Fact Check: CBS Claims Badgers Scoring 1.28 PPP

Fact Check: CBS Claims Badgers Scoring 1.28 PPP

The growing usage of KenPom.com by fans and media can be a great thing. However, quite often the commentary from folks regarding KenPom’s predictive system is flawed and illogical.

Just another example is Matt Norlander of CBSsports.com, who this week wrote, “The Badgers have an offensive rating of 127.5, meaning they are dousing foes at a blazing rate. UW’s scoring nearly 128 points per 100 possessions. Unheard of at the college level… hat UW is flirting with 1.3 PPP on a per-game basis is flatly freaky.”

The truth is that Wisconsin’s offense isn’t producing anywhere near 1.3 points per possession.  Let’s take a further look…

KenPom’s ranking system is predictive – that is, the adjusted offensive efficiency is not a figure that represents what has happened. It’s a figure used in a system designed to predict the future.

Wisconsin’s adjusted OE sits at 127.5 as we head into the Final Four. However, Wisconsin’s OE is only 1.216 (“only” being relative to adjusted OE – their OE by itself is still sensational) on an actual results basis that includes equal weighting for all minutes played. KenPom.com will show a 1.215 OE due to averaging games and effectively removing the true weighting impact of overtime games.

So, whether we use 1.216 or 1.215, it’s clear that Wisconsin’s offensive isn’t “scoring nearly 128 per 100 possessions.”

To go just a bit further, beware those who say KenPom’s adjusted efficiency numbers simply “adjust for the opponent” as this is untrue. There are other factors, including location of game, recency and what is effectively a cap on margin of victory. The margin of victory cap, by its nature, favors lower-possession teams such as Wisconsin.

If we look at the difference between adjusted offensive efficiency and offensive efficiency for all 351 teams for 2014-15, Wisconsin gets the largest boost of all teams from KenPom’s predictive system on a points per possession basis (and is near the top on a % increase basis).

Across all teams, the average difference between adjusted OE and OE is essentially zero (+0.2 rounding), with just over half the teams receiving a positive adjustment.

Top 5 positive adjustments in 2014-15

TeamName OE AdjOE Diff
Wisconsin 121.484 127.4974 6.0134
Oklahoma St. 103.2472 108.7224 5.4752
Drake 98.9403 103.9572 5.0169
Iowa St. 111.4001 116.3609 4.9608
Auburn 98.2002 103.1423 4.9421

Wisconsin vs. Kansas Example
Kansas has an adjusted OE of 110.6 vs. an OE of 106.5, a difference of 4.1 or 3.9%.

Wisconsin has an adjusted OE of 127.5 vs. an of 121.5, a difference of 6.0 or 4.9%.

Did Kansas play weaker defenses than Wisconsin? KenPom’s adjusted DE for Wisconsin’s opponents is 98.2. That’s good for the 10th toughest schedule from an adjusted DE perspective.

Kansas, though, played even tougher defenses per KenPom. The 96.8 adj DE of their opponents was the most difficult in the nation.

So, why the bigger boost to Wisconsin?

LNH calculated and averaged individual game results, adjusting for each opponents’ adj DE and for game location (using a 1.4% adjustment). In addition, we further penalized KU by treating all five of their “semi-home” games as home games.

The result? KU’s offense receives a positive adjustment of 5.7, while Wisconsin is only adjusted by 5.0.

Again, we go back to the other factors that go into KenPom’s predictive calculations.

Is CBSsports.com correct in saying that Wisconsin is flirting with 1.3 ppp? No; not even close.

Are people who say Wisconsin is scoring 1.28 ppp when you adjust for the opponent correct? No, they are not.

Wisconsin’s 1.318 ppp against Arizona in the Elite Eight was only their 11th best offensive game of the year. However, when adjusting for the opponent and location, it was second only to a 1.522 dissection of Iowa in which the Badgers had only 1 turnover in 54 possessions.

That said, Wisconsin’s adj OE heading into tonight’s game against Kentucky has benefited from the recency of the Arizona game. Is that a good thing for predictive purposes? Perhaps not.

The Badgers’ offense rebounded poorly (17.4%), turned the ball over more than usual (14.1%), and shot far less than average  on their 2-point attempts (48.1% vs. a season average of 55.2%) against Arizona.

Frequent trips to the line – including 10 over the last 4:13 of the game – helped. More than anything, however, was Wisconsin’s 12/18 performance from 3-point range. Despite a lackluster 2FG%, Wisconsin’s eFG% was an excellent 68.9% – their second best this season.

Is it wise to bump up predictions of Wisconsin’s offense against Kentucky because the Badgers had a little luck and/or were just “in the zone” against Arizona? We’d argue “probably not,” but with a static predictive system you’re stuck.

This doesn’t mean the predictive system is bad or unhelpful – rather, it means you should understand what has really happened and how the system processes those actual inputs. When you don’t, you’ll believe things that aren’t near the truth like what Mr. Norlander reported.

Technically unrelated, but a prediction for tonight: unless Wisconsin is superb from long-range again, we like Kentucky in rather convincing fashion. Duke, as it has since before the season began, remains our pick to win it all.

Sharing Options:

Leave a Reply

Your email address will not be published.