Based on Round 2’s tendency to upset our mathematical model for football predictions during the Illinois High School Associations playoffs, we have generated a new value for α and Δ for each game from here on out.
Under the old model, which was based on all games played so far, we had 27 upsets of the 64 playoff games in Round 2.
The resetting was done in such a way that only games played on or after Oct 1 were counted. This reduced the number of “upsets” in Round 2 to 18. And, the win by North Greene over Abingdon-Avon in Round 2 doesn’t look as extraordinary as it once did.
What this resetting essentially does is completely throw out the performance of the teams during August and September, using only games played in October and November when computing the team strength.
The playoffs are now in quarterfinals
Based on the new numbers, we make the following predictions for quarterfinal games, shown here with the percentage chances we give the winners.
|
Class 1A Carrollton over Bismarck-Henning, 57% Central-Southeastern over North Greene, 51% Stark County over Forreston, 72% Galena over Lena-Winslow, 85% |
Class 5A Geneseo over Sycamore, 60% Montini over Marian, 55% Sacred Heart-Griffin over Taylorville, 78% Peoria over Morris, 77% |
|
Class 2A Maroa-Forsyth over Athens, 62% Tuscola over Chester, 50% Fieldcrest over Clifton Central, 59% Rockridge over Eastland-Pearl City, 54% |
Class 6A Hinsdale South over St. Francis, 53% Notre Dame (Peoria) over East St. Louis, 66% Nazareth Academy over Lake Forest, 55% Lemont over Richards, 58% |
|
Class 3A Williamsville over Auburn, 57% Newman over Byron, 92% Robinson over Mt. Carmel, 57% Wilmington over St. Joseph-Ogden, 67% |
Class 7A Cary-Grove over Geneva, 73% Lincoln-Way East over Mt. Carmel (Chicago), 76% Libertyville over Fenwick, 83% Providence over Wheaton Warrenville South, 60% |
|
Class 4A Bloomington Catholic over Rochester, 58% Herrin over Carterville, 65% Coal City over Manteno, 56% Rockford Lutheran over Phillips, 60% |
Class 8A Stevenson over New Trier, 66% Simeon over Naperville Central, 78% Glenbard West over Barrington, 62% Homewood-Flossmoor over Bolingbrook, 68% |
Our metholdology, in brief
The predictions in the above table were based on a simulated playing of 100 football games between each pair of teams. Given their past performance on offense and defense, compared with the expected performance based on their opponents’ average performance over the same period, we simply ran 100 games with the help of a random number generator.
To compute α and Δ, we use the following model:
where a = average points scored by Team t, b = average points given up by Team t’s opponents, c = average points given up by Team t, and d = average points scored by Team t’s opponents.
In other words, the model uses α and Δ to represent team strength and then just subtracts one team’s strength from the other for each pair to predict which team will win the game. The percentage is not a point spread; it’s just the number of simulated games, out of 100 for each pair, that the winner came out on top.
In the case of a 50 percent prediction, the simulator predicted the same number of wins for each team in the pair. However, in every case in this year’s quarterfinals, one team had a slight advantage in α. Whenever α and Δ predicted different winners, as it did in one game above, we called the game for the α winner.
Look for upsets in the quarterfinals
If I were to say which games might be upsets in this round, at least in terms of the games that would cause our nice mathematical models to go haywire, I would have to say Mt Carmel and Fenwick, both in 7A.
The Caravan of Mt Carmel have a knack for winning championships under coach Frank Lenti, and Fenwick already brought great shame to our computer when they beat Highland Park in Round 1 of the playoffs and won what we thought should have been a close game against Carmel by more than two touchdowns.
