The "Thrill" of the Race

[TurtlePowered]

Sorry Stan, the lights on out finish line tells the first, second, third and fourth place finish. I guess i'm a heads up kinda guy a timeing deal may be alright for us but we are not 8 years old.

Oh! Okay! We are both arguing for head-to-head racing.

Are you advocating head to head racing but using a timer so you don't have to eye cars coming within 1/1000 of a second?

[Stan Pope]

Are you advocating head to head racing but using a timer so you don't have to eye cars coming within 1/1000 of a second?

No, I am advocating head-to-head racing using finish line electronics which show finish order as one of the judges. 1/1000 seconds is not sufficient resolution.

[TurtlePowered]

Are you advocating head to head racing but using a timer so you don't have to eye cars coming within 1/1000 of a second?

No, I am advocating head-to-head racing using finish line electronics which show finish order as one of the judges. 1/1000 seconds is not sufficient resolution.

I haven't particularly studied the other elimination methods so forgive my question that seems fairly obvious. Sooner or later I would think, using a elimination type method, there will be a final race. Somebody will get a better lane than the other, correct?

[Stan Pope]

I haven't particularly studied the other elimination methods so forgive my question that seems fairly obvious. Sooner or later I would think, using a elimination type method, there will be a final race. Somebody will get a better lane than the other, correct?

In the case of "elimination" racing, this is true. It is "inaccurate" to the extent that lane differences exceed car differences. It is "unfair" to the extent that lane assignment is not random.

There is a middle ground between "elimination" racing and "timed" racing which preserves the excitement of head-to-head racing. This middle ground is called "final standings" which is based on points earned by individual heat results. "Final standings" is "inaccurate" to the extent that heat and lane assignments are not balanced. If one racer is assigned extra heats (don't laugh ... it has happened) then he has an advantage or disadvantage, depending on how points are assigned and compared. If one racer is assigned more heats in the faster lanes than in the slower lanes, then he has an advantage. If one racer is assigned more heats against slower opponents and fewer against faster opponents, then he has an advantage. So, balance in the chart is an essential component of accuracy. This is why Stearns sometimes has a problem and why PPN and Stearns are best used as a screening race for a PN or CPN finals.

[TurtlePowered]

I. So, balance in the chart is an essential component of accuracy. This is why Stearns sometimes has a problem and why PPN and Stearns are best used as a screening race for a PN or CPN finals.

What is PN and CPN?

[Stan Pope]

What is PN and CPN?

I'm glad you asked!

PPN:
Each car races the same number of times in each lane (which implies that the number of races is a multiple of the number of cars)
Equality of opposition is optimized, i.e., no head-to-head matchup count exceeds another by more than 1.

PN:
Each car races in each lane the same number of times.
Each car races each opponent the same number of times.

CPN:
Each car races in each lane the same number of times.
Each car races each opponent the same number of times.
Each head-to-head matchup between cars appears elsewhere in the chart on opposite lanes.

These are detailed in notes on page
http://members.aol.com/standcmr/ppngen.html

[TurtlePowered]

What is PN and CPN?

I'm glad you asked!

PPN:
Each car races the same number of times in each lane (which implies that the number of races is a multiple of the number of cars)
Equality of opposition is optimized, i.e., no head-to-head matchup count exceeds another by more than 1.

PN:
Each car races in each lane the same number of times.
Each car races each opponent the same number of times.

CPN:
Each car races in each lane the same number of times.
Each car races each opponent the same number of times.
Each head-to-head matchup between cars appears elsewhere in the chart on opposite lanes.

These are detailed in notes on page
http://members.aol.com/standcmr/ppngen.html

Just for the fun of it I generated a chart for 50 racers for 6 lanes. Looked like each racer races 6 times which is the same as our average time method does. Your earlier notes specified a final grouping to determine a championship, how many cars do you normally take to a finals?
It's interesting when differing methods come to different conclusions when racing.

[Stan Pope]

Just for the fun of it I generated a chart for 50 racers for 6 lanes. Looked like each racer races 6 times which is the same as our average time method does. Your earlier notes specified a final grouping to determine a championship, how many cars do you normally take to a finals?

Number of finalists is set so that there is a high probability that those deserving trophies are among the finalists. For most PPN charts and tracks, taking number of finalists at about twice the number of trophies gives a probability of about 100% that the deserving are among the finalists. Cory's chart simulation program includes this measure.

It's interesting when differing methods come to different conclusions when racing.

Depends on what is important. What does timed racing do when a car is pushes the envelope to the point that it "blows its engine" (i.e., flies off the track) in about 10% of its runs? A rerun may not be fair, because it is the car's fault. Does he get an infinite time for the "did not finish?" Does one DNF move a car from first to last place overall? Or, does everyone get to discard its worst N times?

Final standings penalizes a poor heat in a limited fashion. To me, that is about right.

What aspect of the car's performance is important to you?

[Cory]

For most PPN charts and tracks, taking number of finalists at about twice the number of trophies gives a probability of about 100% that the deserving are among the finalists.

TurtlePowered: Note that a chart of a particular type (PPN, CPN, PN) is not guaranteed to exist or to have been "discovered" for a particular combination of cars, lanes, and number of races.

For example, there is no known 2-round PPN chart for 12 cars and 3 lanes. Same goes for 14 cars. If you enter these parameters into ppngen.html, the Javascript will produce a 1-round chart. Any number besides 12 and 14? No problem.

The more accurate PN and CPN charts are much less likely to be available than are the PPN charts.

So in addition to Stan's guideline, I always pick a number for which I know there exists a right-sized CPN chart. For our 4-lane track (or our old 3-lane track), this number is 7.

If there's a tie for 7th, then you have a choice to make. Let 8 cars into the Finals and use a less accurate 8-car PPN chart, or break the tie with a race-off and use the more accurate 7-car CPN chart.

[TurtlePowered]

Just for the fun of it I generated a chart for 50 racers for 6 lanes. Looked like each racer races 6 times which is the same as our average time method does. Your earlier notes specified a final grouping to determine a championship, how many cars do you normally take to a finals?

Number of finalists is set so that there is a high probability that those deserving trophies are among the finalists. For most PPN charts and tracks, taking number of finalists at about twice the number of trophies gives a probability of about 100% that the deserving are among the finalists. Cory's chart simulation program includes this measure.

It's interesting when differing methods come to different conclusions when racing.

Depends on what is important. What does timed racing do when a car is pushes the envelope to the point that it "blows its engine" (i.e., flies off the track) in about 10% of its runs? A rerun may not be fair, because it is the car's fault. Does he get an infinite time for the "did not finish?" Does one DNF move a car from first to last place overall? Or, does everyone get to discard its worst N times?

Final standings penalizes a poor heat in a limited fashion. To me, that is about right.

What aspect of the car's performance is important to you?

What our pack does if a car flies off the track is to re-run the same lane to see if bad staging caused the problem or technical problem with the car. Cars that fly off twice are not allowed to continue. I have seen a problem with one car starting up, can't recall if car flew off track or just went in wrong lane at beginning of race but have not seen a double disqualification.
We have 1st 2nd and 3rd place for each rank with about 12 scouts per rank and top 6 or 8 overall. According to that 1/2 of the scouts make the finals, not a bad thing I suppose. Do the scouts that don't make the finals lose their interest?
Running races this way, do cars that lose to other cars by the smallest of margins tend to even out over the course of the contest?
I think in most cases, the fastest car with come out ahead with either method being used. The chart method does make it easier to compensate for a bad heat but I don't really know how often that comes up. I think the average time method has one advantantage of the computer doing all the math.
It would be interesting for some packs using chart methods to keep the times and do the time averaging vs chart method as far as car placement is concerned.

[Stan Pope]

So in addition to Stan's guideline, I always pick a number for which I know there exists a right-sized CPN chart. For our 4-lane track (or our old 3-lane track), this number is 7.

If there's a tie for 7th, then you have a choice to make. Let 8 cars into the Finals and use a less accurate 8-car PPN chart, or break the tie with a race-off and use the more accurate 7-car CPN chart.

Having gone to the trouble to estimate the accuracy of the preliminary race method and run it to pare the field, I would always aim for either a PN or CPN chart, breaking ties as necessary to get there.

Of course, if I were aiming for a 7-car CPN chart, and there were an 8-way tie for 1st thru 8th, I've got a problem! ... Actually, I've got two problems. Breaking the tie and using a track that desparately needs to be upgraded.

An subtle advantage to this process is that no one car is singled out as the slowest of the bunch. The precision of Stearns or PPN charts just doesn't justify that conclusion. The CPN chart does have that precision, but now, it is telling the slowest of the fastest, which is not so hard to take.

BTW, Final Standings can be run on paper or on computer, just as timed races can. If the computer goes down, the Final Standings arithmetic is a lot easier to do on paper.

Another thread showed the full chart for another race ... times and heat placement, but did not show the results summary. The results from the two scoring methods, timed vs. points, correlated well, but not exactly. I haven't ascertained the causes of disagreement, yet. Cory did an interesting analysis, tho.

[glaforge]

Cory,

Nice start on the analysis. I went back on one point you made about the group 1 races (not the others). You spoke of "timing concerns" with respect to heats 11 thru 15 being a car length slower than heats 36-40. I think we can attribute that to the mix of cars racing (as you said). I charted the cars that raced in both groups of heats and found that cars 145 (17 pts), 149 (25 pts), 152 (22 pts), 154 (twice and 20 pts), and 157 (32 pts) raced in heats 11-15 that did not race in heats 36-40 and car 151 (9 pts), 153 (twice 21 pts), 156 (9 pts), 161 (10 pts) and 65 (15 pts) raced in 36-40 and didn't race in 11-15. In other words a slower mix of cars raced in heats 11-15 than did in heats 36-40.

I might also disagree with your order of finish switching 2nd and 3rd place around between the total time and place finish. I would have to review the mix of cars each raced before I could buy into that one (guessing they each raced a different mix of cars a different number of times).

Now with respect to group 2 races, the "alarming" difference between the times of each racers 7th race and 8th might be attributed to solely car 125 which ran 1.2 secons slower in its 8th heat than it's 7th. Nearly all the 7/100ths difference in total average can be attributed to this car alone. Once we factor that in, now all cars are within aproximately 3/100ths across all heats...what would an acceptable variance be? Interestingly enough about the same (approx 3/100ths of a second) variance as in group 1.

Since we are seeing a 3/100ths across two groups, I'm beginning to think that there is some consistancy to the data and the 3/100ths might be normal and acceptable. Need to check the other groups and also verify against this year's race (when we have it).

One might also be concerend with the time differences between group 1 and group 2 as you stated, but since I know (and you could not have possibly known) that the races between all the groups we comingled. In other words, there were some group 1 races, then group 2 then group 3 then group 4 and then back to group 1, group 2 and so on. Basically 8 rounds like that. The difference in times may be attributed to experience in building the cars themselves because group 1 were my oldest scouts (again you could not have possibly known this).

The 1.2 second difference in car 125 7th and 8th race, as Stan should point out, is a heavy price to pay for him, especially when we are seeing a 3/100th of a second variance between a cars heats. One might use this example alone to argue for place scoring methods.

However, Stan also wrote of 3 points in using timing finishes, the last being the benefits outweighing the problems. Let's call this the intangibles that measuring time/placements cannot possibly consider, For example, 1) All scouts get the same number of races, hey, some slower cars may have spent more time in making the car and deserve's just as many races. And believe it or not, this is important to more scounts than you think 2) The entire audiance is there to cheer, clap, etc the winners because they didn't empty out once the "winner's bracket" was announced. 3) With large groups such as ours, it would be impossible to keep the number of races consistant (See point 1...it is big to the kids) across all groups and still get out of the derby in a reasonable amount of time. 4) It gives the kids in every group a way of seeing how they did in the overall race without actually having to race the kids in the other groups (yes, we do print reports with times only, no names).

After having done both methods over the past few years, I had many fewer problems with the total time method...especially with the intangables which seem to make or break how much fun the entire group as a whole has.

[Cory]

Your points are all well taken. I really appreciate that you posted your data so there could be some substance behind this analysis for a change.

Some comments (or maybe attempts to clarify my position):

I might also disagree with your order of finish switching 2nd and 3rd place around between the total time and place finish. I would have to review the mix of cars each raced before I could buy into that one (guessing they each raced a different mix of cars a different number of times).

In practice, we do not disagree. I don't advocate awarding trophies based solely on the results of large, point-scored PPN charts. Rather they should only be used as "trimming tools" for selecting finalists.

Since we are seeing a 3/100ths across two groups, I'm beginning to think that there is some consistancy to the data and the 3/100ths might be normal and acceptable. Need to check the other groups and also verify against this year's race (when we have it).

Perhaps. As you may know, I've done a bit of car testing, on what is probably an average track. If you look at the Control Set averages on my first day of testing ( http://pack146.nova.org/pinewood/expres1.html ) you'll see they are 2.5942, 2.5597, 2.6005, 2.5944, 2.5942, and 2.5954. This is a 6 millisecond spread, approximately, versus your 30 millisecond spreads. Thus, 30 milliseconds seemed a little high to me. IMO, jury is still out on this one, you could be right.

Also, keep in mind that, to some extent, my perceptions are influenced by the complaints of parents who have attended our District derby. Maybe if you had been running our District derby they wouldn't have complained.

After having done both methods over the past few years...

I'm assuming you mean time-scored PPN as one method, point-scored PPN followed by point-scored CPN as the other method?

I'll work through the other groups as time permits. I think the aggregated heat numbers is the best statistic, I'll probably compute only this one.

Thanks again for all this data.

[Cory]

As promised, more "analysis" for those interested:

Group 3 consisted of 16 cars.

Car# Total Time Sum of Places
90 19.639 28
91 19.232 23
93 21.583 31
95 18.766 11
96 19.189 20
98 19.387 21
99 22.694 31
100 19.106 19
101 18.374 8
102 18.635 12
103 19.183 21
105 18.886 15
106 20.381 29
107 18.533 9
108 19.392 23
109 18.964 16

Average times aggregating by heat number:

Average of 1st heat times for all 16 cars: 2.444625
Average of 2nd heat times for all 16 cars: 2.432063
Average of 3rd heat times for all 16 cars: 2.430813
Average of 4th heat times for all 16 cars: 2.422313
Average of 5th heat times for all 16 cars: 2.452750
Average of 6th heat times for all 16 cars: 2.443563
Average of 7th heat times for all 16 cars: 2.443500
Average of 8th heat times for all 16 cars: 2.5628

Spread is about 30 milliseconds, which is consistent with Groups 1 and 2, provided you omit one significantly "oddball" time.

Group 4 consisted of 13 cars:

Car# Total Time Sum of Places
128 18.583 14
129 18.205 10
130 18.915 23
131 18.485 16
132 18.414 13
133 19.23 24
134 17.955 8
135 19.237 27
136 19.16 27
137 19.819 28
138 18.715 20
139 24.896 32
140 18.586 16

Average of 1st heat times for all 16 cars: 2.400692
Average of 2nd heat times for all 16 cars: 2.405615
Average of 3rd heat times for all 16 cars: 2.415923
Average of 4th heat times for all 16 cars: 2.419923
Average of 5th heat times for all 16 cars: 2.405154
Average of 6th heat times for all 16 cars: 2.400462
Average of 7th heat times for all 16 cars: 2.391231
Average of 8th heat times for all 16 cars: 2.414846

Spread is 28 milliseconds, which is consistent with the other groups.

Trophy summary (car #, total time) (4th place included for comparison)

Group 1
151 18.022 **
161 18.023 **
156 18.039 **
155 18.376

Group 2
110 18.031
114 18.090
118 18.304
119 18.484

Group 3
101 18.374
107 18.533
102 18.635
95 18.766

Group 4
134 17.955
129 18.2050
132 18.4140
131 18.4850

Take a look at the top three spots of Group 1.

This derby consisted of 132 group-intermingled races, according to Greg. Both ends of the track got handled over and over again. What are the chances of maintaining timing consistency across that many heats within 16 milliseconds (1st thru 3rd places) or within 1 millisecond (1st and 2nd place)?

I would have really liked to see cars 151, 156, and 161 go at it head-to-head for 5 or 10 minutes.

[glaforge]

Cory,

Again great analysis.

I read you tests pages with great interest. Perhaps the close tolerances you had are attributed to a couple of factors like 1) You track being shorter (sharter track=less time=closer tolerances). 2) warming up your cars before you did your time trials (a luxery we don't have in running a PWD). With your analysis, I'm now comfortable with 30 ms...especially when you factor in 4 different independant groups.

Certainly group 1 should have all our attention as there is really no dispute in the results of groups 2-4. Both the top two cars in group 1 raced in the district derby, with 161 beating 151 that day. But that is after seveal days and on a different track and so on...to many variables, some of which we may not know.

How many more times do you want them to go at each other? In essance they went at each other 8 times already, Another 5-10 minutes may show a different result and it may show the same. At some point the buzzer must sound and the race must end. Yes, the order may have been a result of handling and it just as easily might not have been. This debate could go on forever with no agreement.

I know the kid that came in second place was bummed because it was only 1 ms...who wouldn't be, but he handled it very well.

Question: Would a perfect N chart fail to pick winner in this instance? Would the cars end up tying after n races..then what, run another n races.

Again, thanks for the analysis