It’s a statement you might hear at a support group meeting, except there may be no one else there.
I’m a college basketball fan, and I don’t hate the RPI.
There, we said it. And now we’ll duck and let the virtual tomato throwing commence.
If that above sentence were given a sub-title, it just might be “How to commit Twitter suicide in one easy step.” For, of all the lock statements in the world, few are more of a sure thing than this: if one wants to talk college basketball and be popular on social media, they do not speak in any sort of positive terms about the Ratings Percentage Index, or RPI, that ratings formula used for grouping data by the NCAA Men’s Basketball Committee, a.k.a. selection committee.
Do not. Under any circumstances. For any reason. Ever.
Ah well, so it goes. Never really liked Twitter much anyway. (No offense, Twitter friends)
Unlike seemingly 98% of college hoops followers on social media, the RPI has never bothered me much. Not in 30 years following the sport, not in 20 years hearing about it regularly as part of the NCAA tourney selection process, and not in recent years as so-called “advanced” metrics came into vogue.
Don’t misunderstand: it’s not that I love it. Like anyone else, there have been times when the numbers spit out by the RPI puzzled me. Occasionally even frustrated.
It does get talked about too much. Could it use some tweaking? Almost certainly. (Hello: more credit for wins, less for the conference schedule you play by birthright)
But hate? Anger? No. Never.
Some do hate it, to the point where you’d think those three letters are a dirty word. The R–, if you will. That’s their choice.
Clearly some are in the sport’s major national media, who have led the rebellion against the R– and-we might add-whose rantings in recent years sometimes cross the line of reasoned opinion, reading as if the R– has inflicted emotional harm on them.
We’re talking about a ratings formula here. Not an ex-girlfriend.
The fear here, though, is that while not doing much for discourse, more importantly the hostility for said ratings formula has distracted from some real issues with the selection process, essentially leading a witch hunt this-a-way while actual problems are ignored that-a-way.
The R– as scapegoat for selection errors, though, led to a meeting at NCAA headquarters a couple weeks ago, where-at coaches’ behest-the men’s basketball committee met with several analytics folks with the intention of replacing the R– with another metric. Perhaps a composite metric that could incorporate any number of ratings formulas, data points, and such.
What this meeting ignored, though, is that as much as the RPI (are we OK using it by name again yet?) gets hammered on, if comparing rankings when they count-late in the season, not in November and December, when schedules haven’t even come close to balancing out-the never-intended-to-be-predictive RPI is still little guiltier of questionable team placements than any other rating.
Sorry, but no matter which formula one thinks is superior, I can furnish examples in a minute to show its weaknesses, too. (In fact, CBSSports.com ironically did the service of displaying exactly that a few weekends ago.)
They’re called outliers, and every rating has them. Might one formula shake out as slightly better than others by March? Perhaps. But there’s not remotely enough of a difference in them to raise this person’s blood pressure.
Of course, the elephant in the room is that the RPI does not incorporate margin of victory. Directly or indirectly, for that it has been called outdated, antiquated, you name it.
Instead of auto-assuming this makes it wrong, though, let’s back up the truck. We would ask the question: why should it?
Perhaps more than any other team sport, the margin of victory in college basketball games doesn’t always represent how a game played out. It’s a fairly regular occurrence that a two-point game with a minute left can end up a six, seven, even 10-point contest with teams intentionally fouling to get the ball back. Or a 25-point game can get knocked down to 10-12 at the end when the coach in the lead lets off the gas.
Rewarding margin of victory still might be OK if it didn’t also mean rewarding close losses, or penalizing teams that win a lot but don’t always win by much. A perfect example last year was St. Bonaventure, a team that won 22 games, including a share of the Atlantic 10 title, yet was hurt badly in non-RPI metrics because a number of its wins were close. The Bonnies were essentially kept out of the NCAAs because they ‘only’ won, and didn’t win big.
Sorry, but that stinks.
Let’s also be honest about this: the use of different metrics in recent years hasn’t done anything to address inconsistent logic in the selection process. It might’ve made the inconsistencies worse.
Going back to St. Bonaventure again: last year the Bonnies were 30th in the RPI, performed well against top 50 and 100 teams, yet were left out, likely because they weren’t nearly as high in the Ken Pomeroy rankings, the chic rating of current times. Meanwhile, Wichita State was a sky-high 12th in the Pomeroy ratings but just 50th in the RPI, with a resume resting almost solely on its Pomeroy rank and missing star guard Fred VanVleet for several key games early in the season.
The Shockers made the field, while the Bonnies were left out. So, the committee was proving more enlightened in its use of metrics, right? Wrong.
The Shockers got into the field-barely, their Pomeroy ranking almost matching their seeding for a play-in game in Dayton. Meanwhile, Saint Mary’s also was rated high in the Pomeroy ratings (and the RPI, for that matter), yet the committee was more impressed by a Michigan team 20 spots below them in the Pomeroy rank (and with a hideous resume) and a Tulsa 30 rungs lower in that ranking than the Gaels.
Just ahead of Tulsa in the Pomeroy rankings was Oregon State, a team the committee not only embraced with its poor non-RPI metric, it handed a 7 seed. And yet, the committee loved Vanderbilt’s high Pomeroy rating, overlooking the Commodores’ much more accurate RPI in the 60s that, evidenced by a 2-7 mark vs. the RPI top 50 in the regular season, foreshadowed their quick exit from the NCAAs.
So, which was it? The committee relied too much on the RPI? St. Bonaventure (among others) showed clearly otherwise. The committee relied too much on KenPom rankings? Ask Wichita State or Saint Mary’s if that was the case.
This isn’t to say complex ratings aren’t enjoyable or don’t have a place. They are and they do, and have for a long time. Computer ratings are hardly a new thing, and it was always a treat as a kid in the 80s and 90s to pick up the USA Today to read Jeff Sagarin’s ratings. Then and now, they’re useful as additional information when trying to differentiate two teams, or when trying to get a read on some of those outliers.
Come NCAA tourney selection time, though, I don’t want wins devalued because they aren’t by enough points. I certainly don’t want a rating telling me a loss is practically the same as a win. And I don’t want ratings telling me one team’s 20-point win is better than another’s 10-point margin. If it is, human beings on a selection committee are intelligent enough to do a little research and evaluate that.
Ultimately, that’s what this comes down to from here: the human element is what I prefer about NCAA Tournament selection. And more than any rating, the RPI minimalizes human influence in the numbers, allowing human beings-imperfect as they are-to interpret data and make decisions.
Wins are wins. Losses are losses. People can then look into the how and why of the results, which lets them then essentially invoke an educated eye test.
As much as the RPI is vilified, this is a good thing. I don’t want the tournament picked by computers. Nor do I want humans inundated with a composite formula designed by humans. We’ve seen before how that ends. It’s not good.
For those who like the composite route, here’s a question: remember this thing called the BCS? It certainly shouldn’t be so hard to recall because it wasn’t long ago. College football’s Bowl Championship Series was a yearly fiasco, and every year its organizers spent the offseason tweaking the formula to fix the results to reflect what people complained they should be. And people still complained-every year.
Is that really what college basketball needs? The BCS?
(The other part of this: for those who think computer formulas are harder to “game” than the RPI supposedly is, football teams figured out right away that running up scores helped them in the BCS computers. You better believe college hoops teams will figure out how to run up scores and efficiency ratings to impress computers-if they haven’t already.)
If (when?) the NCAA goes to a composite rating next year, this is almost certainly what we have to look forward to. Yuck.
This is why I don’t sweat the RPI (much). There will always be complaints about which teams are in the field and which aren’t. And no ratings formula will ever pick a field that everyone is happy with.
All of this isn’t to say the selection process can’t be improved. It most certainly can be. But from here, those improvements have infinitely more to do with how humans are interpreting numbers than with the numbers they receive being flimsy.
Those improvements in many cases can be summed up in one word: context.
We can start with how in recent years, a team’s quantity of top 50 wins has been elevated to ridiculous levels of importance to the selection committee, while it completely ignores the actual record in such games. It’s to the point where the committee seems to think 3-7 against the top 50 is better than 2-2.
Seems here that all a team going 3-7 against the top 50 proves is that, when it plays a reasonable sample size of NCAA at-large caliber competition (which is essentially what the top 50 signifies), it will lose 70% of the time. Such teams regularly flame out early in the NCAA tourney. How a 30% winning mark against top 50 teams is considered an asset, we’ll never understand.
The other side of that is, a 1-1 mark against the top 50 isn’t proof that a team can’t beat top 50 teams regularly. That should be the most basic common sense, so obvious it shouldn’t even have to be said, yet somehow in this sport it’s been accepted as proving just that.
Besides that, there’s plenty of information available-whether talking statistically or looking inside results-to tell us a team just might’ve gone at least 3-7 vs. the top 50 if they had the chance to. That’s especially the case if it can go 6-2 against the top 100-something Saint Mary’s did a year ago.
Committee members must go deeper than just citing big wins and ignoring losses in similar games. Teams’ entire resumes need to be looked at and judged accordingly. Games against teams ranked 51-100, even 150, still can be decent wins. Road game performance matters. The committee must acknowledge that every game counts-not just games against a select group of teams.
The need for context also goes with gaining proper perspective of losses. That includes those to teams outside the 200, something that has been elevated into a death knell to teams’ at-large hopes, when often the committee has clearly not taken the first bit of effort to understand just how common such losses by good teams can be-when they play them.
Such was the case last year with Monmouth, which did absolutely everything asked of it in scheduling-taking on a host of big names and beating many of them-and played a ridiculous 23 of 34 games away from home, yet was penalized for three sub-200 losses-again, all on the road.
Those were the type of road games a Monmouth in the MAAC can’t avoid, but that teams from BCS conferences lost at about the same rate the Hawks did the few times they actually played them. Again: it shouldn’t be so much to ask the committee to understand this.
Whether you liked the Bench Mob or not (wasn’t our flavor), the lack of nuance in evaluating and the seeming intent to find a reason to keep Monmouth out was disturbing. As evidenced by last year’s use of RPI and Ken Pomeroy rankings, too often the committee uses whatever numbers it wants to support the teams it wants to pick, and discards numbers or finds convenience samples to support excluding those it wants to.
Is it brand name bias? (It’s hard to think otherwise when teams like UCLA in 2015 and Michigan a year ago with hideous records vs. the top 100 still get in.) Is it lack of education of what some numbers mean? Perhaps. But then it’s time to get educated. The committee can do better, and should.
Of course, the Hawks might’ve also been viewed more favorably if the committee went back to something it used to do and put more value on regular season conference championships. As we noted last year, the reduction in regular season champions in the field is a disturbing trend in selections going back 20 years, and it’s not a good one.
When in doubt-and we’re typically talking a couple bids a year, tops-the committee should be giving preference to achievement and excellence. Again: wins against teams even outside the top 100-especially on the road-can still be good wins. Even in a mid-level conference like the Colonial or Mid-American, it’s not nearly as easy to go 16-2 or 15-3 as many seem to think.
Rewarding these wins also is the very best way for the committee to combat the scheduling imbalances that are a known fact in college basketball, where leagues with the gold can literally buy NCAA bids by spending almost the entire non-conference season at home and actively avoiding the best teams from leagues outside their economic realm. It’s really quite simple: if the committee didn’t reward it, they wouldn’t do it.
It should be noted: it wasn’t RPI that told the committee to keep Monmouth out of the tourney last year. In fact, the RPI had Monmouth ranked much more favorably than margin of victory ratings. Nor was it the RPI’s fault that conference champions from leagues outside the top 8-9 leagues have become increasingly forgotten in recent years.
Regardless, it’s not so much to ask the committee to understand context in all of the information available on teams, and then show reasonable consistency in its messages. The written criteria for selection isn’t nearly as complicated as the NCAA likes you to think. If it were, you wouldn’t see dozens of at-home bracketologists who every year generally agree on what the field will look like, with only a few exceptions.
It’s also not too much to ask that the committee evaluate every resume with an understanding of the inherent advantages and disadvantages certain teams and leagues have. It did a wonderful job of just that in 2012 and 2013, fairly evaluating teams like Iona, La Salle, Middle Tennessee State and Saint Mary’s, and those teams performed just fine.
The committee doesn’t have to be perfect. It does have every obligation to aim for consistency but also nuance in evaluating wildly different schedules.
These are realistic goals to work towards, and committee members can do this. They’ve done it before.
Twitter: @HoopvilleAdam
Email: [email protected]
