I've looked at 3 different stats over the past 3 weeks, so I thought it would be a good time to combine them all together, look at the 2008 season, and see how well I could predict the games. A quick refresher on how I set up the predictor:
- Each of the stats I broke into 2 Levels (Above and Below Average), and into 4 Levels (Bad, Below, Above, Excellent). I wanted to see if the number of levels mattered.
- For this test, I used the actual results from the games for the Predictor. In practice, this (obviously) won't be the case, so these results will be better than the real thing. I'm just looking at the first half of the predicting process.
Since the Red Zone Efficiency data only goes back to 2001, that's what I'll use to set up the Predictor. I then took the game stats from every game in 2008, and tested the predictor. Here are the results when using 2 Levels for each stat:
Predicted | Actual | ||
Home | Away | Tie | |
103 | 43 | 0 | Home |
16 | 93 | 0 | Away |
0 | 1 | 0 | Tie |
That is an overall record of 196-60, or 76.6%, which is pretty darn good. Again, when used in practice this will come down, but it is a good starting point.
I thought I'd look at each of the stats at 4 different levels, to see if you can give a little more emphasis to better teams. Again, I tested the predictor with the game stats from 2008, and here's the results for 4 Levels:
Predicted | Actual | ||
Home | Away | Tie | |
117 | 29 | 0 | Home |
23 | 86 | 0 | Away |
0 | 1 | 0 | Tie |
That is an overall record of 203-53, or 79.3%, which is slightly better than the 2 Level. It isn't significantly higher, so I think I'll keep both of them around for now.
For fun, let's see how the Colts 2008 schedule looked in the Predictor:
Week | Opponent | Location | Score | Predicted | Probability |
1 | Chicago | Home | 13-29 | Chicago | 65.5% |
2 | Minnesota | Away | 18-15 | Indianapolis | 58.2% |
3 | Jacksonville | Home | 21-23 | Indianapolis | 55.0% |
5 | Houston | Away | 31-27 | Indianapolis | 50.2% |
6 | Baltimore | Home | 31-3 | Indianapolis | 87.4% |
7 | Green Bay | Away | 14-34 | Green Bay | 54.5% |
8 | Tennessee | Away | 21-31 | Tennessee | 51.9% |
9 | New England | Home | 18-15 | Indianapolis | 76.1% |
10 | Pittsburgh | Away | 24-20 | Indianapolis | 63.7% |
11 | Houston | Home | 33-27 | Indianapolis | 58.7% |
12 | San Diego | Away | 23-20 | San Diego | 69.8% |
13 | Cleveland | Away | 10-6 | Indianapolis | 50.1% |
14 | Cincinnati | Home | 35-3 | Indianapolis | 91.3% |
15 | Detroit | Home | 31-21 | Indianapolis | 53.1% |
16 | Jacksonville | Away | 31-24 | Indianapolis | 64.2% |
17 | Tennessee | Home | 23-0 | Indianapolis | 88.8% |
The Predictor was 14-2 with the Colts' games this season, only missing the first Jacksonville game (which the Colts should have won), and the SD game, which they squeaked out at the end. Again, this is all with the actual results, so predicting 14 games is probably a little high, but to know it is possible, this early in the process, is fantastic.
Coming up over the next 3 weeks I'll look at RB Success Rate, Plays / Drive (thanks to Nideak for the suggestion), and 3rd Down Conversion Percentage.