Why a Straight Indexing System Won't Work!

[Eric]

In recent weeks there has been a fair amount of discussion regarding going to a straight indexing classification system, possibly as early as 2003. I have been privy to emails amongst the Solo 1 committee members, one of which suggests that 3 of the 4 Solo 1 organizing clubs, namely HADA, OMSC and TAC have committee representatives that support the notion of a straight indexing system. In the case of at least HADA and OMSC some sort of poll within their respective clubs has been done to show there is club support for this idea. I find it somewhat frightening that there seems to be so much support for an idea that was proven to be unworkable a little over two years ago when the CCC considered it. The purpose of this post is to give the Solo 1 community some insight as to why a straight indexing classification system won’t work to allow for an informed opinion in light Dave Pratte’s Solo 1 Car Classification Poll which hopefully you will respond to.

The primary reason a straight classification system will not work is that a straight indexing system requires the entire population of vehicles to be ranked using one indexing system and one weighting of those indices. This has been proven not to work.

A little background information is required to explain this. Under the current classification system a performance index is established to rank the cars within each family into a somewhat logical order for the purposes of establishing 3 “stock” classes within each family. There are 4 “objective, measurable” attributes that are considered in coming up with a performance index, horsepower, torque, weight and handling (granted the handling index is not that objective or measurable, however the values assigned to specific vehicles were done in as objective manner as possible). Those attributes are converted into these sub-indices (the scale used is beside the index), horsepower (50hp to 550hp), weight-to-hp (6.0 lb/hp to 35.0 lb/hp), weight-to-torque (6.0 lb/tq to 35.0 lb/tq) and handling (10% to 85%). Under the straight indexing system proposed the actual sub-indices may change, but the 4 attributes they are measuring won’t.

The scale for each of the sub-indices is converted into a percentage and then each sub-index is weighted to come up with a final performance index (PI) number. So, for example on the horsepower scale 50hp = 0% and 550hp = 100% and each interval of 50hp = an interval of 10%. In A family horsepower is weighted 20%. So a horsepower value of 200hp has a horsepower index of 200 / (550-50) x 100 = 40% and this makes up 40% x 20% = 8 PI points of the final performance index number for the given vehicle.

Initially, all the families were weighted using the A family weightings, HP: 20%, W/HP: 30%, W/TQ: 15% and Handling: 35%. However, what the CCC found was that the C family cars (lighter, small displacement cars) were not ranked correctly using these weights. There were a few cars that were out of place. If you have a copy of the classification spreadsheet (perhaps Christian can provide a link to the current one) you can see this for yourself. Use the A family weights for B and C families and re-sort. Take a look at how the ranking changes. If memory serves me correctly the Austin Mini Cooper S is significantly affected by this. It moves up substantially.

Why is this? It basically comes down to the scale that is used for the indices. Going back to the horsepower example. The scale goes from 50hp to 550hp. For C family cars the relevant range is probably more like 50hp to 220hp. So if you use the 50-550 scale, you have to weight the horsepower attribute for C family cars more heavily relative to the other families to ensure this attribute correctly influences the ranking. In C family a 50hp difference between two cars is significant, much more significant than a 50hp difference between two A family cars, yet in a straight indexing system with the same weighting of the indices across the entire population those 50hp work out to 10 index points (50/(550-50) x 100) [times the weight assigned to the HP index] regardless if it is a Honda Civic or a Camaro. I think you can appreciate that a 150hp Honda should be indexed much more than 10 index points above a 100hp Honda whereas 10 index points for a 350hp Camaro over a 300hp Camaro is more realistic.

This issue with the relevant range of the scales applies to not just the horsepower scale, but to a scale for any attribute. If the relevant range for a group of vehicles you want to compare happens to be towards the lower end of the scale then that particular index will not be weighted as much as you may intend it to. Lets assume you pick these four attributes to index, horsepower, torque, curb weight and handling and you decide to weight them 25% each to come up with your PI. I am trying to make this as simple as possible. I’ll pick two C family cars for this example, a 1991 Honda Civic Si and a 2002 Acura RSX Type S.

First lets use a relevant range for C family for each of these attributes. HP: 50hp to 220hp, Torque: 50ft-lbs to 180ft-lbs, Weight: 1800lbs to 3000lbs and Handling: 0 to 100. Using these ranges the PIs for the two vehicles work out as follows:

1991 Honda Civic Si
HP: (108-50) / (220-50) = 34.1 ==> x 25% = 8.5
TQ: (100-50) / (180-50) = 38.5 ==> x 25% = 9.6
WT: (3000-2260) / (3000-1800) = 61.7 ==> x 25% = 15.4
Handling: 50 ==> x 25% = 12.5

46.0

2002 Acura RSX Type S
HP: (200-50) / (220-50) = 88.2 ==> x 25% = 22.1
TQ: (142-50) / (180-50) = 70.7 ==> x 25% = 17.7
WT: (3000-2767) / (3000-1800) = 19.4 ==> x 25% = 4.9
Handling: 60 ==> x 25% = 15.0

59.7

Now lets compare the same two vehicles using scales for the attributes for the entire spectrum of vehicles. HP: 50hp to 550hp, Torque: 50ft-lbs to 550ft-lbs, Weight: 1800lbs to 4200lbs and Handling: 0 to 100. Using these ranges the PIs for the two vehicles work out as follows:

1991 Honda Civic Si
HP: (108-50) / (550-50) = 11.6 ==> x 25% = 2.9
TQ: (100-50) / (550-50) = 10.0 ==> x 25% = 2.5
WT: (4200-2260) / (4200-1800) = 80.8 ==> x 25% = 20.2
Handling: 50 ==> x 25% = 12.5

38.1

2002 Acura RSX Type S
HP: (200-50) / (550-50) = 30.0 ==> x 25% = 7.5
TQ: (142-50) / (550-50) = 18.4 ==> x 25% = 4.6
WT: (4200-2767) / (4200-1800) = 59.7 ==> x 25% = 14.9
Handling: 60 ==>x 25% = 15.0

42.0


Using a relevant range for C family cars the PI spread between these two vehicles is 13.7 PI points. When you use the range for the entire population of cars that spread gets reduced down to 3.9 PI points. What effectively has happened is that for these two vehicles (and all other similar vehicles) there is a shift in emphasis from HP and Torque to Handling (the Handling scale still being in a relevant range for these two vehicles). Since the handling spread between these two vehicles was not significantly different for these two vehicles their final PI numbers have come closer together.

(This example also shows the shortcomings of using curb weight alone as an attribute which is why the CCC elected to use Weight to HP and Weight to TQ, buts that’s another discussion.)

Using a relevant scale for each of the sub-indices for similar cars (i.e. a family) is one way of overcoming this problem. The CCC chose to keep the scales the same across all the families, but change the weightings given to the sub-indices to overcome this problem. So, if you were ever wondering about those weighting charts by family in the rule book, now you know why they exist.

Before I leave this discussion, lets look at that PI spread again. Currently the RSX Type S is classed a C1 car and the Civic Si is classed a C3 car. The PI spread in the example above is 13.7 PI points, so effectively this spread covers 3 classes, C1, C2 and C3. Presumably one of the reasons for going to a straight indexing classification system is to have more classes available at the bottom of the PI scale (the so called “bottomless pit” problem). So if you wanted to have two classes between these two cars rather than one, you would have to fit 4 classes within a PI spread of 3.9 PI points. To get the ranking correct within that small a PI spread in order to cover 3 or 4 classes is virtually impossible.

To summarize, the problem with a straight indexing system is you don’t have the luxury of shifting the weightings around for vehicles with different characteristics, nor can you change the scale used for the index for different vehicles as you do in a family based system. You pick your sub-indices assign a weight to them and let the ranking fall out as it may. The final PI number will possibly be more heavily weighted by the attribute that has a scale that is most relevant to the vehicle in question rather than in the ratio you intended to weight the attributes by. At the lower end of the PI ranking the PI index numbers will tighten up for vehicles that would normally be considered to have substantially different performance levels. These two factors will make it virtually impossible to get a decent ranking for the “bottomless pit” vehicles (basically the current A3, B3 and C3) to establish classes at the lower end of the PI scale.

I hope you found this insightful :)

Eric Juraschka
Chairman – Car Classification Committee

[ctenche]

Good post Eric. I does shed some light on how the indexing system works and what the tradeoffs are between the various weighting factors.

However, you are making one potentially significant assumption. You are assuming that the straight performance indexing (PI) system will be calculated using the weightings and method you specified.

If that is not the case then perhaps the straight PI system might work. I guess what it comes down to is that we don't have enough information at this point to say definitively that the system will or won't work.

I'd like to hear the counter point to your argument from those who are working on this proposed system and maybe have a look at a draft spreadsheet that shows the straight PI classification system.

[turboawd]

Eric, thanks for the info. I think the answers lie in between the lines of your post. I would drop HP and curb weight as a PI weighting alone and use only HP/weight, T/weight and handling. I would think Braking would also be important but realize that race brake pads and mods throw that out the window. This should improve the current problem you hi-lighted. There is no doubt that a linear system would require a lot of massaging but you need to start somewhere as a baseline then develop it. Lets say one recalculates the PIs as I propose. Set up 10-12 classes and see where they fall. Then review it and if a car looks grossly mis-classed, then look at its PI and see whats wrong and adjust the PI if neccessary. There is no doubt that you will need to run through several iteratrions. Other series, i.e. Pro-Solo, re-class cars all the time.
You could also throw in some small x factors for things such as RWD, displacement, LSD, brakes, stability control...etc if it produces a better/fairer spread. Its an option.

The problem with family class system is that all root classes funnel down to one SP class. Then you end up with grossly mis-matched cars in the same class. For example, a moded Suzuki Swift GT vs. a moded Type R, all in CSP. With a linear system, its highly unlikely that they would have to compete against each other. If the Swift guys throws on a turbo then it most likely would compete against the Type R and maybe they should. I think it can work, it just needs development.

[Dave]

FYI, the Solo 1 Competitor Survey is now available online at:

http://www.soloontario.com/solo1survey.htm

Due to the results of the survey the Solo 1 Committee found itself in the position of needing to review the classification system further if we were going to take the results seriously. This lead in turn to a healthy discussion at the Solo Workshop, where several proposed changes were presented (including the straight indexins system Eric objects to in this thread), which I later outlined in the listserve poll I sent out a few days ago as I was instructed to do at the end of the Workshop.

Eric wrote Under the straight indexing system proposed the actual sub-indices may change, but the 4 attributes they are measuring won’t.

This isn't quite true. Under Christian's straight indexing proposal he uses weight instead of horsepower, along with the handling index, weight-to-horsepower, and weight-to-torque. This is in recognition of the fact that when indexing all cars together the curb weight of each vehicle becomes considerably more important. The specific weightings he used for these 4 indices are: curb weight (10%), handling (40%), weight-to-horsepower (40%), and weight-to-torque (10%). By placing considerable emphasis on curb weight and weight-to-horsepower the spread of cars in Christian's initial straight indexing spreadsheet look pretty logical to me, though of course there is always room for fine-tuning. Have a look for yourself:

http://www.hadamotorsport.com/new/flat.htm

The columns that are important to look at in this spreadsheet are obviously the first one where all the cars are ordered, the 'Class' column that shows the 12 root classes labelled A1 thru A12, and the final column where the total PI is shown.

In terms of Eric's example of a 91 Civic Si and a 02 RSX Type S, we see the Civic Si indexes out as an A10 car and the RSX Type S as a A5 car, with total PI's of 50.7% and 64.9% respectively. This seems like a pretty significant PI and root class spread, unlike the very minimal spread we see in Eric's example. Point being, there are ways to weight the indices, in my opinion, such that we do get a logical ordering of cars from top to bottom of all the cars in the spreadsheet.

As far as I'm concerned (and this is just one man's opinion), we can make a number of different classification systems work as well as what we've got now, including a straight PI system. Every system is flawed, has its own set of pros and cons, and these strengths and weaknesses need to be considered when choosing a system that's best for the series. So for me the choice comes down to what system will allow us to grow the series the most. In other words, which system will help us retain current competitors, bring back competitors who have left the series because of the current system, and attract new competitors and new sponsors. I don't know the answer to this, thus all the consultation with the Solo 1 community via the survey and now the listserve poll. At the end of the day the Director and Solo 1 Committee will have to decide what direction to take the series based on the feedback we've received and based on a reasonable application of logic. Obviously there are some differences of opinion here, and I for one value this as it's part of the process, as it will hopefully allow us to identify the best choice for the series. I don't know what that choice will be, but all I can do is represent HADA as best I can, and in my opinion this means supporting a classification system the majority of HADA's members prefer as long as I don't feel (as the club President) that we'd be doing harm to the club, the OSO, or the Solo 1 series by doing so.

Cheers,
Dave

[Eric]

ctenche wrote:

"Good post Eric. I does shed some light on how the indexing system works and what the tradeoffs are between the various weighting factors.

However, you are making one potentially significant assumption. You are assuming that the straight performance indexing (PI) system will be calculated using the weightings and method you specified."

---------------------------
Caius, thanks for taking the time to read my post. Regarding any assumptions in my example, my post was plenty long without considering different weightings or looking at presenting the sub-indices in a different manner. I tried to present my example in as simple a manner as possible to explain a very complex issue and in a way most people could relate to. If I were putting the system together I would use different sub-indices and different weightings then in the example. However, having said that, the example is relevant to illustrate the point I was making. Also, I tried to use ranges for my scales that were realistic to ensure the my findings were credible.

As a side note, if you were to change the weighting to improve the situation at the lower end of the PI scale you will create a problem somewhere else in the scale, likely near the top.

I hope I was able to illustrate that while on the surface a strickly PI based system, without families, seems very scientific and very objective, if you look a little deeper at how those PI numbers are generated there are some very serious problems. Those problems are are not easily overcome and will make implementing such a system an impossible task.

Eric J.

[turboawd]

Dave, why are there 5 indicies HP, W, Susp, W/HP, W/T?
Why index HP, W and then W/HP and W/T ? If the rational is that W plays a large part in cornering speed/ability then try it with 4; W, W/HP, W/T and Susp. I don't understand the relevance of have HP indexed seperately from HP/W. Can you exlpain please.

[Christian Sorensen]

turboawd:

Check out the Tables tab in the spreadsheet. That shows the weightings applied to each index. The ASOLO1 column is the relevant one, reading like this:

IDX_W: 10%
IDX_HP: 0%
IDX_SUSP: 40%
IDX_W-HP: 40%
IDX_W-TQ: 10%

As you can see, the IDX_HP index is not being used at all in the current calculation of PI.

The IDX_W and IDX_HP indices are logically mutually exclusive if IDX_W-HP is being used as an index. Which one you use depends on how you answer the question "If two cars have identical weight-horsepower, weight-torque and suspension ratings, which one would be faster on our Solo 1 tracks, the car with less weight or the car with more horsepower?"


Eric, et al:

As David Pratte has mentioned, the purpose of a classification system is to work in such a manner as to retain existing competitors and attract new ones.

The function of a performance index is to rank cars in a way that would be generally agreed-upon by those with experience (ie, our competitors). Choosing breakpoints within the index to group cars together for competition is a subjective exercise, naturally, as is devising a schedule of prep points or table of authorized mods is also subjective.

Have a look at the spreadsheet the HADA folks posted to their website. Are these cars ranked correctly, and if not, which ones are incorrect?

I believe that a straight-indexed approach can be designed so as to provide 'close' competition amongst both stock and modified cars. The fundamental innovation in the currently proposed system, as compared with anything we looked at previously, is the use of a 'ladder' to 'step up' cars as they are modified so that they compete with other, lesser modified cars from further up the ladder and other, more heavily modified cars from further down the ladder.

Previously, we were completely separating stock, super stock and street prepared levels of preparation and creating a sort of 'pyramid' of classes which, by definition, required there to be a large number of purely stock classes. In order to keep the overall class count down we couldn't have anything like 12 separate root classes. The less divisions, the larger the PI spreads within the classes, the more likely competitors will have the 'wrong' car.

Furthermore, since most of our competitors are not bringing stock cars to Solo 1 events a large number of classes would be permanently underfilled.

A flat, 'laddered' system based on a performance index that works would allow all cars to be modified up to the level the owner desired. We simply have to ensure that there are enough classes on 'top' of the ladder for those with level 1 and 2 cars to make a reasonable series of steps. Nobody would have to 'take parts off' of their cars to avoid being in a class dominated by much faster cars. That is one reason why implementing such a change could work for 2003.

I have made my views on this matter clear, but for the record I will not be forcing this down the competitor's throats. Rather, I want as many competitors as possible to give us their feedback on their preferences, and from this feedback the Solo 1 Committee will decide how to proceed.

Chris Sorensen

[Eric]

Dave wrote:

=============================================
Eric wrote

quote:
----------------------------------------------------------------------------
Under the straight indexing system proposed the actual sub-indices may change, but the 4 attributes they are measuring won't.
----------------------------------------------------------------------------



This isn't quite true. Under Christian's straight indexing proposal he uses weight instead of horsepower, along with the handling index, weight-to-horsepower, and weight-to-torque. =============================================

Dave, your statement is actually an agreement to my quote. The four attributes we are both talking about are horsepower, weight, torque and handling. Each one of these is considered in the sub-indices used in Christian's straight indexing proposal. In my examples I didn't combine the attributes when I came up with the sub-indicies in the interests of keeping it simple for illustration purposes. It does not mean I am an advicate of having one sub-index for each attribute - in fact quite the opposite is true.

What would help in this discussion is if you could post the ranges used for each of the sub-indicies in the proposed system, so that we can assess how relevant those ranges are to various groups of similar vehicles.

Eric J.

[Chris P]

Class PI-Range Varience
------------------------------------

A1-------87.8-80.0-------7.8
A2-------79.0-74.5-------4.5
A3-------74.0-69.6-------4.8
A4-------69.5-67.0-------2.5
A5-------66.7-64.1-------2.6
A6-------63.9-60.0-------3.9
A7-------60.0-57.0-------3.0
A8-------57.0-54.0-------3.0
A9-------54.0-51.1-------2.9
A10----- 51.0-48.2-------2.8
A11------48.0-40.0-------8.0
A12-------40.0-17.4------22.6



That is based off of Christians proposed system

[Shaman]

It should be noted that HP differences make MUCH more difference at larger tracks. Even the Pro track at Shannonville is not a great "horsepower track," it's just barely fast enough to give the high power cars any advantage at all. Most places, I'm convinced it's more the torque (delivered to the wheels) and the way it is delivered that keep the high horsepower cars as fast or faster, since they can fall "off the boil" and still blast forward, they simply have more room for error and are never for want of power. Of course, you also have the Sorensons that will blast right over curbing with disregard for suspension components to avoid that... ;)

What I'm saying is that as soon as you get a track that doesn't match the ones we have now (say, we went to Mosport and ran the big boy), your handling indice would take a big back seat to power, though it would of course still be significant. Then, when the 450hp Camaro with stock suspension gets out on the track, it is going to easily blast the doors off, say, a NSX-T simply by seconds shaved off on the straights (especially the back one). So where it will fall down - without lots of particulars for each particular track - is in actual implementation. Likewise, that 450hp Camaro is going to have a hard time making use of any of that extra power on Nelson - you just get your foot "in it" and it's time to turn again... all that power is wasted because it never gets to do its magic.

I can certainly see where Eric is coming from, and this is my own spin on it. If you're not grouping similar cars, you're going to run into this sort of thing. Cars and tracks are simply not equal, or one-dimensional.

[Dave]

Eric wrote Dave, your statement is actually an agreement to my quote. The four attributes we are both talking about are horsepower, weight, torque and handling.

Doh! When you put it that way, I see your point. We do have four basic sets of data we're using to create the various indices.
I hate it when Eric's right! :p

I'll post a fully functional spreadsheet to the HADA website so anyone who wants to play around with it can download it directly.
I'll throw a link to it up here once it's ready to go.

Cheers,
Dave

[Eric]

Dave Pratte,

It is unfortunate that you couldn’t make it to tonight’s Solo 1 Committee meeting. At tonight’s meeting I officially stepped down as the Chairman of the Car Classification Committee. While I still hope to be involved somewhat with the CCC, I am no longer in a position to commit the time I previously did to chair the committee. As you know I have a very precious baby girl now and I feel it is very important for me there for her and for our family especially during her development years.

I know you have volunteered to become the new Chairman and I would like to be the first to congratulate you on your new position. I think you are an excellent choice. You are devoted, full of energy; seem to have the necessary time available and always have a good sense of humour. I am sure you will do a fine job and do what is best for the sport. Good Luck!

That being said, I still don’t agree with your viewpoint on a flat indexing system. I have had a look a Christian’s proposed linear system spreadsheet. As I suspected it is full of ranking problems. I plan on doing a separate post looking at those problems class by class, but for this post I’ll address some general concerns and respond to some issues you raised in your last post.

Firstly, look at the sub-indices used in this proposal: handling (40%), WEIGHT to horsepower (40%), WEIGHT to torque (10%) and of course don’t forget WEIGHT (10%). A little heavy weighting to WEIGHT, don’t you think? :)

I want to look at the ranges for the sub-indices and the implications of the ranges selected in a separate post as well, but in case someone else wants to pick up on this train of thought, here they are:

Weight to HP: 6.0 lbs/hp to 35.0 lbs/hp
Weight to TQ: 6.0 lbs/ft-lb to 35.0 lbs/ft-lb
Handling: 0% to 100%
Weight: 0lbs to 7,500lbs

(BTW, thanks for sending me the spreadsheet)

Based on my original post to this thread, the obvious place to look for problems is with exceptionally light weight cars since the maximum relevant weight for small light weight cars is nowhere near 7,500lbs and since the weight attribute is so heavily favoured in this proposal. This proposal has these light weight cars classed as follows:

Lotus Super 7 vs. Corvette C4 ZR1
Integra Type R vs. Camaro SS LS1
early Toyota MR2 vs. Camaro SS LS1
MINI Cooper S vs. Camaro LT1
Austin/Morris Mini Cooper S vs. Camaro IROC
base MINI Cooper vs 5.0L Mustang
Fiat X 1/9 vs 5.0L Mustang

The really light weight vehicles are the most poorly classed. At the DDT some of these groupings may be competitive, but on most tracks we run the light weight car is going to get killed.

You mentioned that under this proposal my example of the 91 Civic Si and the 02 RSX Type S show a significant points spread with total PI's of 50.7% and 64.9% respectively. The 91 Civic Si indexes out as an A10 car and the 02 RSX Type S indexes out as an A5 car. What you failed to mention is what happens to the Civic as you modify it. The intent of the linear system is that an SS2 level Civic goes up one class to A9, an SS1 level Civic goes up to A8 and an SP Civic goes up to A7. So basically a street prepared level, light weight ’91 Civic that is permitted an engine Swap to the RSX Type S 200hp motor gets to compete in A7 while a stock RSX Type S competes in a class 2 levels higher in A5. That’s hardly fair! Furthermore, this modified Civic will be able to proposition the stock RSX Type S.

It looks like in this proposal the point spread between these two vehicles was achieved only because the WEIGHT attribute was very heavily weighted. As I had suggested in my earlier post, by fixing one problem you create another in a linear system which is exactly what happened here. These two vehicles obviously shouldn’t be so far apart in PI. If you tighten up this PI spread by changing the relative weighings of the sub-indices I am fairly confident that the a problem will crop up somewhere else on the scale.

Back to the drawing board,

Eric Juraschka

[Chris P]

Eric, what do you think of this idea for dealing with modified cars.


-------------------------------------------------------------


I've been looking over how to convert Prep points into index points. Basical, it seems to work pretty good. However a few issues, one being the lower classes(A12 and perhaps A11) where the index spread is high. I donno maybe a 22.6 point firefly could run with the top A12 car :) Maybe using a different Index point coeficient(explained a little later) number will help.

Basically my vision for the index system would work by assigning an index amount to modification. I think the hard work has already been done. The wieghting was already completed with the creation of prep point schedual. All that needs to be done is to make a good ratio for converting 1 prep point to XX index points. A 1:1 ratio seems a little low and a 1:2 ratio seemed a little more resonable. Try putting a few preped cars into the mix.

For example, with the prep point method my 1st gen CRX has 8 points, so it would move up to A8. I would probally pull 1 modifcation out of the car to make 7 points so A9 would be my destination. Is that fair, no, the A9 cars would suffer under the mighty 1st gen CRX.

However, with a Modifcation index using a 1:2 ratio it would be

CarPI+(PrepPoint*IndexCoefficient)=Total Car PI

so 47.5+(7*2)=61.5 the new class would be A6 and i would be allowed 2% more modifications. A6 seems a little more resonable for my car then A9 wouldn't you agree?

I'll do a current CSS1 Integra Type R, 69.7+(3*2)=75.7 The cars new class is A2 and allowed 4% more work to be done before moving to A1. If you look at some times from Pro track, i did a 78.599 in Alex Burnes Type R. The E46 M3 is also in A2 and Blair P managed a 78.494. Hmmmm, that seemed to work alright but maybe the value "2" is a little high in this case.

The other issue with both modification systems is dealing with cars that have a higher proformence potiential. With my little system all you do to solve this is determine a cut off point and change the index coefficient.
With this, it really doesn't matter where you car starts, bottom of a Index group, or top of a index group. The car at the bottom can do a few modifications so that it is now compeditive with the cars at the top of the group.

------------------------------------------------------------------------

[Dave]

Eric, thanks for the kind words. I really appreciate it and I'm sure the CCC will consult with you regularly as the reassessment process continues.

Taylor and Wired Motorsports has been kind enough to host a downloadable version of the MS Excel file containing Christian's straight indexing spreadsheet. Anyone who'd like to play around with it, follow this link:

http://www.wiredmotorsports.com/reclassproposalflat.xls

I agree with Eric that this version of the flat index has quite a few potential problems with respect to car match-ups. The Lotus 7 continues to be a concern to me because it ranks so high. I personally think the index weightings used in this version of the flat index spreadsheet over-estimates the importance of curb weight and under-estimates the importance of horsepower, but overall the ordering of cars doesn't look like a bad starting point IMO. I think a lot of the match-up problems you've identified can be sorted out with careful "tuning" of the index weightings, but that may or may not open up other mismatches. I'm still optimistic that we can find a good balance that has a minimal (and acceptable?) number of "problem" cars, but if we can't I'm certainly not going to recommend a change of this magnitude to the series. We will have to measure the number of "problem" cars against the number we have now, as well as evaluate all the other relative advantages/disadvantages of the flat index vs. a family system. Time will tell and I can't believe I signed on for more of this punishment! :)

Cheers,
Dave

[Dave Barker]

I agree with Chris P's analysis ( do you have to be a Hada member to refer to him as Krispy? sounds like burnt toast or an overpriced donut). I would see cars advancing on the basis of their prep points i.e. a PI increase of 2 for each prep point, and not having the break outs as previously prescribed ( SS2 , SS1 and SP ) . This could mean that a 12 point car might increase its PI by 24 which seems pretty high and I think we could pro-rate the scale so that possibly beyond 7 prep points the increase ratio in PI would be 1:1.
One of the big advantages to this would be that cars that are lower in their root class might put in a point or 2 and not move up the ladder where cars at the top of their root class will move up given the same mods. ( Seems fair to me and much more fair than our present system where both the top and the bottom of any given class have the same points to play with at the next prepared level even though their PI spread may be as high as 15 to start with )

One issue though that would have to change from Christian's analysis would be that the break outs between different root classes would have to be the same , i.e. an increase of 5 or 6 PI between all classes.

I stongly believe that Eric's concerns can be dealt with by how we calculate PI and no it is not perfect but I think the linear system deals with issues that arose from the survey , namely too many classes , a thumbs up to the PI calculation principle and a general thumbs down to the famly classification . Just think of it being a logical extension of shedding the B family and ending up with a 2 family system to ending up with a one family system .

[ADAM]

if you look at the spread sheet i think it is pretty fair in the way the cars have been classed....


as far as how cars progress...in the power dept its actually a very simple way to calculate the "potential" power of a car, we can easily get to i think around +- 5% using some simple engine math calculations....basically.....an engine is a big air pump...so how much air is it sucking in.....thats the power level it can achieve...simple as that... (we would assume the designer would use the correct A/F ratio)

the hard part of the PI system...is the handling aspect of it....which i really have no clue how anyone can address that...

...other than as the cars get higher prepped...the handling aspect will converge regardless of the car........this is conversly different from the power index...which will diverge at higher levels of preparation.....

meaning.......a 4cyl car with very high points will handle about the same as a very high point V8 car......BUT the very high point v8 car will out power the very high point 4cyl car....

so all things being equal the V8 car will be superior (ahhhh am i actually saying that ) this would be at the extreme end of the scale.....

this is why we need some sort of sliding system to account for the difference in power levels that will occour at high point levels between various displacements...

and we are in luck...since that can be calculated

[Rob McAuley]

A couple of points...

First, I love Chris P's suggestion of converting prep points to PI, and allowing cars in the bottom of their class to move up and be more competitive. We could even support fewer classes, and have more competitive cars in them! Those in the "Black Hole" of C3 (or A12) can appreciate that.

Second, can we not do some regression analysis to determine the appropriate weightings and factors on the primary indicies? I remember using SPSS at university to do something like this. Surely someone with some influence on the CCC has access to a big university system that can crank out the appropriate factors!

[Dave]

haha! Rob, if you only knew how many hours I've spent doing regression analysis (and therapy!) using SPSS. My Masters thesis was based on this type of analysis for the most part. Oh, the memories! But these days I'm more of a qualitative researchers, so my SPSS skills are definitely rustier than Krispy's old CRX :p But hey, I'll crack open the old text books and boot up my DOS (!!!) version of SPSS and play around a bit. And if that fails I'll see about getting a Stats prof to help me out with it. Could be worth the effort!

[Chris P]

hmmm, were learning SPSS stuff in my Decision analysis class so i have access to the software, do i know how to use it.......ummm, lets see what happens in my lab tommorrow. :eek:

rusty CRX, i think you better devert your attention from my car to Wes's, that thing is looser then a school girl after being sweet talked too by Preet........opps, this isn't the HADA board......