The View From Wisconsin

Just a random set of rants from a Sports Fan from Wisconsin.

Monday, May 09, 2005

Extended Racing Points

Several years ago, when it was fashionable to bash NASCAR's Winston Cup Series point system, I toyed around with an alternative method to determining a champion in a given season.

The idea came from the then-new Stock Car Racing Encyclopedia, Macmillan's attempt in 1996 to create a NASCAR version of their venerable Baseball Encyclopeda. In this tome, Peter Golenbock and Greg Fielden presented a simplified means of determining success as a driver: Performance Points. It wasn't anything new, of course; Fielden took the existing Formula One point system of awarding points for the top six finishers (10-6-4-3-2-1) and applied it to every driver in NASCAR's Grand National/Winston Cup series history.

Only one problem: if I wanted to actually go through and see who finished in sixth place for every single race over the past season, I'd be spending more time in front of the computer than I really wanted. So I made the first "adjustment" to the Performance Points system by awarding one point for all finishes from 6th through 10th. That way, all I would have to count would be the top-five finishers, and once I had those totals, simply subtract them from top-10 finishes as was commonly listed on and other racing sites. This system is what I called Adjusted International Points, or AIP.

The idea was enough for a while, but as it came out in 2000 that NASCAR was seriously considering tweaking the system, I started to play around with AIP. I got a basic idea of what the Frances might want in terms of a points system: every driver gets a point for starting; bonus points for leading the race; winner should get more points than anyone else.

So I made some small changes to AIP: first of all, I wanted to extend the point awards out to the top 20 drivers in a given race. The concept was that if there wasn't much difference between finishing 21st and 43rd, the rolling wrecks wouldn't stay out there just to gain points. To do this, though, I had to bump up the award for finishing in the top 10. The simplest way to do that was to simply multiply AIP by 10 – that is, award 100 points for a win, 60 for 2nd, 40 for 3rd, 30 for 4th, 20 for 5th, and 10 points for each finish 6th through 10th. Then, you award 3 points for 11th through 15th place finishes, and one point for 16th through 20th.

To appease the promoters who want every starter to get at least a point in the standings, I added one point per start. This actually bumps the awards by one, but it's really a "bonus award" for the driver who starts the race. I also made the same decision that NASCAR did in 2004 – award an additional 10 points to the winner of the race.

Everything was going pretty well with this new system, with one exception: most of the major racing stats sites didn't include "races led" in their summaries, but instead gave "laps led." Thus, I had a difficult decision – how can I award points based on simply laps led and not by the NASCAR-traditional races led? The solution was an elegant compromise of sorts: if a driver leads a lap at any time during the season, he (or she) is awarded five points in the standings. For every 50 laps that the driver leads during the season, he or she gains an additional point in the standings.

For example, in 2000 Rusty Wallace led the most laps in the Winston Cup Series with 1,731 of his 9,925 completed laps. He got one point for that first lap he led, and 38 more for the remaining 1,730 he led, for a total of 39 bonus points. His AIP for 2000 was 73 (4-1-1-3-3-8), so that made for 769 points. Add his four wins, and he had 809 points. He had 25 top 15 and 26 top 20 finishes, so that makes for 825 points. Add his 34 starts to that total, and you have his final Extended Racing Points (XRP) total of 859. That was good for only fifth among Winston Cup drivers that season; Bobby Labonte ran away with the title with 1,120 points.

I'm obviously thinking I'm onto a good thing, here, and I think I even posted this to the newsgroup at one time. Then came a small argument among some of the posters on that vaunted newsgroup: what about pole positions? NASCAR has always been loathe to award points based on starting positions, mostly because of the rainout effect – the persons starting on the pole might be there because of their points standings, and not because of qualifying times, when there is a rainout of qualifying.

But there were those who were adamant about awarding something for winning the pole for a race. To be fair, I decided that the two drivers in the front row of most NASCAR races should split a 10-point bonus for posting the fastest two times in qualifying. The pole-sitter would get 6 points, and the outside pole-sitter would get four. To adjust for this in XRP, I changed the bonuses for 11-15 finishes to 6 points and 16-20 finishes to 3 points. This total, then, was what I called New Extended Racing Points, or NXRP. (When Nextel announced they were taking over as series sponsor from Winston, I decided that the N should probably mean "Nextel" instead of New.)

Of course, Nextel and NASCAR decided to go with the Chase for the Championship format in their last five races, which actually blew away an amazing season by Jimmie Johnson and teammate Jeff Gordon in 2004. Johnson racked up 1,511 NXRP, the single highest total since I began recording these points in 2000. What was ironic is that only two of the top 10 NXRP totals since 2000 were of drivers who won the Nextel/Winston Cup – and that was only Labonte and Gordon in the first two seasons (2000-01).

I honestly don't think it's a bad thing that NXRP deems another driver to be better, because the two systems are measuring different things. The correlations between the two point systems aren't that much out of whack: every single year had a correlation of .79 or greater between the two systems, with the new Nextel Point System for the Chase for the Championsip correlating at a five-year high of .86.

For those of you who didn't pay attention during math class, correlation is determining the relationship between two sets of numbers through a statistical formula. Theoretically, if every number in group A has a correlated number in group B, the two sets are considered to have a correlation coefficient of 1.0. Basically, the closer that coefficient is to 1, the greater the similarity between the two groups of numbers. Ideally, you would want the coefficient to be somewhere in the high 0.90's, but getting it to .86 is good enough in the real world.

I have not, as of yet, sat down and crunched the numbers for this season's Nextel Cup drivers. When I do, I'll post an update in a future entry.