Running 2023-24 NCAA Basketball Thread

What's the problem with a system that actively disincentivizes coaches from developing bench players by playing them in blowouts?

 

I brought it up on a twitter discussion a few weeks ago when some team gained like 25 spots for beating a bad team by 40 and was corrected and told it wasn't score differential it was an efficiency metric. But don't they go hand in hand? I mean, if you are beating someone by 40 you are being pretty efficient on offense or defense or both.

Anyway, I believe it should be capped at 20. I think the first year of it it was capped at 10 and they got rid of it. That might be too low. You should be rewarded for beating a good team by 20. But it is absolutely insane that you can get such a push by doing that to bad teams. There should also be some sort of way to devalue big wins against bad teams.

I am not smart enough to know the big differences between the NET and RPI, but I bet this is a main one, because in the RPI a win was just a win. Of course that changed the SOS across the board and everything else. I do think the NET is a better judge of who is where, but if it is going to be your main tool, tweak it so that it is as accurate as can be. Let the other predictive metrics worry about what a 40-point win means.

 

Weirdly, it kind of does matter, though. A 15-point game coulda been competitive for 2/3 of it. A 25-point game was probably an ass whooping for large portions. Computer models almost always use margin of victory and then adjust for opponent quality because it usually produces better results, whether we're talking college or pro sports.

 

RPI sucks because it basically just rewards you for compiling wins, even if most of them suck. IIRC, the basic formula is that it is half your win percentage, a quarter your opponents' win percentage, and a quarter your opponents' opponents' winning percentage.

I'm on an Atlantic 10 message board, and one older gentleman was pissy because Fordham at one point last year was like Top 25 or 50 in RPI, and as a result, he wanted them getting bubble consideration. By KenPom and Sports Reference, I think their SOS was in the 300s. They still finished 83rd, because they won 25 games, even if they lost to pretty much any team remotely near the bubble (to Arkansas by 40, twice to Dayton, once to VCU).

 

UAB played an amazing first half. 61 points, 8 of 14 from three, 67% from the field. Memphis basically didn't play defense in the first half. First half fouls, UAB 16, Memphis 7.

This team has no poise, no killer instinct when it gets a big lead. They gave up a 7 point run going into the half. They were up 15 and it felt like being up 2, because that's their pattern. Also, this team simply is not a good three point shooting team, they live on points in the paint. I *knew* 3's were not going to continue to fall at that clip.

Memphis came out with intensity and ate them alive in the second. Threes dropped to 1 of 12, shooting percentage to 36%.

The refs were shit, Memphis shot nearly twice as many free throws, but that wasn't what caused the loss.

I trust this UAB team to win a close nip and tuck game late far more than I do that they will build a big lead and hold it, let alone put a foot on the other team's throat and put the game away.

 

But the metrics don't account for if a game was close at a time or not, just what the final score is and how a team overall did in the game. To an extent even more reason to have some sort of cap, whether scoring or efficiency. It could go both ways, a team could be getting beat by 15 most of the game, the other team lays off the gas, and the team that has been losing hits a couple of trash 3s and loses by 8 and can be slightly rewarded for that. Also, I really don't think NET and KenPom at least adjust for opponent quality that much because I have seen too big of jumps for teams beating up on bad teams.

Don't get me wrong, things like KenPom and the NET are about the best things there are right now, I just think especially in terms of the NET, do something a little more to try and make it the best and differentiate it some from other metrics, especially if it is your main tool.

 

A lot of people still jump to the RPI to justify their team. Maybe it is old hat, some of just not knowing how it works these days. Always have to say if it makes you feel better look at it, but it isn't used at all anymore. It has no impact in selection in any way.

 

It's within the bounds of technology these days to have an average margin over the course of a game and compare it to an average of a typical game at the same point and work an efficiency metric off of that. So if the average halftime lead is, say, five, and you lead by 10, you get credit for that and vice versa.

Failing that, the margin of victory should be capped around 12-15, i.e., the margin that takes late-game fouling out of the equation which often artificially increases win margins (see anyone who bets on basketball for advice on that) balanced against domination of an opponent from start-to-finish. Once you get past 15, it's overkill that shouldn't be rewarded.

Me? I couldn't care less how much a team wins by, which is why I lean a bit more towards the results-based metrics like KPI, SOR, etc.

 

Can we call it an impending departure yet? They’re not gonna have a resolution any time soon, I don’t think.

 

Only relevance I see here is with a team that has multiple lopsided wins. Or losses, more likely. Virginia has a good overall and league record, but a ledger with 5 losses by 22 or more points has to give the committee pause.

 

Also? College basketball writers need to question the metrics more and not treat them like gospel or as an infallible reference source. Roscalbo brings up several flaws in the logic underpinning them which isn't always so logical.

I have nothing against Kenpom, for example, it's an interesting tool, especially for things like tempo, etc., and often accurate (though I prefer Torvik), but there are flaws in the predicative metrics (Kenpom prices in previous season data, especially near the start of a season, which in this transfer portal age is well nigh a disqualifier) that simply don't get challenged often enough.

My opinion is that Kenpom specifically drafted off the heat of being an alternative to the RPI (which was flawed, but not as horrid as people said it was, the principles of it were at least sensible) and basically became this unchallenged source on all things predictive. Sagarin was that for a long time, but he fell out of favor too - and doesn't even have a college basketball rating anymore for reasons, so far as I know, still unexplained because he still rates other things.

Too often you see copy without context that says "Auburn is 7th in Kenpom" as if its been handed down on a tablet straight from God's Top 25 poll without asking why Auburn is 7th? Particularly when they only have one Quad 1 win? The method of what got them there should be questioned when there isn't an obvious underpinning of evidence supporting their spot. It doesn't mean the methodology is wrong, but it should be better explained.

 

My crude unofficial stat for one team thoroughly dominating another is how much time the spend with a lead that matches or exceeds the minutes remaining in regulation. It is possible to come back from that, but not very likely, particularly when it extends beyond a one-possession game. So if a team wins by 16 but maintained that margin throughout the second half, to me they’ve proven the same point they would have winning by 20-plus.