Standard Deviation?

[bracketracer]

Do any of you guys log the standard deviation for runs on your track? I'm nearly finished building a four lane 32 foot wooden track with an Arduino-based timer. I've got it done far enough to test it out (still need to apply a finish to the track surface), but I had to see it work so I slapped four wheels on an uncut block and made about seven groups of ten runs. The "car" has four straight axles so it could be just poor alignment causing this, but I'm seeing a standard deviation of around .026 sec. I thought that seems high, but I don't have anything to compare it to. I did work on the cars alignment after the kids went to bed tonight, so I'll make another group of runs in the morning to see if it helps. I do have an electric gate release, the .026 was using it compared to .032 with the manual lever. The run times appeared to be more consistent with the electric gate, but the car would occasionally throw a couple fast or slow runs in the mix and blow the average. I made twelve runs and discarded the highest and lowest. Any idea what a reasonable number would be? Stephen

 

[txchemist]

Run 4 cars down the track and record each run as you switch lanes. Then look at average time of all four cars run to run variation to see the timer variation. ( variation lane to lane is usually very small) Then you can decide if your car or your timer is causing the variance.

 

[bracketracer]

I'll need to build a few cars before I can try that, since I only have my son's car from this year and this block with wheels. I did a quick alignment last night (2.5 bend on the rears and the DFW and adjusted it on a board for 4" over 5 feet. I had the youngest help run two sets of twelve runs this morning, after discarding the high and low runs I'm seeing a SD of only .007 (all test runs were in the same lane). The car will still throw an odd run now and then, I had three slow runs out of these two sessions. I should probably run another group and just delete those when they happen. Each time it got the wobbles it was a trash run. I'm thinking it's because I didn't tune the COM at all, it's heavy on the front and the DFW is making it wobble between the diagonal opposite wheels. You can hear and see it when it's on a bad run. I think the timer is fine, we just need to build an actual car now.

 

[topspin.d]

One factor that significantly impacts times is staging. If you just plunk them down for each run, you'll have good times on some and bad on others... the standard deviation will increase with inconsistency in staging.

 

[bracketracer]

I set the car body parallel to the guide, DFW just touching the guide, both rear wheels and the NDFW pulled to the nail head each time. I don't know if that was the way to get the fastest time, but it was a constant I could hold. I just let the little guy count down and hit the GO! button on the remote.

We made some more runs with my oldest's car from January. It was showing a SD of only .005 with a dozen runs in the same lane, so we made a dozen more but split it up to 3 runs in each lane. Same result, .005. His car had the COM farther back, at 1", but it has straight axles and it also suffered from the occasional wobble just not as often as the plain block. I need to build a car now with what I've learned since I've joined this site and see how repeatable it is.

 

[pony express]

I've always been surprised by the deviation in the times on my track. I don't know if it's the timer or the hinged gate. I've tried lubricating the hinge and it seemed to help for a while. Even with the time deviation though, if I race one car agaist another, the fastest car will always win despite their times.

 

[txchemist]

I see the same thing Pony sees. A quick check with only two cars can be to run a few races, enough to get the typical car variation you think you have, and compare the results average and Standard Deviation, also look at the deltas between the cars for each race and get average and deviation for them. In my case, the delta times have a std. Dev. of only 0.003 which is a third of the race to race time variation. If the cars were really changing speeds due to staging or car variation so much, car B could have beaten car a twice out of 5 races.If every car in a race gets on the track for enough runs, the average, and especially the average with the slowest reading tossed out will be accurate, but if you are running a race with many cars, and your track is like mine, use the point system and dynamic scheduling. When you look at the winner deltas, it looks almost impossible to get a run where car B beats car A. This of course is with a good track where there is no "slow" lane.

 

[bracketracer]

Well, if those are actual numbers, maybe I don't have a problem.
I guess I was expecting to see .000! lol!

Here's what my son's times look like with three runs in each lane:
(These are as observed, I did not discard the high and low time like I normally do)

Lane 1 1.716 +0.002
Lane 1 1.717 +0.003
Lane 1 1.711 -0.003
Lane 2 1.726 +0.012
Lane 2 1.709 -0.005
Lane 2 1.710 -0.004
Lane 3 1.724 +0.010
Lane 3 1.716 +0.002
Lane 3 1.713 -0.001
Lane 4 1.706 -0.008
Lane 4 1.707 -0.007
Lane 4 1.712 -0.002
Average 1.714
Std Dev 0.006

Those +0.010 and +0.012 runs had noticeably more wobble in the car, so I'm sure that's what caused it. I need to work on my alignment skills to see if I can minimise that. The way my track is set up, I was planning on using lane 4 for all of my testing. I want to add incremental sensors in that lane. I see it's a little quicker than the others, probably because it has about a foot less wire to the sensors.

Thanks for the advice guys!

 

[OPARENNEN]

Maybe an obvious point, but being careful to stage the car exactly the same each time is point 1.

2nd point: Some starting gates come down fast, and make a clunk (like mine), even with a solid base. I tried putting a soft cloth in my hand, and catching the starting gate softly, totally eliminating the "clunk". The times decreased dramatically.

When I send my cars in to NPWDRL races, I determine the ratio between the NPWDRL times and my cars immediately after they are returned. I use this Ratio to determine how fast my cars will be in the next race (assuming John doesn't change something on this track). In the past, I found that my predicted times for the next race are quite accurate.

By looking at the previous months NPWDRL times, I pretty much know in advance where I will place, which in the past couple of months has been 10th place or worse.

 

[txchemist]

OK, taking my own advice, and looking at the good spread that bracketracer got, I could tell my cars were OK but my track was not up to snuff. I had to put my track away so my wife could have an unreasonable crochet club meeting, but while it was down I worked on the gate some. Earlier work on the gate added better foam to stop gate smooth with no bounce. I smoothed the area where the notch rubs the gate, and then covered both surfaces with scotch tape ( I still use a hand release, not a solenoid yet) , and the arm connected to the gate that trips the laser start was loose and that was tightened. Now a lot of interesting things pop up.
1. identical variation car to car by race.
2. Std. Dev. on both cars tightened up to now be very low.
3. Car b was not re-prepped after one year, just new oil added and now looks to be slowing down so will get a re-prep and Car A gets a second coat re-prep

 

[bracketracer]

Those numbers look much better TX!

Over the weekend I had time to build a new car. I wanted to try a plank style rail runner, something simple. I made it 142g with an 11/16 COM. I made 5 runs all in lane 4. The observed times varied by .007, with a SD of .002. I added a touch more steer to it and then it ran 1.700, 1.700, 1.699, 1.698, 1.699, 1.700! Made for an average of 1.699 and a SD of .001! You guys are good teachers! Thank you!

 

[Guest]

MERKIN THE WORLD!!!!