clock menu more-arrow no yes mobile

Filed under:

The Providence v. Analytics Conundrum

We all know Providence is winning games and yet metrics still seem to not like them. Let’s dive into why and why no one should be mad about it.

COLLEGE BASKETBALL: FEB 01 Providence at St John’s Photo by Gregory Fisher/Icon Sportswire via Getty Images

Look, Friar Faithful.

I want to start with a take that you may not like:

The metrics are not going to change any time soon, and constantly getting mad at what the metrics are saying is not going to help out anyone.

The fact of the matter is, whatever the metrics say, you’re still winning games. You’re still atop the Big East, cruising to top-4 protected seed. You are having arguably the best year since your last Final Four run back in the 1980s. Those are things to be proud of and a massive reason why you should love your squad. I know I am constantly ever more impressed by the grit, the clutch, and the heart the Friars play with, and I want to let you know this is not a column meant as a hit piece. It’s a simple explanation of why the metrics say what they do, how to put them into context, and why at the end of the day the metrics are mostly good at describing what we see but do have some blind spots (which are usually by design).

Alright, so let’s talk public enemy number one: the NET rankings put out by the NCAA.

Providence is currently sitting at 29th in the NET, behind the likes of Marquette (split the season series), Xavier (1-0 against the Musketeers, the win coming in Cincinnati), UConn (ditto X, but the win coming in the XL Center in Hartford), and Villanova (yet to play, two games still on the schedule). I know, what the hell, right? Providence is 5-1 in Quad 1 games, 5-1 in Quad 2 games, and undefeated otherwise. Prima facie, should much better than Marquette (6-5 Q1, 2-1 Q2, 3-1 Q3, undefeated Q4), Xavier (4-5 Q1, 5-0 Q2, undefeated otherwise), etc.

Let’s start breaking that down though. For example, Marquette. Sure, Marquette has 4 more Quad 1 losses than the Friars. However, Marquette has played 5 more Quad 1 games, so the strength of schedule is much higher, and Marquette has the same road record in Q1 games as the Friars, including (according to NET rankings) the much better win in beating #5 NET Villanova in Philly, while the best win by NET rankings Providence has is at #17 NET UConn.

Furthermore, even in losses at home, Marquette has played every Q1 and Q2 loss close, while in their losses, the Friars have been thoroughly, well, whooped. Furthermore, of the 10 Q1 and Q2 wins Providence has, Providence has exactly 2 wins, both Q2, of more than 10 points: one at home against Vermont (70 NET), one away against DePaul (110 NET). Old enemy (for purposes of this article, because they’re the closest Big East team in the NET at 27 to Providence) Marquette? 4 of the same types of wins, including the 30 point beatdown of the Friars (29 NET), neutral court against West Virginia (62 NET), away at Seton Hall (41 NET), home against Xavier (18 NET).

While the NCAA has moved away from using a pure margin of victory as an input into the NET (it used to be a capped margin at 10 points), that margin still matters for efficiency metrics. The margin of victory matters because, while you’re still getting rewarded for notching a win in the win column no matter what, the truly elite teams (as defined purely by the NET and efficiency metrics) are the ones that beat the opponents they’re supposed to beat and by a comfortable margin.

This leads to the second public enemy and a point that ties into the margins and efficiency ratings.

Public enemy number two: KenPom.

Why doesn’t KenPom like Providence, who is ranked 48th by adjusted efficiency margin (basically, the difference between offensive and defensive efficiency)? Well, it goes back to those margins of victory. Yes, Providence is finding ways to win in crunch time. That rocks. Objectively, unequivocally rocks. Wins are better than losses. I know. We can scream and shout all we want about it.

But it matters how you win. For example, let’s go back to the Marquette comparison (Marquette is 30th in KenPom). As stated above, Providence only has 2 wins in Q1 and Q2 of over ten points, both coming in Q2 away at 110 NET DePaul and home against 70 NET Vermont. Marquette has 4, all of which are against better NET-rated teams than the two 10+ point wins Providence has.

Side note: I’m not trying to tell you that Marquette is better than Providence. I’m trying to tell you that this is what the metrics are looking at and why they stand where they do because of them. That’s the entire point of the article.

Digging in further, Providence is a super-balanced team: 47th in offense per KenPom (110.5 adjusted offensive rating), 58th in defense (96.2), good for a (110.5-96.2) +14.3 adjusted efficiency rating. Providence doesn’t boast an uber-elite offense OR defense, but it’s pretty good at both.

Contrast with Marquette. 68th on offense (108.7) - 23rd on defense (96.2) for a +16.5 adjusted rating. Marquette isn’t playing as good of offense as Providence, but it’s playing much better defense over every minute played so far this year. And that matters for the overall efficiency ratings, not just being very good at both come clutch time (I’d kill for a team that was as good in the clutch. I AM a Marquette alum and watching them fight back in the first three Big East losses just to fall apart when it mattered sucked. Ditto watching them winning most of the game in the Dunk just to get out-clutched by the Friars. We can talk efficiency metrics, but the wins and losses columns still do matter, and being clutch enough to get them matters).

And then there’s the often-memed and often-derided Luck statistic. This is how Ken Pomeroy defines it: (emphasis the editor’s)

Luck, which is the deviation in winning percentage between a team’s actual record and their expected record using the correlated gaussian method. The luck factor has nothing to do with the rating calculation, but a team that is very lucky (positive numbers) will tend to be rated lower by my system than their record would suggest.”

Luck has no actual outlook on some subjective idea of how things go in favor of a team. It is simply defined as “what my metrics tell me will happen versus what happens in real life.” As said, it has no bearing on how teams are rated; rather, teams that are lower-rated in efficiency metrics with better records will have a higher luck rating inherently.

Sure, Providence is enjoying a historic run of high luck ratings per KenPom. It’s number 1 this year with a +.234. For context, the second is +.203. That’s a very large gap. The gap last year between 1 and two was .020, and the high number was +.199. In fact, in the KenPom era (since 2002), this Providence team currently has the highest rating ever (the highest ever after a full season being Wagner in 2008 with a +.226). The gap between 1 and 2 isn’t as bad as that Wagner team (a massive .110 between Wagner and second-place Hartford).

It does sorta reflect in the way Providence has won. Providence is 11-0 in single-digit games and 8-0 in games decided by 5 points or less. Again, prima facie, that looks lucky. Every team would kill to be undefeated in those circumstances (coming from a Marquette alum, please Jesus). And most, if not all analytics sites, would call those types of games coin flip results and expect most teams to go, at least in the games decided by 5 or less, 4-4 on average.

And that’s where I was building this article to. By analytics’ measure, Providence IS a lucky team. The way the analytics measure games do not account for strong execution and decision making, good coaching, however you want to classify refereeing (more on that in a paragraph or two, and I am very anti-referee conspiracy, I don’t think they’re just getting calls), and other things that go into winning clutch. The analytics are not designed to measure only these things. From a purely mathematical point of view, the analytics would expect Providence to break even in these games.

And they haven’t. They’ve won. And that’s good!

Sometimes you can’t do anything about the damn analytics measuring what they’re set up to measure. Providence in ‘21-’22 is living in an analytic anomaly where they’re very good at playing close games and closing them out. God, so many schools would kill for that! The analytics are measuring overall performance, beyond just wins and losses. And they’re saying picking up exactly what they should be: Providence is a super-solid team across the board that is playing close games and winning them. The analytics don’t love it, but the win-loss column sure does.

Providence has the makeup of a team built to win ugly, close games. They’re a team that excels at getting to the line, especially late in games, doesn’t foul and give opponents free throws, and is an incredibly veteran-laden roster that has been here before and knows what to do. They do not make mistakes in crunch time, have a tough team of bucket getters and lockdown defenders, a defensive enforcer at the rim, and a coach getting the most out of his guys when it matters most.

For every minute of basketball the Friars have played, they haven’t been the sexiest team in college basketball, and they tend to not always prove how much better they may be than their opponents over the course of literally every minute of basketball they have played, which is what the analytics measure. That doesn’t make them bad, it just means they’re not playing games where they stomp all over their competition. They play close games and close them out every. single. time. (Not to jinx it, but so far. fingers crossed, Providence fans).

I know what I’d rather have come March, damn what the analytics say.

I want to promise the Friar Faithful this: the analytics are not biased against you. They’re telling you how the team plays, on average, throughout every second of every game, as they have defined measuring it. They are not built to quantify a “clutch” gene or “toughness.” So before we all collectively freak out about the analytics not loving the Friars, please remember what they are and are not measuring.

There’s another big discussion about how and where analytics should be used in NET rating and tournament selection, but I assure you: the analytics will not cost you what you think they will.

It’s easy to get caught up in perceived disrespect for a team that is winning games and leading the league that doesn’t seem to be getting what it is due in rankings or national attention or whatever. Yes, AP voters suck. Yes, the NET is imperfect. It’s still a damn sight better than RPI. And I will stand by the use of metrics outside wins and losses in analyzing teams because otherwise, we’d just be back to using RPI. That was a pretty awful system (granted, Providence would be a one seed under the RPI this year, but I’d like to point out 4 seed Iona, 5 seed St. Mary’s, 2 seed LSU over 4 seed Houston, 2 seed Alabama over Arizona/UCLA/Gonzaga/Purdue/etc., a 6 (!) seed Gonzaga and Duke, et al.)

It sucks that Providence seems to be disrespected considering what they’re accomplishing this year. I know. It’s brutal. I just want to promise you that, at the end of the season, all of the disrespect bracketologists and AP voters have given the Friars will not matter to the committee, and that the analytics the committee will use in addition to the NET will not harm you like they may appear to be doing at the moment. As long as Providence keeps winning, and beating the teams coming up (2 dates with Villanova, hopefully a reschedule against UConn, beating Xavier and Creighton), they’ll be respected in the analytics, in the NET, and the bracket.

Stay the course, Friar fans. The respect is coming.