clock menu more-arrow no yes

Filed under:

UNC Basketball: Peeling back the layers of Roy Williams’ late-game timeout usage

New, 1 comment

This exercise probably raises more questions than answers, but here are some numbers to chew on.

NCAA Basketball: North Carolina at Virginia Geoff Burke-USA TODAY Sports

In college basketball, there are essentially four main objectives of calling a timeout:

1. To keep possession of the ball.

2. To provide players rest or recovery from injury.

3. To stop the opponent’s momentum.

4. To draw up a play.

UNC coach Roy Williams has a fairly well-deserved reputation for being peculiar with his timeout usage and lack thereof. The first two functions above are ones likely every coach uses from time to time, though it does seem like Heels players are discouraged from calling a timeout to escape pressure until at least the very end of the game.

The third function is the one that probably angers Heels fans the most – at least me. Almost universally speaking, Williams simply refuses to call a timeout to try to stem an opponent’s run and often just hands over bunches of points that could have been limited and kept Carolina in the contest. Anything short of a 15-0 onslaught – and sometimes even that – will not elicit a timeout from Williams. This is something literally every other basketball coach does on all levels. Heck, in the NBA, played by professionals, all it normally takes is two or three opponent buckets in a row during the second quarter of a meaningless regular season game in December for a coach to stop the action. The main justification is that Williams wants his players to learn from and grow through adversity and essentially figure things out themselves for long-term benefit.

However, that angle is a little harder to study, at least with the resources I have. My bet is that taking a timeout has a relatively significant impact on quelling an enemy run, particularly early in games, but who knows.

Williams receives almost as much or equal criticism for his stance on the fourth and final purpose; that is, he typically does not call a timeout when tied or trialing in crunch time, even with several in his pocket, and instead lets his players go for broke. But fans might be surprised to learn how often he does call a timeout. At least in a crude way, this is probably the easiest one to examine empirically for its potential effect on the scoreboard.

So is there anything to all of this bellyaching by a segment of Carolina fans and haters about Williams and his tactics? A lot of times after a failed final play without a timeout, I have thought to myself, “Wait, the game could still be going on right now and UNC would still have a great chance to win with a brilliant play!” Is that just crazy hindsight and sore loser talk? Well, sort of, and sort of not.

To narrow this down as simply as possible, I have attempted to take a survey of situations in the 16 seasons of the Roy Williams era within about 80 seconds of the end of regulation or overtime in which the Heels possessed the ball and the score was either tied or Carolina trailed by three or fewer points. A bunch of caveats spring from this criteria, which I will touch on throughout the rest of the article, but this working definition captures what I am seeking pretty well. It might not snapshot the true last-second phenomenon, but it also expands the sample size to include fairly similar situations and strategies.

I found 135 valid instances across 68 of Williams’ 586 games at Carolina – on average almost exactly two per close contest, which constitutes 11.6% of outings. The most recent two were in the final 31 seconds of the ACC tournament semifinal loss to Duke in March when UNC trailed by one point each time. Williams let it roll on both, and Cameron Johnson misfired on a three-pointer before Coby White slipped and missed a long two following two bricked free throws by Duke’s RJ Barrett (Williams did call TO before Barrett’s misses, but I did not include what was only a couple of this type of timeout for reasons discussed later). The last successful example was at the end of regulation of the memorable battle against Miami in the Smith Center in February, when the Heels trailed by three with 16 seconds left and Williams called for a nice pick-and-pop play for Luke Maye. The legend drilled another big shot and Carolina went on to win in overtime 88-85.

As many as six crunch-time scenarios occurred in a single game (vs. Duke in a 2004 OT loss), and as few as two in an entire season (2012-13, both in an OT victory vs. Virginia Tech). It’s important to keep in mind that this certainly isn’t a representative cross-section of all Carolina teams under Williams; for example, the infamous 2009-10 squad dealt with an above-average 11 of these situations. The Heels have compiled a 29-39 overall record (.426) in these games, which is not too shabby and perhaps quite impressive considering that they were trailing in over 70% of the examples and by multiple points in nearly half.

Williams called 56 total timeouts in 47 of these situations and did not call a timeout in 88 others (nine times the setting was virtually the same and Williams just called another TO). The better than one-third rate (34.8%) of calling a timeout hardly qualifies as “Roy never calls timeouts,” at least at the end of games. It’s also important to remember that, while Williams’ rate might well be lower than most other coaches’, there are naturally a lot more “no’s” than “yes’s” because timeouts are in fact a finite resource and passing one up in a tight squeeze is not super abnormal in up to an 80-second time frame.

In any event, the results of each type of possessions are as follows:

UNC possessions after late-game timeout or no timeout

Timeout called by UNC? Number of instances FGM-FGA 3PM-3PA FTM-FTA OREB TO PPP
Timeout called by UNC? Number of instances FGM-FGA 3PM-3PA FTM-FTA OREB TO PPP
No 88 26-72 (36.1%) 8-28 (28.6%) 27-33 (81.8%) 15 8 0.989
Yes 47 14-36 (38.9%) 8-23 (34.8%) 11-17 (64.7%) 6 5 1.000

The stats are basically the same in each scenario, with a slight uptick in field goal and three-point shooting, and a negligible edge in points per possession (PPP) when UNC calls timeout in a given crunch-time possession. I’m going to spare you and me p-value analysis and disclaim that the discrepancy in these figures is almost surely not statistically significant. For example, is the worse free-throw percentage following a UNC timeout because the break builds pressure and ices the Carolina players? No. That has to be random noise.

Other numbers do make some straightforward sense, like the lower than usual offensive rebounding in both scenarios because limited second-chance opportunities on last-second plays influences the data. In addition, the PPP of exactly or almost 1.0 would be a pretty lousy overall figure when factoring in the normal flow of the whole game – for reference, last season’s Carolina squad posted the 19th-best mark in the country of 1.13. But in these hard-pressed situations, I suspect that stacks up well in comparison to other teams (I won’t want to hear this after the next loss to Duke when Roy passes up a timeout, however).

This information appears to let Williams off the hook – hey, it’s a wash, especially in the most important category of PPP. Not so fast, though.

A quick aside: Williams has stated on at least a couple of occasions, most recently after Luke Maye’s heroic shot against Kentucky in the 2017 Elite Eight on the way to the national championship, that he likes to call a timeout in a close game when six or fewer seconds remain on the clock, and let his players continue to play with seven or more seconds left. This was awfully convenient to say after Maye won the game on a possession without a timeout that happened to start with 7.2 ticks on the clock, but is it true? Yes, generally speaking, UNC calls a timeout the majority of the time under seven seconds and declines the majority of the time with at least seven seconds.

However, let’s take that principle and ask: what if Williams were willing to use more timeouts with more time left? Instead of looking on the possession level, here’s a big picture look at the results of games in this data set in which UNC called at least one “non-Roy” timeout (seven seconds or more) versus games in which they took none:

UNC record in games with and without late-game timeout

At least one timeout called by UNC with 7-80 seconds remaining? W-L record
At least one timeout called by UNC with 7-80 seconds remaining? W-L record
No 13-23 (.361)
Yes 16-16 (.500)

Again, I’m not asserting statistical significance, but it does make you scratch your head a bit. To recap: UNC has won 36% of tight-pinch games in the last 16 years in which Williams did not take a timeout with seven or more seconds to go; the Heels have won 50% of games in which Roy did take at least one such timeout. Perhaps the biggest surprise of all to UNC faithful might be that Williams has called TO in almost half of games in which your average coach “should” or customarily does call at least one.

What explains this difference? A big part of it is that not all of these late-game scenarios, clearly, are created equal. The neat thing about this dataset is that the four different margins (0, -1, -2, -3) are roughly equally represented among the 135 situations, but Williams’ inclination for when he calls a timeout is not:

UNC late-game possessions by margin

Margin Number of instances Number of times UNC called timeout FGM-FGA (timeout and no timeout) 3PM-3PA FTM-FTA OREB TO PPP
Margin Number of instances Number of times UNC called timeout FGM-FGA (timeout and no timeout) 3PM-3PA FTM-FTA OREB TO PPP
0 36 17 (47.2%) 11-28 (39.3%) 4-12 (33.3%) 5-8 (62.5%) 3 3 0.861
-1 35 9 (25.7%) 10-32 (31.3%) 5-11 (45.5%) 9-10 (90%) 9 4 0.971
-2 29 10 (34.5%) 11-22 (50.0%) 1-7 (14.3%) 12-17 (70.1%) 4 1 1.207
-3 35 11 (31.4%) 8-26 (30.8%) 6-21 (28.6%) 12-15 (80%) 5 5 0.971

There is a lot to unpack here. First, Williams apparently prefers to call timeout with the score tied, doing so almost half the time, which was particularly true early in his tenure (it has happened only three times in the last six seasons). This sheds light on the middle chart above – Williams calls timeout more frequently when the game is tied, which is obviously when his team needs fewer points and thus it’s easier to win, whether on that trip down the court or later. This confounding variable is mostly why UNC is more likely to win when Williams calls at least one seven-seconds-or-more timeout in a close situation.

Interestingly, though, that is also when Carolina produces its worst offense by far among these margins, whether following a timeout or not. When Williams likes to call timeout is when his team is about to play worse than all 352 Division I teams besides Chicago State did in 2018-19. That doesn’t sound too flattering, but it’s also somewhat of a chicken and egg deal because he might call a timeout suspecting that that’s when his team needs the most help, plus at least a handful of the tied situations are no better than heave chances. Being closest to victory bumps up the overall return and win percentage, taking into account the superior efficiency, yet similar win expectancy of UNC’s offense when trailing by a point.

Carolina’s two-point-behind approach, meanwhile, is apparently enough to sustain an offense greater than those of all teams last season besides Gonzaga, despite having made only one long ball and subpar free-throw shooting. The same warning about sample size exists, but it does intuitively make a little sense that this scenario is conducive to the Heels regular offense – they are willing to run, aren’t trying to burn clock for a final play, don’t need to rush a quick two or chuck a bad three, etc. I’m not sure why it has been so much more effective than UNC’s attack down one point, which happens to have produced exactly the number of points on exactly the number of possessions as when Carolina is down three. But as always, it could be noise.

Despite hitting six three-pointers and some free throws in 35 possessions when trailing by three points, the Heels have had a miserable time pulling off wins in these cases. Just twice have they done so, including the aforementioned OT win over the Hurricanes last season. You might remember a certain negative example from late in the 2016 season, which included a miraculous three-point shot by Marcus Paige to tie the title game against Villanova and force overtime … oh wait. Anyway, if you remove this margin, UNC has won over 55% of the other games in this dataset. Clutch.

There are many different ways to skin a cat. Williams clearly calls some late-game timeouts. These examples don’t even include when he takes one after a made basket or free throw or dead-ball turnover to set up the defense, which he does regularly and matches Carolina’s style of play, or when the opponent whistles timeout for him anyway. And it’s easy to forget the times Williams calls a TO that leads to a big basket or game-changing stop outside of the last couple of minutes that don’t get underscored – he did so with about five minutes remaining and the Heels on the ropes in wins over Arkansas and Kentucky on the way to the 2017 title. But I tried to pin down just these last-couple-of-shot-clocks, do-or-die moments when critical and/or nervous fans are usually screaming “take a $%*!^#@ timeout!” and Williams has the decision solely in his hands knowing the deficit to make up for certain as the offense gathers possession.

It’s fair to ask if Williams and his staff could look at ways to potentially better leverage UNC’s timeouts at the end of close games to find even more success. Williams is fond of hoarding timeouts for what one could only assume is doomsday strategic purposes, and I love the thing he occasionally does with minimal time on the clock when he calls a play to advance the ball near midcourt before quickly calling another timeout in a more advantageous position. But the chance of scoring/winning using this poor man’s NBA timeout has to be at most 5 or 10% anyway, right? Down one or two points with 50 seconds left, the data clearly show it could be a different story. Remember that there appears to be at least some small benefit of the Heels taking at least one timeout with substantial time on the clock during a given close game.

I don’t quite have the time and resources to crunch these numbers for the rest of the coaches, so it’s hard to know how usual or unusual, effective or ineffective, Williams’ timeout practices are. I don’t mean to tell a Hall of Famer how to coach. It’s possible he is already aware of this, as he has shifted more timeouts from tied to trailing positions in recent years. But without too much conviction, I suggest that Williams was/is either not teaching tie-score plays well enough or he’s taking timeouts when they aren’t of much use anyway. There’s no way I thought I’d ever say that before looking at the numbers.