant difference in the averages and standard deviations. It even has included instructions.

Though, if there is a significant difference, it is up to you to determine the "why" by closely examining the track and addressing any issues.

Re: How to tell if you have a "Slow" Lane

Posted: Thu Feb 15, 2007 9:50 am
by tmbnorm
I used the spreadsheet on our track times and did not find any significant difference in the standard deviations.

The spread sheet is helpful.

Thanks

Re: How to tell if you have a "Slow" Lane

Posted: Thu Feb 15, 2007 10:21 am
by gpraceman
tmbnorm wrote:I used the spreadsheet on our track times and did not find any significant difference in the standard deviations.
How were the averages? Standard deviations are only half of the picture.

Re: How to tell if you have a "Slow" Lane

Posted: Thu Feb 15, 2007 3:03 pm
by Stan Pope
Noted that at least the t-test required assumption that the populations had "normal distribution". Does limiting the sampling to "faster cars" provide such a distribution. My observations of car times is that they do not satisfy the definition.

Is there a test that can be applied to justify the assumption?

Re: How to tell if you have a "Slow" Lane

Posted: Thu Feb 15, 2007 3:15 pm
by gpraceman
Stan Pope wrote:Does limiting the sampling to "faster cars" provide such a distribution. My observations of car times is that they do not satisfy the definition.
I would still think it to be normally distributed. Even the faster cars will demonstrate variation in their run times, especially across different lanes. I would expect that faster cars would have a tighter standard deviation in general, but the distribution should be close enough to be considered normal.
Stan Pope wrote:Is there a test that can be applied to justify the assumption?
There likely is. Though, for me it has been quite awhile since my stats courses, so I don't recall exactly.

Re: How to tell if you have a "Slow" Lane

Posted: Thu Feb 15, 2007 10:56 pm
by NealOnWheels
When calculating lane statistics based on finish times using many cars the standard deviation is really meaningless as a measure of lane bias. The standard deviation will be dominated by the variance of car construction.

Reducing this to the fastest 10 cars may improve the calculation but you would still not know for sure what impact car construction has on the standard deviation.

Running a single car many times on each lane would yield very good results. The standard deviation would be a good measure of the degree of confidence of the lane bias determination. Unfortuneately this would only determine how one particular car construction performs on each lane.

I would propose the following method to determine lane bias:

1. Calculate the average finish time for each car.
2. For each finish time calculate the difference between the finish time and the car's average time. A positive value indicates a slow run and a negative value indicates a fast run.
3. Calculate the average and standard deviation of the differences for each lane.

The lanes average would indicate on average how much faster (negative value) or slower (positive value) cars ran on each lane. The standard deviation would be a good indicator of much of an issue lane bias is.

Anyone have any raw data to try this with?

Re: How to tell if you have a "Slow" Lane

Posted: Fri Feb 16, 2007 6:41 am
by mossk
The way to do this assuming normality is to put the data in a regression model along with a variable that would identify each car. Using all the data available should increase the accuracy (power) of your model. But you have to realize different cars my act different in each lane. For example, you may have one car that is a rail-rider with a left dominate wheel and another with a right dominate wheel. There are certainly a lot of other reasons why cars would act differently in each lane. But putting car number into the equation will sort this out. I think there was another tread about what to do if you identify a slow lane.

My guess is the data is normally distributed. But you could test for normality. If you find it is not normally distributed then a non-parametric test is appropriate.

Re: How to tell if you have a "Slow" Lane

Posted: Fri Feb 16, 2007 6:53 am
by mossk
Noted that at least the t-test required assumption that the populations had "normal distribution". Does limiting the sampling to "faster cars" provide such a distribution. My observations of car times is that they do not satisfy the definition.

Is there a test that can be applied to justify the assumption?
To answer Stan's question there are many tests to test for normality. One of them is the Shapiro-Wilk test.

Re: How to tell if you have a "Slow" Lane

Posted: Fri Feb 16, 2007 11:25 am
by SuperDave
Three quick comments:
1. Get a good track. (highly recommended)
2. Stop using scoring methods that depend on lane balance. (highly recommended)
3. Or, use the Soap Box Derby method. (not recommended)

For those not familiar with the Soap Box Derby, every heat consists of two runs on two lanes on a city street. I race you and the differential time is recorded. Nobody cares about actual run time, just the difference at the finish. We swap lanes and wheels (yea, wheels!) and do it again. Again, the differential is recorded. Biggest difference (if a split decision) wins. This, in theory, balances the lanes but works only if you have a fairly small number of racers and still takes all day

Re: How to tell if you have a "Slow" Lane

Posted: Fri Dec 05, 2008 5:25 pm
by ScouterDave
I have become the "Keeper of the Track". I have been storing it in my (dry) basement. Should I prepare the track for the season?

Re: How to tell if you have a "Slow" Lane

Posted: Fri Dec 05, 2008 6:16 pm
by ohiofitter
I don't know about all the data inputs...but our troop track is a plywood track with a lament on it.and up until last year we ran a double elimination race....But we had a lane that is truly slow so that's why they went to a average race.......IMHO the grain has probably lifted under the lament.....affecting the one lane... Cars can run a 2.4 to 2.5 in lane 2 and lane 3...but all of a sudden a 2.9 to 3.3 in lane one.........I know people can say that it might be poor axle/wheel prep..and if it wasn't the hole troop that this happens to I would agree...lets face it we always have about 15 cars in each den.And there nobody that can polish....I think not...I'd like to think I have a grasp on axle/wheel polishing so This happens to our cars as well...The only thing that the boys like see if they draw lane 1 first there times look more competitive after that one......And for the boys that draw that lane last you can tell there is some wonder as to why there car was fast and now it's not

Re: How to tell if you have a "Slow" Lane

Posted: Fri Dec 05, 2008 6:39 pm
by gpraceman
ScouterDave wrote:I have become the "Keeper of the Track". I have been storing it in my (dry) basement. Should I prepare the track for the season?
I think it is always good practice to take out the track well ahead of the race and make sure there are no problems with it. Then you have the time to address any issues that you find, instead of trying to do so the night before the race or on race day (when you are in a panic).

If you have some old cars, run them down each lane several times and see if there are any spots the cars jump around. You can also run you hands down the lane guides and along the joints to feel for burrs, bumps or divots.