Blown Leads, Comebacks, and Close Games Around the NBA

Note: the data for this piece was compiled as of January 5, 2019, prior to that night’s games.

I’m not sure if you knew, but the Sixers played the Suns a couple days ago and there sure was some gnashing of teeth after the 132-127 road win due to the...difficulties of the fourth quarter. Before I recommend to Elton that he should trade Joel, Ben, and Jimmy for a few second round picks, I wanted to investigate if the kind of fourth quarter shenanigans that occurred were Sixers-specific or a part of the NBA as a whole.

The first step (which should be familiar with the readers of this website) is to sort of divorce ourselves from the results of games and focus on the process. If the Sixers are up by 15 at the start of the fourth quarter and then win at the buzzer when the opponent misses a potential game-winning field goal, is that game appreciably different than if the Sixers are up by 15 at the start of the fourth quarter and then lose at the buzzer when the opponent makes a shot at the buzzer? I say no — despite the outcome being different, how the Sixers got there is mostly the same.

Basic Methodology

For every single game played in the NBA this year, I took the scoring margin at the start of the third quarter (Q3), start of the fourth quarter (Q4), and at approximately 60 seconds left in the fourth quarter. My thought is that to fully remove the outcome, we can treat every game with a five point or less margin within 60 seconds to go as a “close game”, regardless of the ultimate result. This way, a 20-point comeback to get to -5 at 1:00 in the fourth quarter is essentially the same (for this specific metric) as a 20-point blown lead to get to +5 at 1:00 in the fourth quarter, or any other permutation of previous scores as long as the margin is +/- 5 at 1:00 left. For bonus fun, I have the win probabilities attached at each time interval as well so we are able to gauge the relative quality or likelihood of each comeback/blown lead. Think of each section here moving forwards as a collection of related chapters/vignettes, not exactly a linear stepwise path to some grand conclusion.

Type of Game Outcomes

For our purposes, there are only three outcome types. Games with margins at 1:00 to go in the fourth quarter (Q4:60) greater than +5, less than -5, and those within 5.

Just from a visual inspection, it seems fairly intuitive. The Knicks, Hawks, and Cavs seem to have about half of their games entering Q4:60 down more than five points, and the Bucks, Raptors, and Thunder seem to only have about 10-20 percent of their games end like that. What’s interesting to me is that it appears that the proportion of games at Q4:60 within five is far more uniform across the board compared to the larger margin categories. Figure 2 takes a look at that in more detail.

I might have to save digging into this for a future article, as I’m not sure what exactly to make of this right now. Or if I’m missing something obvious, let me know in the comments.

Win Probability at Q3 and Q4

Combining NBA.com second-by-second win probability data, we can start to examine actually how win probabilities change from Q3 to Q4. My thought is that a team prone to collapsing would likely have a higher (relative) win probability at the start of Q3 than at the start of Q4, whereas a team who is strong later in games might have an increased (relative) win probability at the start of Q3 compared to the start of Q4. Figure 3 presents the spread of those deltas for each team.

Again, it seems to more or less pass the eye test. The Warriors, Raptors, Bucks, etc. all have median increased win probabilities at the start of the fourth quarter compared to the third quarter, whereas the Hawks, Cavs, and Wiz (hah), etc. are the opposite. If you’re trying to squint to figure out if Brett Brown should be immediately fired or given a lifetime extension based on if the Sixers’ value, it’s -0.00695.

Good Teams and Their Q4 Margins and Q4:60 Margins

My hypothesis is that good/sustainably successful teams enter the Q4:60 up by more than five in order to prevent a series of stupid or unlikely events causing them to lose. With an average of about 14 second per possession, and assuming the team in the lead might run some clock a bit, that’ll get you about 2-3 possessions per team. If we assume the team in the lead at Q4:60 scores 1.0 ppp for either the two or three possessions they have, the only way to lose is to give up between 7-9 points on 2-3 possessions, a fairly unlikely scenario. Figure 4 takes a look at the relationship by team between the average margin at the start of Q4 and the average Q4:60 margin. Obviously, using averages smooths out the bumps, but that is precisely the point. Of note, the line here is a reference line, not a regression line.

I colored the sections of the graph based on quartiles of team quality assuming my hypothesis to be true. Lo and behold, I think this does a nice job of separating the teams. I also think this is a solid indication that the Kings might be due for a bit of a backslide, and that the Pels are probably better than their record indicates. Also, the Bucks are good and Jason Kidd might be the worst NBA coach ever, aside from Jim Boylan. Plus, we might be underrating the Pacers a bit.

Actual Counts of Close Games, Wins, and Losses

Let’s put actual wins and losses back into the equation at this point. Using five-point chunks for largest leads (Table 1) and largest deficits (Table 2), we can count up the number of games that ended up as wins, losses, and close (within 5 points either way at Q4:60). These are for the entire NBA, so the tables are large. You can search and filter through them at your leisure.

Upon a brief review, it appears the Sixers do not profile as well as the Raptors or Bucks (shocker). However, we are talking about a game or two in either direction. I certainly do not see anything that indicates the Sixers have any sort of significant WIP caller CHOKE ARTIST situation going on. All teams blow leads to certain extents. However, they do seem to have a knack for letting teams back into games a bit. Of Philadelphia’s 20 games that they led at any point by 16 points or more, seven of them (35%) ended up within 5 points either way at Q4:60 — they did win 18 of those 20 games (90%), however. For comparison, the Warriors have let three of 19 games they led by 16 or more points (16%) get to within five with at most 60 seconds left in the fourth quarter, winning 18 of 19 (95%). The Celtics are at 1-of-14 (7%) and 13-of-14 (93%), and the Bucks are at 2-of-20 (10%) and 19-of-20 (95%), respectively. I have a strong suspicion that if we gave Joel robot legs to play him 42 minutes per night, we wouldn’t be seeing as many large leads dwindle.

Update: Requested by Commenter

Here’s what the stats look like for the 2015-2016 Process squad: ten games with leads of 16 or more points, with four turning into close games at Q4:60 (4/10 = 40%) and winning five of the ten (5/10 = 50%).

If you like these kind of articles, respond to the poll or leave any ideas you have (broad, narrow, well thought out or random pieces of an idea, etc.) in the comments.