Most folks, when making a prediction on a season, more or less guess a won/loss number, rounding up for their favorite teams.
If they do examine it on a per game basis, they just guess for each game as well, with comments like “we win that one, it’s at home”, or my favorite “we lost last year, so the team will want revenge and we win”.
A more thorough analysis is to use historical won/loss trends on a per opponent basis, and assign a risked win percentage for each game. It’s more or less what’s called Decision Risk Analysis, as used in engineering and business, where you assign percentages to each action and come up with an overall project likelihood outcome and a risked cost.
Usually I’d use 4 years of data, to capture a full recruiting cycle, but for Texas, that would include the last year of the spectacle that was the Charlie Strong Experience. Coaching changes are the biggest drivers of changes in win / loss trends, and need to be factored in. So for Texas, just the past 3 years of data will be used.
Data for the B12 is easy to use, as every team plays each other during the season, so there’s 3 years of data. For the out of conference games, it’s more difficult. The 3 years under Herman have had two broad categories of out of conference games – P5 teams on a home/away basis, and non-P5 teams that are hired to play.
Herman has had 9 total out of conference games – 5 to P5 teams (Maryland twice, USC twice, LSU once), and 4 to non P5 teams (San Jose State, Tulane, La Tech, and Rice, all wins).
The record against P5 teams is bad – only one win, against USC in 2018, with home / away losses to Maryland, and losses to USC on the road and LSU at home. This poor record is one of the knocks against Herman, especially in the 2017 and 2018 seasons, both of which stated with losses to Maryland.
Below is a listing of the 2020 opponents (in schedule order), won / loss records for the past 3 years, raw risked win percentage, and then adjusted win percentage. Per game discussion follows the table.
2020 Opponent 2017 2018 B12 CG 2019 Percentage Adjusted %
S Fla Mar - L Mar - L LA Tech - W 20% 0.75
LSU SJS - W Tul - W LSU - L 20% 0.25
UTEP USC - L USC - W Rice - W 100% 1
Kansas State W W W 100% 1
Oklahoma L W L L 25% 0.25
West Virginia W L W 67% 0.67
Texas Tech L W W 67% 0.67
Baylor W W L 67% 0.67
Kansas W W W 100% 1
TCU L W L 33% 0.33
Iowa State W W L 67% 0.67
Oklahoma State L L W 33% 0.33
Total wins 6 9 7
sum 7.59
First up is S Fla, who fired C Strong at the end of last year, and has a new head coach.
As noted, Herman’s record against out of conference P5 teams is 1/5, while for hired chumps it’s 4/4. S Fla is not a P5 team, nor a hired chump, so difficult to categorize. Closer to hired chump that P5 team, and while Charlie may have baked another cake, he also left another team with a losing record. I adjusted this win percentage to 75%. Texas should win this game, but just like the Maryland double losses, it’s not out of the question that they could lose.
LSU – clearly in the P5 bucket, so raw win % is 20%, rounded up to 25%. This game is what I refer to when I say that most folks just guess or pick a winner based on who they root for. Comments for this one will refer to “revenge, payback, the team’s mad”, etc. Based on 5 previous P5 regular season games, Herman’s won just 1, so his history gives a low win percentage.
UTEP – 100%, hired team.
Now onto the B12, where there’s 3 years of data for each team. For these, I’ve gone with the historical win percentage for the risked wins for 2020, as they all seem in line with trends. KSU at 100%, OU at 25% - harsh but Herman has won only 1 of 4. OU will have a new QB, but that’s been the case for the last 3 years, without issues for them.
West VA, Tech, Baylor, ISU all 67% - more than likely to win each game, but not certain based on history. TCU and OSU at 33%, again based on history. The OSU win was the best win of the regular season last year, an indicator of how poor the season was.
All that total up to 7.59 wins. Rounding up, and with the fact that Texas does have a senior QB who’s one of the better ones in college ball, pencil it in at 8 wins.
Hindcasts are useful in seeing how well a method performs as a forecast. Using the same method for the 2019 season, it predicts 8.33 wins.
2019 Opponent 2017 2018 B12CG Percentage Adjusted %
LA Tech Mar - L Mar - L 100% 1
LSU SJS – W Tul - W 25% 0.25
Rice USC - L USC - W 100% 1
Kansas State W W 100% 1
Oklahoma L W L 33% 0.33
West Virginia W L 50% 0.5
Texas Tech L W 50% 0.5
Baylor W W 100% 1
Kansas W W 100% 1
TCU L W 50% 0.5
Iowa State W W 100% 1
Oklahoma State L L 0% 0.25 (never 0% chance for Texas to win)
Total wins 6 9
sum 8.33
Based on the good Sugar Bowl win to cap off the 2018 season, I’d have rounded this up to 9 wins. Instead Texas won only 7 games last year, showing both that the method isn’t unduly pessimistic, and how disappointing a season last year was.
So, based on this method, I predict an 8-4 season, with losses to LSU and OU, a loss from either of TCU or OSU, and then one from the W VA/Tech/Baylor/ISU grouping.
Would be nice if it was higher, but there just isn’t the data to suggest that. Texas regressed in the 3rd year of Herman, not improved, with pitiful losses to Baylor and ISU, and 5 in total. By the end of the year, the DC was fired, as were lots of other coaches. It’s hard to realistically see more than 8 wins on the schedule based on prior results.
-
Like x 2
-
poop x 1
Last edited: Jul 28, 2020