"My rule was I wouldn't recruit a kid if he had grass in front of his house.
That's not my world. My world was a cracked sidewalk." —Al McGuire

Monday, December 01, 2008

Dayton Numbers Recap

Just wanted to break down the Dayton game a bit by numbers before we move on to Central Michigan and then the hated Badgers. After all, it is rivalry week.

No surprise, but Dayton won on almost every aspect of the Four Factors. Even the one area where Marquette enjoyed an advantage was negligible. The numbers tell us that this game was lost on the defensive end. Marquette's offensive efficiency was definitely worse than their season average (1.15 points per possession), but it was the inability to stop Dayton defensively that was the real killer.

I'd like to look at two primary culprits, eFG% and TO Rate.

If Marquette averaged a defensive eFG% of 56.1%, that would be #325 in the nation. If we averaged a defensive Turnover Rate of 14.6%, that would be #340. From my own modeling efforts, both of those results ended up outside the expected 90% confidence interval.

I calculated that Dayton should have gotten an eFG% somewhere between 23% and 55%, with a most likely view right around 42%. After all, Dayton's season eFG% is 46.0.

TO Rate

TO Rate tells a similar story. Dayton coughs up the ball 20.2% of the time and Marquette forces a lot of turnovers (well, they usually do). Dayton should have had a turnover rate somewhere between 21% and 35%. For Dayton to protect the ball at 14.6% is another result outside the range of expectations.

So was it just one bad game, or was it a result of Marquette being exposed? It's way too early to tell, but if I was forced to make a judgement, I'd lean towards one bad game. If these teams played again, I'd be shocked if Dayton was in the top 5% of expected eFG% and Turnover Rate.

At least we'll get another opportunity to see how Marquette stacks up against a good team on Saturday. Plus, there's a game tomorrow! Let's move on.


rugbydrummer said...

Hey, Henry Sugar, can we expect to see the P values to show statistical significance? Are you calculating standard deviations and T/Z-scores and all that?? Where are the confidence intervals?

I want to start seeing Relative Risks and Odds Ratios for our pregame predictions, if that's possible, ha ha.

Thanks for the stats-fix!

Rob Lowe (aka Henry Sugar) said...

Someone's taken a few stats classes... probably more than me and I'd wager probably associated with med school.

It's unlikely that I'd share the p-values from the regression analyses that I run. This isn't an academic site, and for many fans, the numbers I do crunch are too many.

However, I do calculate standard deviations, and all my projections are based on a 90% confidence interval. The two ranges listed above in the Dayton recap are both based on a confidence interval of 90%.

Thanks for the feedback. We'll try to balance your requests with the need to keep the site interesting to a broad range of readers.