By the Numbers: 2019-20 Season Wrap Up
With a weird ending to the season, I want to congratulate all teams this year and also thank them for keeping my winter busy and more enjoyable. I truly look forward to putting scores in each night to see what changes occur in the ratings. I eagerly await each score to come across my screen so I can either think, “This predictions thing isn’t too bad,” or “Yikes, I really need to adjust this prediction thing!” But it the end, I try not to focus on one particular game. So let’s focus on all 5,235 games I was able to record this season. While I want to predict the correct team to win, the numbers I am more concerned with are the error in the margin of victory, the error in total points scored between the two teams, and the bias of total points scored between the two teams. Listed below are those results.
Full Season | After Jan. 1 | |
Average Margin of Victory Error | 13.8 points | 13.1 points |
Average Total Points Error | 15.1 points | 14.4 points |
Average Total Points Bias (+ or -) | +1.9 points | -0.9 points |
Overall, you expect to see these numbers improve for the back portion of the season as the computer gathers more data. While the numbers are still a little high for my liking, I do believe that it is an indication that things are going well. For reference, Ken Pomeroy’s numbers for each category in men’s college basketball are 8.9 (margin error), 12.9 (total points error) and 0.2 (total points bias) for a game that is much more controlled are predictable. Those numbers are mostly for me, but what about the wins and losses?
Date | Predicted Record | Actual Record | Win Percentage |
All Season | 4499-742 | 4278-963 | 81.6% |
November | 239-48 | 203-84 | 70.7% |
December | 1223-226 | 1147-302 | 79.1% |
January | 1598-234 | 1532-300 | 83.6% |
February | 1373-216 | 1335-254 | 84.0% |
March | 65-19 | 61-23 | 72.6% |
State Tournament | 20-6 | 14-12 | 53.8% |
As is typical, the rating system got better as the season went on, minus the State Tournament, which for lack of a better word was a disaster for picking games correctly. Last year, the model correctly predicted 36 of 44 winners at the State Tournament. The non-State Tournament March predictions were 47-11, which was better than the 45-13 expected record. The overall win percentage of 81.6% was slightly below last year’s mark of 81.8%, which out of 5,241 games would have been nine more correct predictions.
Predicted Percent Chance | Record | Win Percentage | Expected Win Percentage |
50-60% | 268-242 | 52.5% | 55.1% |
60-70% | 332-200 | 62.4% | 65.3% |
70-80% | 404-166 | 70.8% | 75.1% |
80-90% | 592-171 | 77.6% | 85.3% |
90-100% | 2682-184 | 93.6% | 97.4% |
This table quite clearly shows that the percentages were below where they were supposed to fall. In seeing the volatility of some of the scores this season, I think I need to pull back the reins on some of the win percentages, which is an easy fix and should help normalize the percentages in each category. Also, I think it is important to understand the impact that November and December have on the overall outcomes. Since the preseason ratings are still being used for most teams through the middle of December, it is difficult to determine how much of a difference I should be willing to accept.
Going forward, as always, the goal is to be as close to perfect as I can be, with the understanding that it will never be completely perfect. I will play around with the formulas between now and November to see if I can create a more accurate system, but the difficulty is, and will always be, rating teams from opposite ends of the state that have zero common opponents.
I want to thank everyone who took the time to look at the ratings this season. It is truly amazing what is being done for girls basketball players around the state. Coaches, parents, journalists, administrators, and the list goes on an on. I look at all of the support that I received this past year and know that the support for the players is exponentially greater, which means that girls basketball in Minnesota is in a very, very good place.
Biggest Upsets
Below are the five biggest upsets this season according to the computer and the percent chance they would occur. I only included upsets that occurred after January 1 so that preseason ratings did not have a major impact on the predicted score.
- Cristo Rey Jesuit 67, Maranatha Christian 65 (0.03% chance of upset) – 1/14/20
- United Christian Academy 47, West Lutheran 46 (0.12%) – 1/4/20
- Hayfield 44, Waterville-Elysian-Morristown 36 (0.16%) – 1/31/20
- St. Paul Academy 47, CHOF 33 (0.27%) – 2/4/20
- Northeast Range 47, Nashwauk-Keewatin 42 (0.34%) – 2/24/20
Highest Scoring Games
Below are the five highest scoring games of the season. The average point total this season was 106.7 per game, over a full point higher than last season. These games far surpassed that.
- 193 – Worthington 98, Windom 95 (2 OT) – 1/16/20
- 186 – Windom 95, Jackson County Central 91 – 1/30/20
- 184 – Rocori 103, Duluth Denfeld 81 – 2/8/20
- 184 – Red Lake 99, Cass Lake-Bena 85 – 2/14/20
- 182 – St. Anthony Village 92, Concordia Academy 90 (2 OT) – 2/22/19