Should the committee consider only wins and losses or should other metrics come into play?

Submitted by Yeoman on March 10th, 2014 at 3:37 PM

I don't think it's directly Michigan-related at this point, but it's not as OT as True Detective or Cosmos so I'll throw it out there anyway.

Curious about what people think about this:

Most of what I see in bracketology commentary, here and elsewhere, rests on a tacit assumption that a win is a win and a loss is a loss, regardless of score. Thump a team by 30 or beat them on a last-second banked-in three, it's all the same to your RPI, your RPI/SOS (and everyone else's), your record against whatever class of teams you want to consider.

But to me there's a big difference between firmly establishing your superiority over a team vs. winning a coinflip at the buzzer.

To make this concrete, two teams each with ten losses playing very comparable schedules. Team A played the #97 SOS at kenpom; team B the #95 SOS.

Team A lost games by 1, 2, 2 (OT), 2, 3, 4, 4 (OT), 4 (OT), 9 and 14. The 14-point loss was at the kenpom #29, the 9-point loss was at the kenpom #1

Team B lost games by 1, 2 (OT), 4, 5 (OT), 7, 8, 10, 14, 16 and 25. The 25-point loss was at the kenpom #90, the 16-point loss was at the kenpom #129, the 14-point loss was at the kenpom #67. And the 25-point loss was their last game of the season--it's not like they're heading into the tournament with a full head of steam.

Team B is in Lunardi's most recent bracket. Team A has never appeared even in his last eight out. (Yes, I know, Lunardi. It's similar across the matrix.)

Why? To me, team A has only been soundly beaten twice, by very good teams. Other than that they've just lost more than their share of coinflips. Team B's gotten completely dominated several times, sometimes by bad teams.

Shouldn't that matter? At this point team B has firmly established that they aren't a top-25 caliber team. With team A I'm not sure--maybe they aren't, maybe they are and they've been unlucky.

Am I swimming against the herd here? I understand that the committee might not want to make a lot of noise about point spreads because they don't want to encourage coaches/teams to run up the score. But should they ignore them altogether?

(If you're still wondering or want to fact-check, team A is Utah and team B is Arkansas.)



March 10th, 2014 at 3:46 PM ^

But you are only looking at losses. Were most of Team A's wins pretty much coin flips against mediocre teams? Any big wins for Team B?


March 10th, 2014 at 4:01 PM ^

They've each won four games against teams in contention:

Utah beat UCLA by 5, Arizona St. by 23, Colorado by 9 and Cal by 4.

Arkansas beat Kentucky twice, both times in OT, and beat Minnesota by 14 and SMU by 11.

Computer rankings:

  • Kenpom: Utah 33, Arkansas 55
  • Sagarin : Utah 33, Arkansas 40
  • Massey: Utah 48, Arkansas 59
  • RPI: Arkansas 62, Utah 83

That's consistent across the ranking matrix. Systems that take scores into account prefer Utah; systems like RPI that ignore scores altogether prefer Arkansas. The bracketologists unanimously side with the systems like RPI.

I'm not so much arguing that Utah is better (I think they are, to be honest, but it's not what I'm arguing here) as: why are they incomparably worse? The only answers I can come up with inolve RPI and its clones.

Zone Left

March 10th, 2014 at 3:50 PM ^

The BCS proudly acknowledged it ignored margin of victory.

Margins can and should be considered. A win by 30 is more impressive than a win by 3.

I Like Burgers

March 10th, 2014 at 4:03 PM ^

This is what will make the football playoff committee interesting.  Since its a much, much smaller pool of results and matchups to look at, things like margain of victory will have to be looked at.  And not just that.  Say two teams beat Team X 42-21.  But the first team pulled their starters midway through the second quarter, but the second team left theirs in until the start of the fourth.  Team A's victory becomes more impressive.  

But what if team A was playing at the start of the season, and Team X had some players that really came on during the season and became major players during the latter part of the season when they played Team B.  Is Team A's win still more impressive since they were possibly playing a weaker version of Team X?

The football committee debates will be really interesting.  Football is a much harder sport to try and pick a playoff from using a selection committee.


March 10th, 2014 at 3:54 PM ^

Games are always skewed by free throws at the end though. A 3-5 point game frequently turns into a 10 or even more game in the last 30-40 seconds of the game alone. There's no way to account for this that I can think of without ignoring free throws in the final minute, but that's just stupid.


March 10th, 2014 at 4:19 PM ^

True, of course.

If we want to get to that level of detail, Arkansas was down 20 to Alabama with six minutes to go in the first half and never got closer than 18 thereafter. They were never close enough to start fouling late against A&M, either. LSU turned an 11-point game into a 14-point game with late free throws.

In general I think this sort of thing tends to even out over the course of a season and it's not a big effect on the computers to begin with. Hopefully Burgers is right and the committee really does look at those details when they're sitting.

I Like Burgers

March 10th, 2014 at 3:57 PM ^

I actually spent the weekend talking to a committee member.  Super nice guy.  His big point of emphasis was that it really comes down to who you play, who you beat, and where you played them. Conferences and all of that other crap don't matter.  Its all on a team by team basis.  

So to answer your question, yes things like this do matter.  If it comes down to it (and it might since they are 21-10/10-8 and 20-10/9-9), they will put up Utah's resume next to Arkansas's resume and compare the two.  They will ask, OK, why did Arkansas get beat so badly in those games?  Or why/how did Utah play their top teams close?  And then they'd probably throw in a third team like say Iowa (also in that 20-11/9-9 range) or one of the three 20-11/10-8 Big East teams and compare their results.  And then they'd vote on which of those teams they will continue to consider, and which ones get left behind.

The committee memebers watch a TON of basketball, have a ton of data at their disposal, and are all very well informed.  Each member is assigned a main conference (like ACC, SEC, Big Ten, etc) and then a couple of smaller conferences to be the "expert" on.  So when "hey, how did Utah play these games close" gets asked, they'll turn to the person that's been monitoring the Pac-12 for some insight.


March 10th, 2014 at 7:11 PM ^

I know that's true to some extent, but it's also true that the bracketology community does a pretty good job of predicting who's going to get in and in doing so they seem to primarily use metrics that don't include scores and other performance stuff. That seems to me to hint that the committee is doing the same, that they may realize the RPI-based metrics aren't all that good but they tend to use them anyway until they get down to those last few teams.

I Like Burgers

March 10th, 2014 at 9:11 PM ^

The guy I was talking to mentioned this.  He was saying that its pretty easy to nail down around 63 or so of the 68 teams in the tournament.  The last 4 or 5 are the really hard ones that take the most time.

Most of the information and tools that the committee uses is also available to the public.  So your Joe Lunardi and Jerry Palm's of the world are using a lot of the same tools.

Ghost of BCook…

March 10th, 2014 at 3:57 PM ^

Jersey color and mascot toughness should come into play. Millions of uninformed people win their office brackets on these factors alone, it's about time the committee took note.

Ghost of BCook…

March 10th, 2014 at 6:09 PM ^

That should be factored in as well.  Using your example:

Mascot:  A single hurricane can wipe out an entire square mile of orange trees.  Edge:  Miami, by a ton. (Miami +15 points)

Team colors:  Syracuse - Orange v. Miami - Orange/Green.  Edge:  Slight nod to Syracuse, by  virtue of not incorporating that nasty green. (Syracuse + 5 points)

City:  Miami, sun, fun, cuban culture v. Syracuse, cold, snow, and Mike Tirico.  Edge:  Moderate advantage to Miami. (Miami +8 points)

Based upon this careful and very qualitative research, if the two teams play, I would expect YTM by 18.  Ken Pomeroy looks like he's guessing compared to this method. 



March 10th, 2014 at 3:58 PM ^

I agree as well. A win is not just a win and a loss is not just a loss - there is context, and margin and the opponent are definitely part of that. I think some people talk about "Oh, Team X lost to Team Y, but they were a TOp 25 KenPom team and it was only by 3", for example, but then come to a place like this and forget to mention the third dimension of the game result. It should definitely be part of the discussion. 


March 10th, 2014 at 4:00 PM ^

Let's not even go there. The BCS was a joke. Of course the human voters were quite obviously influenced by margins of victory, but the computer programs weren't allowed to take it into consideration. Which, of course, handicapped the computer programs to a degree that they couldn't be expected to produce decent results. That is, if the computer programs actually followed the directives. The programs were not even required to publish their formulas, so there is no way to know what they actually did, or to attempt to reproduce the results. It was an appaling nightmare.

In general, yes I believe easy wins vs. close wins is a factor, but it's more important in football, where the sample size is much smaller, and where you are only choosing two or four teams to allow into the tournament. Here, with larger samples and a cutoff line that allows nearly 70 teams in, it doesn't matter so much.

I'm inclined to just consider wins and losses. Advanced metrics are fine but at the end of the day, the point is just to win the game. All of these more sophisticated methods introduce assumptions which are difficult to validate. Conference titles, tournament titles, everything is handed out on the basis of wins and losses -- nothing else. The degree to which these advanced metrics are better predictors of future wins and losses isn't clear to me. The team that earns the national championship is the team that gets six wins, not the team that leads the tournament in efficiency.

edit: this was supposed to be a reply to the guy who brought up the BCS. I'm new-ish around here, OK?!


March 10th, 2014 at 7:07 PM ^

Sagarin and Massey do two football ratings, one without margin of victory that they provide to the BCS, a second that includes margin.

In both cases the version that includes margin is better for predictive purposes. By a lot. They've both stated quite clearly that the only reason they bother with the scoreless version is that the BCS asks for it.

Naked Bootlegger

March 10th, 2014 at 4:10 PM ^

I thought it was standard practice to misspell Loonardi's name as much as humanly possible, since, I guess, a lot of people apparently don't like Lonbardi's prognostications.  I, myself, don't follow Ludnardie very much, so I'm not really familiar with his schtick. 


March 10th, 2014 at 5:50 PM ^

That guy has the best job in the world.  You only do it for three months out of the year, and nobody cares about it until the last month.  And you don't even have to be right.

When you are getting up for work early on the cold, dark Monday after Selection Sunday, just remember . . . Joe Lunardi is done for the year.



March 10th, 2014 at 4:19 PM ^

one reason Utah might not be in the discussion is their non-conference slate was awful.  Looking at the RPI rankings on espn, Utah's non-conference SOS is ranked 347th.  The Hog's non-conference SOS isn't anything great either, 201, but 347 is just awful.  There are 349 teams.


March 10th, 2014 at 4:35 PM ^

Non-conference SOS was introduced into the conversation to defend lesser-conference teams like Gonzaga against the charge that they hadn't played anybody. They don't have any choice about who they play in conference, after all, so why not give them credit for scheduling good teams when they have the chance?

It's now being used as a defense of an SEC team.

And rightfully so. They don't have any choice but to play a terrible conference schedule.


March 10th, 2014 at 5:40 PM ^

Yes.  Our 2 point last-second loss against Arizona is worth at least as much as our win against Stanford.  

You can have good losses (and bad wins) and it should be taken into account for seeding. 


March 10th, 2014 at 6:25 PM ^

"Good teams win the close games." Bill James, the baseball statistician, investigated that statement back in the '70's when I was reading each yearly Almanac that he published. IIRC he crunched a bit and found no pattern in one-run games -- they could go either way -- but that good teams won the 14 - 1 laughers.


March 10th, 2014 at 7:52 PM ^

Michigan is eighth in the AP poll, ninth in RPI and a No. 2 seed in Joe Lunardi’s Bracketology. In BPI, however, Michigan is 22nd. Of the Wolverines’ seven losses, four have been by at least 10 points; of their wins, seven have been by five points or fewer. Also, Michigan is 8-1 with an 88.7 BPI against opponents missing at least one of their top five players (in terms of minutes per game), and BPI de-weights those games.