Probullstats Blog

Oct 1, 2021, 9:42 am

Predicting Rider Success

By: Slade Long

From the time I earned my PRCA card to today, the sport of bull riding has changed in one fundamental way.

In the old days, there were two variables affecting rider success. The rider's ability, and luck of the draw. Many riders stayed on the majority of their bulls, but it was more difficult to draw a bull you could win money on.

Today, you can win money on practically any bull in the sport. The average pro level rider gets on many high quality bulls in a year of normal competition. So, the most important variable now is: can the rider stay on the bull?

Earlier this year, I decided to see if I could leverage data at Probullstats to predict whether any given rider would stay on a any given bull. This question is similar to odds making in sports betting, but the primary goal in odds making is to encourage bets on both sides. My goal here is to simply be able to predict whether rider x will stay on bull y with some degree of accuracy.

So I wrote an algorithm that can make a prediction for any matchup between bull and rider. Rather than giving a simple yes or no answer, it gives a probability - a number between zero and one, representing the probability of the rider winning the match by making a qualified ride.

For the better part of this year, I've been letting the algorithm make a prediction on every matchup that goes into Probullstats. Once data for any event is collected and ready to be loaded into the database, I let the algorithm generate a probability for each rider/bull matchup using only pre-existing data, then store that along with the rest of the data for each out. This allows me to test to see if the predictions are accurate, and also gives insight into rider performance, bull difficulty, and event quality.

Here are the results of 10,846 predictions:

0% to 10% predicted chance of a qualified ride - 973 outs / 53 qualified rides = 5.5% success

10% to 20% predicted chance of a qualified ride - 2685 outs / 357 qualified rides = 13.3% success

20% to 30% predicted chance of a qualified ride - 3437 outs / 791 qualified rides = 23% success

30% to 40% predicted chance of a qualified ride - 2174 outs / 726 qualified rides = 33.4% success

40% to 50% predicted chance of a qualified ride - 898 outs / 426 qualified rides = 47.4% success

50% to 60% predicted chance of a qualified ride - 284 outs / 157 qualified rides = 55.3% success

60% to 70% predicted chance of a qualified ride - 98 outs / 70 qualified rides = 71.4% success

70% to 100% predicted chance of a qualified ride - 17 outs / 15 qualified rides = 88.4% success

The highest probability the algorithm has given is .788 or 78.8% to Sage Kimzey on Diamond G's -51 Mr. Knockout in Coalville. In general it will only predict over 70% when it's a really good rider on a really easy bull. Riders like Sage, Stetson Wright, Jose Vitor Leme tend to have a lot of higher probability predictions because all of their measurable stats are great already.

Next steps for this are to test to see what the typical prediction score is for outs that end up resulting in 90+ points.