I'll be starting a new weekly column here taking a statistical look at the shape of the Conference. So, if there is anything else you would like to know week to week about the ACC feel free to ask in the comments.
Conference Win Totals
These numbers are generated by a model I'm currently working on. I hope to have a full explanation up soon but basically the model estimates each team's contribution to their Expected Points Added per Play and then uses those estimates in a regression to predict the number of points scored in each game. I take those predictions and simulate the season 1000 times to get an expected win total for each team. The model predicts the winner of each game at about a 73% accuracy level, not quite Vegas but good enough to draw some conclusions. Ok, so what does this tell us?
- It's important to note how much a team's schedule plays in to these numbers. Miami gets Clemson, Virginia Tech, and Georgia Tech at home. My model really likes their performance so far this season and they are only underdogs at FSU next weekend.
- My model really doesn't like how Virginia Tech has played so far this season. Even though they don't play Clemson or FSU they aren't the favorite in the Coastal.
My model also produces estimates of how many points better or worse each team is compared to an average team. Here are the numbers for the ACC. The Defense ranks are in terms of how many points, compared to an average team, an offense would score against them. So a negative number implies an offense would do worse than average.
|North Carolina State||6||12.4||8.3||-4.2|
Yes, I see it too. I'm still tinkering with the model but gosh I have no idea how NCST is the 6th best team in the country. Their schedule has been pretty light to start the season and they have played pretty damn well against those teams. I don't think my model does a good enough job early in the season of adjusting for opponent yet, so hopefully NCST will come back down once they start playing better teams and the teams they have beat so far continue to get blown out.
Conference Games this Week
The following shows a boxplot of predicted scores for each conference game this week. A boxplot shows the 25th percentile (lower quartile), median, and the 75th percentile (upper quartile) of the home team margin based on 1000 simulations of the game predictions from my model. That's a lot of fancy jargon for saying here is a visual spread of game scores for this week, along with the home team win probability:
- I'm ok with most of these. I think FSU should probably be a much larger favorite. My model gives home teams a larger Home Field Advantage than I have seen in other models, so I think this is Wake just getting too large a boost. NC State is a huge favorite hosting Louisville, but it's not like Louisville has looked too impressive.