I promised a Classic Car valuation calculator to anyone interested in a copy. I finished it last night and have written a short instruction guide.

There was quite a bit of discussion regarding the applicability of statistic modeling with classic car values. I don't want to get into that discussion again - but I DO want to make sure that anyone who wants a copy of the calculator knows that it is available.

The calculator will give an objective price for cars in different quality levels (a 50% car is average in quality - a 60% car is better than 60% of the cars sold in the same time period, etc). Determining what percentile a given car fits into - that is still highly subjective, and something the calculator can't help with.

If you want a copy, PM me. I'll e-mail it out to you. If you don't, that's fine too.. I'm offering it to everyone, but not forcing it on anyone. And it might be fun even if you just play around with it. I do ask that if you use it and then publish the results, that you use it legitimately. If you plug in crap data you'll get crap for a result - please don't do that and then come back saying how silly it is. Follow the instructions and you'll see that it works...

 

Pm'ed ya.

 

I did express doubt in your other thread but I'll give it a try for my 1974. Yes I am objective. I have PMed.

 

Everyone who requested a copy (thus far) and gave me an e-mail address should have it. Let me know if you do not...

 

Make a believer out of me, PM sent.

 

PM sent

 

We all know a stick is more valuable than an automatic. A convertible is more valuable than a coupe. A 68-72 is more valuable than a 73. A 73 is more valuable than any later year. A LT is more valuable than a stock 350. Leather is more valuable than comfort weave. A black interior is more valuable than saddle. Original color is more valuable than a color change. Likewise a numbers matching engine. But how much more or how much less? That is a judgment call. Also how to do raise or lower price based on condition? Another judgment call. That is the fallacy of your program. Quantifying your personal judgments. That is followed by each user using their judgment to self rate every car for each category of your program.

In the end, your program has little value. It fails to quantify judgment.

 

Of course it doesn't quantify judgement. It's a tool - no more and no less than a business executive looking at similar data before making a business decision. The business executive still has to interpret the data and apply it to his/her situation before making a decision. But does that mean you call the data 'worthless'?

The calculator will give you a breakdown of different values for different qualities of cars of the same make, model, and option package (or whatever you use in your sampling), and then tell you based on condition what the value is for that car in each percentile. Yes - you still need to judge the car you are looking at and decide where in that scale it fits. And yes - that is still largely subjective. But the calculator does give you an accurate scale to fit it to.

If you take a wrench and fit it to a bolt do you throw the wrench away because it doesn't turn the bolt for you? This calculator is a tool, but it doesn't do the whole job. It just gives you good information to help in your decision-making.

 

I totally respect your program. The problem is that 10 people using your program will likely get 10 different results.

 

They would vary if they used different sampling techniques, but I think you'd find that the results of the program don't vary all that much when used consistently. What varies is the subjective part that comes into grading the car you are applying against the calculator.

 

yes but the same person using the program for himself on 10 different cars could get good data because his judgments will be the same for all of the entries.

 

You don't put in judgments. You put in random actual sale prices for cars matching the one you are looking at. The calculator then generates a normal distribution (aka bell curve) showing with a 99% confidence level where all sales for that make/model/option package are going to fall. Rather than displaying the actual bell curve it shows what the different car values are at different point along the bell curve. In other words it says "10% of the cars of this make and model and with this option package will sell for $X or less," and so on for every percentile up to 90.

That part is accurate as long as you really do get a random sample of cars from the same make, model, and option package.

If I'm looking at a 1978 Corvette L48/manual, I would go online and grab maybe 20 recent sales from my geographic region of 1978 Corvettes that have the L48 with a manual transmission. The calculator will then tell me what a 1978 Corvette with an L48 and a manual transmission is worth at each percentile.

Now when I go look at a 1978 Corvette with an L48 and a manual tranny, if I can gauge where it fits on that scale, I can accurately gauge it's value on the open market.

That's how you use it - and it will work when used that way.

 

I would like to try it, it could be a cool I-phone app if its possible to make it so.

 

Not for those of us who own automatics.

 

 

I think I understand where some of the confusion is coming from. When you enter random samples of recent car sales, all you enter is the price. You do NOT need to judge those cars, and should not concern yourself with their quality. The calculator gauges the price for different quality percentiles for you. That's where the statistics come into play...

The ONLY car you judge the quality of is the car you are thinking about buying (or selling - you could use the calculator to sell a car just as easily as to buy one). You don't key that in anywhere either. You just decide if a car is in the 10th to 90th percentile (10th percentile being a car that runs but needs lots of work or if it doesn't run the body makes up for that, the 90th percentile being a show quality car, and all the other percentiles being somewhere in between) and then you look for the appropriate value on the calculator.

 

I once met an accountant fascinated with statistical tools done on the computer with which to make decisions where people were involved in both sides of the transaction.

Wonder what he's doing nowadays, government work?

You sure are determined, WG. best of luck. If you see William The Accountant be sure and wave to him. He might need a lift by now.

l.p.

 

Believe it or not Lance, but I can relate to that comment. I'm not an accountant. I'm on the business/operations end. But I do have an MBA (in Lean Manufacturing), and I'm almost done with a Masters of Science in Manufacturing Operations.

Yes - I love statistics, but as a lean guy I absolutely abhore statistics that take a long time to make and then don't tell me anything. And yes - the number one mistake made with statistics is to use them where they don't really apply. There is nothing like looking at numbers all day when none of them tell you anything.

But the number two mistake I see with statistics are people going on intuition when data is readily available and statistic modeling is applicable. And I can tell you that stats, while not the be-all-and-end-all, will 'out guess' intuition. You might beat well applied stats sometimes, but consistently? Not unless you are doing a better statistic model yourself... You won't do it guessing.

Classic cars are an excellent example. Used car sales for a given make/model/option package of car fall into a normal distribution. Some sell much higher than the average and some sell for much less. Some really crappy cars sell for a fair chunk of change. Some really nice ones sell cheap. But they fall within a normal distribution. The overall condition of a group of cars is normally distributed too. You'll have an average quality, and then you'll have quality levels normally distributed against that mean. If you take a used car (classic or otherwise - this works for both) of any given condition and chart where against the 'mean' for that car different sales prices fall, you'll find that for each condition a car can be graded in the prices are normally distributed too, and that the normal distributions for the different grades fall into a normal distribution pattern when charted against the total distribution.

What all that means is that if I take a normal distribution for ALL sales for a given make/model/option package and then chart out different sales levels, it'll match different quality levels too.

You can do regression analysis to check that if you like. Used car prices for a given make/model/option package have a STRONG coorelation to vehicle condition.

Mix differnet option packages and you'll still get a normal distribution but it will get wider. You can still use it - and it'll still beat even an educated guess - but it will show a wider distribution than if you at least separate an L48 from an L82.

Mix multiple makes or multiple models in a single sample, and it's pretty much worthless. You'll still get a pretty little bell-curve, but it'll be too wide to be of any practical use.

I can't tell you accurately what a given car is worth, but I can put in your hands a spreadsheet that will tell you that for this given make and model with this option package (or at least engine), sale prices are normally distributed, based on vehicle condition. And because of that I can give you meaningful business intelligence to go shopping with - the median price for each possible vehicle condition.

Are you really telling me that you can't make a better decision with that kind of information at your fingertips? That's like telling your kids not to go to college because they'll have to take out loans. If they get a good education in a field people hire, they'll also be able to pay those loans back while driving their new Corvettes.

If I'm going to drop tens of thousands of dollars on something, and if statistics will help me make a more informed decision about that purchase, you are damned right I'm going to use statistics.

Statistics don't lie. People lie. Sometimes they lie with statistics - but it is the people and not the statistics that are to blame.

I think some people want something where they can take a picture of a car and it'll say, "$15,500". I can't do that. But I CAN give you a sound price guide for whatever car you want to look at - provided you don't mind taking the time to gather random pricing information to plug into the spreadsheet.

 

Good for u. I ran a big n.american co. For years...saw it all.

 

It most certainly would. As an IPhone app it could be programed to do the sampling too! That's a great idea! But I'm not an I-phone developer... Sorry! If you know one I'd be happy to talk to them though!