My interest in sports rating, ranking, and prediction algorithms go back over 35 years to an operations research course that I took at Penn State in 1971. For my course project, I used an exponential smoothing technique commonly used with inventory control systems to develop offensive and defensive power ratings for the major college football programs. With those power ratings I predicted the outcomes of the 13 bowl games at the end of the 1971 season. Those predictions were published in the local newspaper (the Pennsylvania Mirror) and my resultant success got me (or, more specifically, my computer algorithm) picked up as a guest columnist for the paper for the following season. However, my football predicting went on hold when I joined the real world in 1974. My passion for it never died, however.
In 1981, my passion was rekindled when I took possession of one of the very first IBM personal computers off of the assembly line. I converted my Fortran computer programs to Lotus 1-2-3 spreadsheets, and I was back in business – albeit only as a hobby. Those spreadsheets still demonstrate to me how far the computer industry has come. Running a prediction cycle for my system would take two separate 6-8 hour runs. I would fire up the first run, the updating of the power ratings, as I went off to the office in the morning, then finish the second run, the score predictions for the next weekend, while I slept the next night.
There was another hurdle that I faced back then. Without the internet, actually tracking down a score for a late Saturday West Coast game was sometimes problematic. Eventually, the effort got tiresome and the passion cooled off once again. Had I lived in Vegas where I could have legally bet on the games, I might have pursued it a little longer.
The next incarnation of my forecasting model occurred shortly after my retirement. Technology had advanced by leaps and bounds – my 6 hour spreadsheet run could now be completed in mere seconds. The internet now provided immediate access to scores around the country. And I now had plenty of time on my hands. That led to eForecasts, my shortlived attempt at developing a web based touting service (perhaps I should have held onto the domain name though – after I let the ownership expire, a domain squatter took it over). My interest waned when my research uncovered that the people who made money in the business weren’t necessarily the ones with the best algorithms, they were, more often than not, the ones who schemed the best to get the money from you not from Vegas (in retrospect, I was pretty naive). There also were still some technology issues in terms of data entry – access was available but I didn’t have ready access to scores in a form that was usable for me.
So why now? Well, for one, I still have an interest and still have the time. Two, interest in computer ranking with the general public has grown tremendously with the advent of the BCS and RPI. Third, sabermetrics, with a firm hold on baseball, is starting to expand to other sports, most notably the NBA. Fourth, Ken Pomeroy’s NCAA basketball statistics, which he so kindly makes available to all, eliminates a lot of the dirty work data entry issues that I’ve had in the past. But most importantly, as with most bloggers, I’m doing it strictly for the passion without any desire to gain any fame or fortune as a result – so what do I have to lose?