ACE: Need your help - suggestions for basketball UFR

Submitted by Ace on November 18th, 2011 at 12:18 PM

As football season wraps up (I know, sadface), I'll be focusing a lot more on basketball, and Brian put forth the idea of coming up with a basketball version of UFR. I have a general idea of what I want it to include, which I'll list below:

  • Rating shot quality: Contested jumper, wide open jumper, uncontested fast break layup/dunk, etc.
  • Assigning points for good (and bad) plays off the ball, like picks that lead to layups, bad fouls away from the play, etc.
  • Forced turnovers vs. unforced turnovers
  • Defensive culpability—who, if anyone, was at fault for a given opponent basket
  • Perhaps a rebound difficulty chart, resembling what Brian does with catches for Michigan receivers

I probably wouldn't post after every game, since that would be somewhat insane, and would also make it difficult for me to do anything else on the site, but I'm looking to put a hoops UFR through a test run during the Maui Invitational—since I'd still have FFFF to post on Wednesday, it might not go up until later in the week, but I want to start ironing out the kinks before Big Ten play begins.

Anyways, I would love to hear suggestions for what should go into these posts, and also any ideas for a name for this feature ('Basketball UFR' just feels like a copout) [never mind—UFR uber alles]. This would be focused on what doesn't show up in a box score or on KenPom, and yes, I will be going through play by play—this is where I thank the Big Ten for playing at a snail's pace.

So yeah, the floor is open for suggestions. I'd love to hear what you all think. Thanks in advance.



November 18th, 2011 at 1:53 PM ^

Let me give you an example:

Pretend Michigan is running a base 1-4 high offense. They run this offense through an entire cycle and end up with 12 seconds to go on the shot clock. After this, Burke resets the team, goes into a high-low post set, utilizes a screen-roll, and gets a 15-foot jumper which misses. The opposition gets the rebound.

Based on what I just wrote, a "results-based charting service" would look at this, possibly see three missed screens (of course, do we know the picker was supposed to screen, or just shadow screen and drop, etc.), maybe see two rebounders and ding them for not getting what looks to be an offensive rebound for the taking.

Of course, the decision on who/how/why to screen depends on not only the offensive player but also the defender and the defense being played. Also, your decision to go for the boards isn't objective, it's subjective (how many fouls, who are you defending, what's the lead, etc.) Finally, what's the defensive playcall - 1-3-1, 2-3, matchup based on where the ball is, press, box and 1, triangle and 2, zone-to-man after 2 passes?

I'm not saying someone can't do this. Best of luck to Ace - I tried doing one half of a Gopher game a couple years back and I quit after five minutes of game action. What I'm saying is that without knowing more information the ground, this becomes an exercise where the analyst is creating facts to fit a story/theme.



November 18th, 2011 at 1:54 PM ^

As a dude who's played a lot of b-ball, I'll tell you it's complicated.  The reason is that there are a lot of open-ended plays.  It's like an option in football, but instead of reading one defender, then bitching or going, you're reading multiple dudes, then it's three passes later than someone scores.  Difficult, but not terribly complicated right?

Well, what happens when it's a mis-read?  Or, rather, you THINK it's a mis-read?  Is it the ball-handler's fault for not reading the options, is it his teammate's fault for not being in the correct position, or is it some combination?  Adding to that, what happens when the play gets derped up, yet THJ just dunks on a fool because the D derped-up worse?  Is that a +3 for "dunking on a fool" or a -1 for "WTF, run the damn play?"

Also, one play leads to another quite often, and plays can sometimes not develop.  For instance, if I'm supposed to run a play where I pass to the big on the block, yet my defender goes "I don't want to defend you" and I drive by for a score, then how do you rate the big?  I mean, he ran his part of the play, but he didn't really add anything USEFUL to the offense, so?...

In summary, it's a complicated picture.  Yes, it can be done.  This is why there are coaches and assistants.  (My last head coach's previous job was to do just this for the Cavs.)  While it can be done, it requires a LOT of video, and a very good understanding of things.  IE, why do people drive to the hoop.  (And I'm not talking about "because it was open."  I want to know WHY it was open. And again, that's complicated.)

el segundo

November 18th, 2011 at 2:08 PM ^

Michigan is playing Ohio State and playing man-to-man defense. The coaches have told the players that, when OSU runs a certain play, one guy is supposed to leave his man and double Sullinger. This assignment only applies on this one play (and maybe only when certain personnel are running it for OSU). The Michigan defender who should double does not, for some reason, and another defender tries to make up for it, running to double Sullinger but arriving late and being ineffective.

If you don't know the defensive rule that the coaches applied, you probably mark the doubling defender down for being late and ineffective. But he's not the one who deserves blame on the play.

I'm not sure basketball is well suited to a quantified system that gives individual player grades, especially not when an outsider is doing the grading.

Why not do something like what NBA Playbook does? That's a terrific website.


November 18th, 2011 at 1:24 PM ^

Ball Pressure-I think it is very important to consider how much pressure we put on the ball defensively.This would be very similar to the pressure metric in the UFR. Putting good pressure on the person with the ball makes it harder for him to make plays for anyone else. Tipped passes or hard pressure that leads to a steal could be pluses while allowing easy passes into the post or allowing good shooters to take easy shots would be minuses.


November 18th, 2011 at 1:34 PM ^

First of all, you've got a decision: do you want to make it based off athleticism, technique, or both?  For instance, at 3:27 2nd half of last game THJ makes a catch that's purely athletic.  Technique-wise this is a poor play, athletic-wise this is a great play.  My guess is you want to lean more towards technique, with a few allowances for great athleticism (for instance, making an amazing dunk, after screwing up the play.)

So, you want to know more than just regular stats.  Obviously, advanced stats are a good place to start.  However, I've got a few additional stats that I watch, based off our team's goals in each game.

Assists: they suck as a stat.  What happens when I pass to my teammate for an alley-oop and he DROPS THE FRIGGIN BALL! (yes, this did happen.  There was no one else near.  It was a good pass.  I want my damn assist.) 
Assisted Shots: are a good stat.  It shows how many times you give people wide open shots, which you directly control.  How I measure this is assisted shots (wide open shots) are the denominator.  Then, the amount of MADE assisted shots is the numerator.  IE, if Novak misses an assisted shot by Burke, Burke gets a 0/1 AS.  If then Morgan dunks an AS by Burke it's 1/2.  You get what I'm saying?  Of course you do, you graduated from UM.

Rebounds: suck as a stat.  Yes, you control whether you've boxed your man out.  No, you don't control whether the ball comes your way, nor whether your derp teammate standing next to you has boxed his man out.  Nor can you control getting screwed on a switch as a 5'10 guard against a 6'10 giant/leviation/behemoth. (This happened also.  I got put-back-dunked on.  I was mad.)  So...
Box-outs: number of men boxed out after each missed shot.  For instance, if Burke boxes out that behemoth then Burke gets a 1/1.  If Novak misses his man, REGARDLESS OF WHETHER THE BALL COMES TO HIM, that's a 0/1.  Add these totals up and you've got X/5.  For the whole game, this should be a good indicator on whether we're keeping up on the boards.  It's better than pure rebounds, rebound %, etc because the players each control whether they get that stat.

On plays: They're super-hard to read in basketball unless you know what you're looking for.  Kinda like FB, but much more free-form.  This gets especially true when the game speeds up and when the shot clock is winding down.  I watch every game and even I don't know all the plays that we use.  (Granted, I don't watch for every play).  So, it's pretty hard to know which play is happenning, especially since multiple plays commonly happen on one possession.  Obviously, you'll want to look for the hand signal of the ball-carrier, but this makes it harder as you then have to pause and slowly watch the OFF-BALL players go through their progressions.  And even then, the play usually goes awry as something else opens up.  So, unless you're watching multiple games, you're not going to know many of the plays fully.  This leads to the problem of "how do you tell whether someone did their job in the play".  I really don't have an answer for that besides "watch a shit-load of plays, yo". 

Random other poop: Our team needs scorers.  But scorers don't just score points.  They go through progressions in a play, in a game, and over a year.  The important point is to notice in which positions they score.  For instance, Burke is a shooter (if he drives, he will dish against better opponents.)  But, more specifically, when he's left in an iso play, he'll drive and look for the lay-up.  THJ is a scorer (both shooter and driver).  But, most of his drives are off isos and screens from the man in the 5 position (Morgan, Horford, or Smot).  So, a good measure for how active offensively THJ and Burke are is "how many isos are called" and "# of screen/rolls with the 5".

I don't know if any of this is going to be used.  I do hope it's been a help.  The main points are:
1.) You aren't gonna know many of the plays if you don't watch a lot, bro.
2.) A lot of stats suck.  Use that brain of yours to figure out which common stats do, and throw them out.  Try to use ones that each player actually controls.
3.) Please, friggin, please don't just say "THJ scored because he beat his man off an amazing dribble."  Know the footwork (was the defender leaning left?  How high was his butt?) and incorporate it.  The players do.

Oh, and uh, good luck.

P.S.  I am a basketball nerd.  I keep these stats for fun (I am a lonely, lonely person).  I would be willing to assist in helping a UFR for basketball in the future.  In fact, I would be enamored with the idea.  Since you should know my email, if you want me to research this stuff further, shoot me an email.  Otherwise, I just hope this helped.


November 18th, 2011 at 1:53 PM ^

My concern is that where football has a lot of straight forward assignments, i.e., scrape over the top, block this guy, cut this back, etc., basketball is much more free-wheeling.

For example, let's say Jordan Morgan sets a nasty screen on THJ's defender, freeing up THJ for a wide open 18-footer. Morgan would be +1'd. But then THJ has the shot rim out. There was no benefit to Michigan on that play minus the screen.

It seems Brian has a value to each play and the +/- portion of the UFR reflects that. An 80-yard run gets so many pluses. A 4th and 1 play action on the road in blistering wind gets the opposite in minuses.

As someone who played and coached and now obsesses over advanced basketball stats, let me know if you need any help.

Quick suggestions:

Guards -

+ for good entry pass to the low block, direct assist for a basket, direct assist leading to two free throws, denying the wing pass that starts the offense, denying an entry feed, steal, making a shot, taking a charge

- for ignoring man in the post with position, turnover, getting beat off the dribble and breaking down the D, getting called for a charge, missing a layup, going iso when the situation doesn't call for it

Big Men -

+ for solid box-out leading to a rebound, offensive rebound, block, help defense leading to a contested miss, man defense leading to a contested miss, pick freeing up another player for an easy basket

- for dropping an easy pass for a basket, giving up an offensive board, not showing well on a screen-and-roll, failing to maintain help on a drive

It's going to be a challenge Ace. Good luck.


November 18th, 2011 at 1:59 PM ^

UFR seems to come from when the NFL first started having replay and the ref would always come back the field and say "Upon further review..." Now they seem to just say whatever the hell they want, but usually "after review..."

I don't know what the basketball version of this is. Is there any standard saying for when a play is reviewed? I think they just make a hand/arm signal.

That is a long way of saying I think you should stick with UFR.

Also, I believe other Michigan/other schools' sites have done something similar. Maybe you could try researching the college bball blogosphere to get some resources, so UFRs are somewhat consistent. I'm sure some nerd at Duke is coming up with something to help you. I wouldn't suggest looking at MSU blogs, though.


November 18th, 2011 at 2:01 PM ^

I think this idea is great and it will definetely help us all understand how our team is developing! Looking forward to it.

A quick concern: A count on ESPN's play-by-play showed about ~100 posessions total from the W. Illinois game.  I feel if you break it down by possession is may be A LOT of work. Have you thought about breaking it down another way? Maybe like minute or two minute intervals? Or team and media timeouts? And then between those breaking everything down?


November 18th, 2011 at 2:36 PM ^

If you want the points to mean something, you need to be consistent with how they are given out.  Maybe the best would be to create a certain number of points per offensive or defensive possession and then just figure out how to split them up based on what happened.

This isn't like football where you can get many results of different value to the final outcome except maybe a bit of magnification of value down the stretch.

Maybe something like this would work:

Assign something like an expected value for a possession - like +1

Offence - +1 for 2 points scored, +2 for 3 points scored, 0 for a single point (FT), and -1 for a possession with no points.

Defense - +1 for a stop, 0 for a single point, -1 for a basket given up, -2 for a 3 pointer

Then the rest is giving credit or assigning blame by judging all the factors involved.  If you don't want to put it on a specific player for a great shot by an opponent, you can create a TEAM category which can also be given the point hit.  But likely, it would be better to just spread the points evenly among the players on the court.

This way, by evolving a consistent assignment of points, you can determine how much each player is contributing to a final result.  Big wins will have a bigger plus result and losses will be the opposite.

In the end, you get something similar to a  + / - charting, but you will be making more specific assignments of value rather than just being on the court.


November 18th, 2011 at 2:39 PM ^

Just want to say thanks to everyone for their input—I'll be taking all of this into consideration when I put the initial UFR together (it probably won't be up next week, since it's OSU week, but early the following week is the goal). If I didn't get to your comment, it's not because I didn't read it, only that there were so many comments (and for that, I thank you all again).

I'll keep checking in on this thread, so keep the suggestions coming if you're so inclined. I really appreciate all of the input.


November 18th, 2011 at 3:23 PM ^

Is a hockey UFR possible in the future? If you can tackle basketball (IMO the hardest part of doing this for a sport other than football is breaking down the beginning and ending of plays consistently, there aren't cut and dry plays in hockey or basketball like football) hockey should be possible as well. 

Obviously if you're doing basketball and recruiting you wouldn't have the time, and Brian's swamped too, I'm sure, but has there been discussion of getting one going after you get basketball up and running?


November 18th, 2011 at 3:41 PM ^

We haven't discussed hockey stuff yet, but that one could be a lot more difficult to do than football and hoops. First of all, there's the issue of plays—football's obviously clear-cut, while hoops has definite possessions. Hockey... not so much. The bigger issue is with TV. To do these, you really need to go over (and over and over) game tape, and hockey just isn't televised that much at this point.

There's also the issue that many, many more people care about football and hoops than hockey, and we all have only so many hours in the day—the time it would take to put together could probably be better spent on other things related to the blog. Maybe it's possible down the road, but I wouldn't hold my breath for a hockey UFR coming out any time soon.


November 18th, 2011 at 3:18 PM ^

a college basketball coach I knew pretty well said one of the most telling stats he focused on was points per possession.  Now I haven't attempted to keep track of that stat in many years, but I remember the number .85 pts/possession as being a benchmark number.  I'd be curious to see if that number is still accurate and if the stat itself is relevant.   Obviously it's a good way to get a gauge of how a team compares to others in terms of possession efficiency, but the benchmark number may be way off.


November 18th, 2011 at 4:19 PM ^

especially, I would be interested in seeing average time of possesion. At least in basketball, the mark of a good defense is running down the shot clock low and often. So I think average time of possession (minus those with offensive rebounds) would be a good benchmark.


November 18th, 2011 at 4:32 PM ^

The assassin metric - those moments in games where someone (usually THJ last year) just turned into an assassin for about five minutes late in the game - I remember a couple of games last year where THJ basically stepped on their throat, threw them in a box and put the final nail in. Try to measure who tries to be that assassin and how successful they are - I.e. Burke may try to go off and finish them off but miss the shots, turn the ball over, etc. Or on another occasion go off and make a steal, a couple of threes, and a couple of great assists to put a team away. This would shed light on who is going to try to step up and who we actual want to step up. And to determine/track whether the team is finding a killer instinct.


November 18th, 2011 at 6:35 PM ^

I would suggest a minimal table of requisite information for each possesion followed by a list of plus minus contributions on an as-appropriate basis for each play. Trying to score everyone in a number of different areas on each possession would be tediuous and produce a hard to read table.

So each possession might start with a header row with Off/Def, clock time, generic play type, possession result, and the game score following possession.

This would be followed with 0-? rows of +/- scores for individual players as appropriate. This could consist of Player name, score +/-, score category (e.g. hustle, passing, shot selection, ... could be quite a few categories), and a  brief specific comment.

An optional general comment on the possession could follow, comparable to the description field of the football UFR.

Summary tables could add up all the plus minuses and break out by category.

Anyway, that is one implementation I would suggest. Of course, I am also looking at this from the perspective of how to turn it into an enhanced version with linked video, so I may have specific biases :)