LeMons Overanalyzed

Pouring over Jalopnik reports, laptimes, and team blogs, I was trying to make sense over the madness that is The 24 Hours of LeMons.  The 24 Hours of LeMons race is an endurance race (13 hours over two days) where teams are allowed to spend $500 or less on their race car (aside from safety equipment).  The races are made up of a totally insane mix of cars, from RX-7s to old cop cars to Ford Pintos to an oooooooold Volvo Amazon.
Although I’m not sure I’ll ever think about entering the race (being in a race with a $500 bucket bumping into explosive Pintos seems a little on the dangerous side), I was curious to see how cars from different countries and makes, as well as models in particular, perform.  Although results of the races are recorded, car information isn’t included there (an email to the race organizer asking for the make, model of each car got me the response “Ah, hell, we’re not nearly that organized.”), so I had to cross-reference between what is recorded on mylaps.com and the information on Jalopnik’s uber galleries (which took a bit of time).  I could only find this information for races going back to Thunderhill 2007, but I will update this page as more races are completed as recorded.
Ok, the results are about to come, but first let me tell you how this is being measured (feel free to skip this paragraph if math bores you and/or you don’t care).  On each race, I took the z-score of each entrant to the race, and I recorded the make, model, and country of origin for each car.  Then, for each of those three categories (make, model, country) I found the average z-score for all the cars entered that fit into that category.  In the charts below, I only plotted categories that had at least 3 samples (so the lonely Volvo Amazon, for example, won’t show up in the model chart, but its performance is included in the country and make charts).
Here we go (These stats include races up until Toledo 2008):
Performance by Country

Performance by Country

Performance by Make

Performance by Make

Performance by Model

Performance by Model

A couple of things to point out: the samples I looked at may be very unreliable.  My math shows Caprices kicking ass, but there were only four of them over the three races I looked at, which means that there is a very small sample size (which leads to poor accuracy/precision).  Also information on some cars were just missing.

Since there was a request for the source data I used, here they are.  Apologies for making them .doc, it was the only thing I could put up easily with wordpress (can’t publish excel spreadsheets).

9 Responses

  1. Can you publish the raw data? I’d love to see what the n is within each model category and what’s actually going on.

    And yes, we’re bringing a 944 to the next Lemons race…

  2. It would be interesting to see a breakdown by model year, if you managed to scrape the model-year info as well.

    As you go back in time, the quality of the purchased car when new goes up, but the effects of father time go up as well. There is probably a model-year sweet spot where the car you’re buying is as fresh as possible, but also wasn’t a piece of shit to begin with.

    In more mathy terms, the function “Z-Score = f(time)” probably has a global maximum at some year, and I’m guessing that year is around 1990. Older cars fall apart, and if you can find a newer car for $500, it probably was either shit to begin with or has something very wrong with it.

    It would be cool to see the average Z-scores for cars built in
    …And so on.

  3. Antoun: Sure, I’ll have it up in a few days. Good luck in the race, I hope your 944 can buck the trend

    Bongle: Yes, I had the same idea also in terms of year, but that data isnt available from the sources I checked. If it was available I would add that analysis too.

  4. quaid… start the reactor

  5. The data does not tell me much as this is a public race. The performance comes down to the team itself and how much they invest and know about their vehicles. Professional rally car racing I believe is a better indicator of performance among make and model

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: