With football season over, I will turn my focus to one of my favorite times of the year: March Madness. I’ll continue to post some football research during the off-season, but the next month or so will be heavy on college hoops.
I had been planning on continuing on the theme of my last couple posts, the difference between “predictive” and “descriptive” measurements. I wanted my first college basketball post to discuss the difference between the “best team” and “teams that have played the best”. Earlier today, John Gasaway at Basketball Prospectus wrote an article advocating the use of scoring margin in determining inclusion and placement in the NCAA Tournament, providing an opportunity for me to debate my point of view.
The “Best” Teams
Say you want to determine who the “best” teams are. Essentially, the question you are asking is: “If these two teams played tomorrow on a neutral court, who would be favored and by how much?” There are numerous methods that can be used to achieve this goal, but the best results have proven to come by using points or other peripheral stats (such as shooting, rebounding, turnovers, etc.) in lieu of wins and losses. Examples of these ratings include Ken Pomeroy (kenpom.com) and Jeff Sagarin’s Predictor (the blue column on this page). Want to know who’s most likely to win the NCAA Tournament? I’d back a predictive rating like these.
The “Most Deserving” Teams
Here’s where I differ. The sole objective in any basketball game is to win. Not to win by as much as possible. Not to look good while you’re doing it. Not to play well but lose a close game. Winning is all that should matter. Over the course of a season, the other factor that comes into play is schedule strength. This is mostly influenced by the strength of your opponents as well as the location of the games (home vs. neutral vs. road). When assessing how well a team has played over a season, the only factors that should come into play are: (1) how often did you win and (2) how difficult was your schedule. Figuring out #1 is easy, while #2 is a little tougher and can be debated. The most difficult part is determining how to combine the two (i.e. how much weight to give to each) and can be debated, but those are the only two factors that should be considered.
Let’s say two teams play an identical schedule and produce identical records, but one team wins all their games by 20 points with their losses all being close, just 2 points per loss. The other team wins all close games, but when they lose they get blown out by 20. Most of us would agree that the former team is “better” than the latter team. If they played each other tomorrow on a neutral court, they’d certainly be favored. However, they are both exactly equal in what they have accomplished, and that is all that you should use to evaluate which teams are the most deserving.
In professional sports leagues–such as the NFL, the NBA, and MLB–all teams play nearly equal schedules (at least close enough in the leagues’ mind to not necessitate being accounted for), eliminating the need to factor in schedule strength in determining playoff teams. These leagues correctly reward teams for what they accomplished on the field of play, and decide solely based on record who advances and who does not. There is no determination of the best teams based on points or efficiency statistics. Take the San Diego Chargers as an example, the #1 ranked team in Brian Burke’s final regular season ratings. If we applied Gasaway’s criteria to the NFL this season, the Chargers would have been in the playoffs while somebody like Kansas City or Atlanta would have been watching from home. (NOTE: I used Burke’s ratings as an example. You may disagree that that is the best way to rank the “best” teams, but that is not my point, I’m simply pointing out that the “best” teams list will differ from the teams with the best records.)
Back to the NCAA Tournament
What this means for the NCAA Tournament is that we need to assess the teams that have played the best, regardless of their scoring margin. To illustrate my point, I’ll use Jeff Sagarin’s ratings, which are well-known, public, and well-respected. Conveniently, Sagarin provides two ratings. His “Elo Chess” rating is essentially what I’ve been referring to as the “most deserving” teams. It only takes into account wins and losses, plus schedule strength. Conversely, the “Predictor” rating is the ranking of which teams are the “best”, and factors in scoring margin.
Washington is a fantastic example this year. Despite rating 7th overall in the Predictor, they are just 18-7 on the season and rank 49th in the Elo Chess rating. This would put them squarely on the proverbial “bubble” to get into the NCAA Tournament (Joe Lundardi’s Bracketology essentially splits the difference and puts the Huskies as a 7-seed). According to Gasaway, they should be a 2 or a 3 seed, based on their impressive scoring margin. I disagree (obviously…have you been reading this whole post?) and think that it is unfair for Washington to be evaluated based on that. They lost 7 games and deserve (right now) to be a double digit seed in the tournament. Would I pick a double-digit Washington team in my bracket? Absolutely! But that’s a completely different question.
There are other teams that fall into this category. Maryland is 75th in Elo Chess, 19th in Predictor. Sorry Terrapins, but you’re out. Alabama is 85th in the Elo Chess, but 41st in the Predictor, and should be out of the bubble discussion. Kentucky’s 21st rankng in Elo Chess says they should be looking at a seed around 5 and not a 2-3 seed that their #9 Predictor rating would indicate. On the flip side, you have a team like Old Dominion, who should be in with a ranking of 30 in Elo Chess despite being 64th in the Predictor.
Conclusion
When it comes down to it, the point of a basketball game is to win. It’s fantastic that we keep score and statistics on the games. It allows us to understand the game better and answer questions like “Which team is the best?”. But when it comes to rewarding teams for their season, it needs to be based on what they accomplished on the court. Did you win or lose? How difficult was your schedule? Use the answers to those questions to select and seed the NCAA Tournament field. Keep scoring margin out of it until after Selection Sunday, then unleash it when filling out your bracket.
Spelling error:assess not asses, get better
Thank you, Andrew. Fixed.
Nice piece, Monto. I tend to agree with you on most things here, except I would rather see a team that lost eight games all season each by 3 points or less make the tourney instead of a team that lost only six games but all were blowouts, since the former has proven they can play anyone close and promises an exciting nailbiter in the first round. So, in that case, I think scoring margin should be part of the discussion. Also, I was surprised that you didn’t mention RPI once. Based on what you’re saying, I would guess that RPI is the perfect rating in your “most deserving” argument since it pretty much only accounts for a team’s W-L record, it’s strength of schedule, and it’s opponents’ SOS. So, why no RPI love??
1) That’s my point exactly. In most cases the team that lost 8 close games is “better” and is likely to be more competitive in the tournament, but the team that lost 6 blowouts but won more games deserves to be in. While I would disagree with it, some may prefer this. However, you’d have to be consistent, and that means seeding a team like Washington near a 3-seed despite their 7 losses and 3rd place standing in the Pac-10.
2) The RPI is certainly one form of this “most deserving” ranking. However, I see some major flaws. First, the weighting is odd: 25% to your team’s winning percentage, 50% to your opponents’ win% and 25% to your opponents’ opponents’ win%. That’s 75% to the strength of schedule portion. Second, their handling of the other portion of schedule strength–game location–is odd. They weight home losses and road wins higher and home wins and road losses lower. They are trying to account for it, but doing it in an awkward way that produces bizarre results in some cases.
I realize this is coming a couple weeks too late, but oh well…
I agree with you Monte that the selection and seeding should be mainly based on “fairness” exactly how you describe. We should REWARD teams, not just let the “good” ones in and keep the “bad” ones out. However, you DO have to think about being fair to EVERYONE. Perhaps Kentucky and Washington should have gotten worse seeds. But then, how is that fair to the unfortunate teams that have to play them? If you are going to say, “congratulations on all the close wins, Kansas State, here’s your #5 seed”, aren’t you actually just arbitrarily rewarding Utah State? Heck, the tournament would look the EXACT same if KSU was #12 and Utah State was #5 (as Kenpom would prefer)!
One solution would be to seed teams based on wins/losses like you advocate, but rather than creating an s-curve, create matchups so that no team is being ARBITRARILY rewarded or punished. This means the “over-seeded” teams play mainly each other while the “under-seeded” teams play mainly each other. 5 seed Kentucky would play 12 seed Clemson, for example, while 5 seed Vanderbilt would play 12 seed Memphis.
If you wanted to make it humorously obvious, you could advertise the two sides of the bracket as the over-rated side and the win-big-lose-close side. The teams in the over-rated side are all rewarded for their good records by having an easier path to the championship.
In the end, though, the teams seeded 5-12 are all close enough in ability so that upsets will happen at an incredible rate no matter how you seed them. So we’re just splitting hairs – it’s not like the tournament needs to be fixed!
Adam, you are exactly correct. By “rewarding” the teams that do not deserve it with worse seeds, you could end up hurting teams that deservedly received better seeds. This really just comes down to a preference thing. If you choose a reward-based system, you run the risk of “underseeded teams” (based on their actual team strength) creating a more unbalanced bracket for the deserving teams. If you choose to do a predictive-based system (i.e. how good the teams are), then you end up burying teams that deserve a better reward with poor seeds. You simply have to pick your poison.