March 18, 2019

# Win Probabilities vs. Ohio State

Below is my first shot at a win probability graph. I realize this graph is not valuable to everyone and, as a matter of fact, some probably think it’s useless.

I’ll just say that the goal of this site is to offer data that you won’t find on other sites and this is an example. But more than that, seeing data like this has changed the way I look at a football game and generally makes me calmer during games (for the most part).

Before I get to that, here’s a quick recap of how these probabilities are determined. I use what I believe is the most important metric, total yards, and turnovers. It’s really simple, the more total yards you have in relation to your opponent the better chance of a victory.

Why do teams with more total yards lose sometimes? There can obviously be a variety of reasons, but generally it’s turnovers. So by combining these two statistics and projecting the total yard differential at various points of the game I know what the likely outcome of the game is going to be.

If Clemson has 20 more yards at the end of the first quarter, this obviously projects to 80 more yards for the game and I know (from other research) what % of teams win when they gain 80 more yards than the other team.

However, if a team outgains the opponent by 80 yards, but is negative 2 in turnovers the win % is going to be less and as you can see from when Tajh Boyd threw the second interception it has a huge effect on the win probabilities.

The problems with this approach? It’s pays no attention to the score. Despite being down 29-20 the model said that Clemson still had a 60.9% chance of winning (because of the total yard differential) and that proved true.

Theoretically, the score of a game could be 28-7 with 30 seconds to go and the win probability wouldn’t be at 100% because the total yard differential may be closer than normal. It happens. Not often, but it happens. I would simply suggest that in a 28-7 game a win probability graph is probably not as useful as in a game like what occurred Friday.

While I understand that many non-metric fans will not agree with this, there is no better metric at determining winner or loser in college football than total yards. Not turnovers, penalties, rushing yards or any other metric.

So even down 29-20 in the 3rd quarter I knew Clemson had a fantastic chance of winning because they were moving the ball. Makes complete sense to me and that’s why it’s changed the way I look at a football game. In the past I would have been petrified at the blown opportunities in the first half and points Clemson had given away and thought “there’s no way we are going to win this game”. Friday, while I certainly wasn’t confident, I knew Clemson had a decent chance to win.

Many of the critics of this type of analysis will point out many times that the team with more total yards doesn’t win. This is true and I would suggest that if you are looking for a model that will tell you with 100% accuracy who will win and lose you’ll be looking forever. These are probabilities. By definition, when I tell you there is a 60.9% chance Clemson will win that also means there’s a 39.1% chance they will lose.

Win probabilities change literally on every play and I could create a graph that reflected that, but chose to chart the turning points in the game. The graph above shows that when Clemson went up 20-9 the win probability was at 100.0%. The assumption was that if the yard and turnover rates stayed the same from that point on there was no way Ohio State was going to win. That obviously changed and the win probability reflected those changes.

Because Clemson was never behind in total yardage and had only one turnover for the majority of the game the model indicates that the Tigers odds of winning never dipped below 50.0% and that occurred when Ohio State tied the game (and total yardage) at 7-7 with 5:44 to go in the first quarter.

This site uses Akismet to reduce spam. Learn how your comment data is processed.