Interview With Mercenary Analytics’ Joseph Gill

Readers of Gator Country know I love analytics and the role they play in basketball. One of my absolute favorites in the analytics community is Joseph Gill of Mercenary analytics, someone whose brilliance in breaking down the game is matched by his ability to communicate those ideas clearly and in a way anyone can understand. He shares my intense passion for the game of basketball and will give you the best conversation (about basketball or anything else) you could ask for. I’ve long wanted to ask him some questions about Florida basketball and he was generous enough to give me hours of his time to break down everything I threw his way. He reinforced lots of the ideas I have written about in the past while also challenging some of them. Everything is broken down masterfully and I think you’ll really enjoy what he has to say.

EF: Florida played really slow last year. Is there a sweet spot for tempo you like to see teams playing at or do you believe teams can have success at any speed?

JG: I always lean towards playing as fast as reasonably possible. I view basketball as a per-possession efficiency battle, and I like to plan on having an advantage in that aspect of the games that the teams I consult with play, so therefore I would like to increase my per-possessions advantage over the most possessions possible.

I would never advocate for an extremely fatiguing strategy like the breakneck Grinnell System that once saw guard Jack Taylor score 138 points in a 2012 game, but the team I consulted with the most heavily last season, D3’s St. Thomas had a Pace of 70.1. That number would’ve tied for 99th out of 353 D1 teams last year, and would’ve been the 12th fastest among Power Five Conference teams. By comparison, Florida has a Pace of 64.0. Florida International was the fastest D1 team last year with 80.0 possessions per 40 minute game, while Virginia was the slowest team in the NCAA last year with 60.2.

Let’s use a team we should all be familiar with as an example to why this per-possessions style of thinking is important: Virginia. Last year’s Virginia team had the 4th best offense by Team Offensive Rating and the 6th best defense by Team Defensive Rating, according to Across their entire season, for every 100 possessions, they outscored their opponents by an average of 25.2 points.

Let’s simply it down to terms that are more accessible. Another way of thinking about this is that in a world where we convert Per 100 Possessions scoring into simple two-points percentages (to adjust for but also “eliminate” offensive rebounding and turnovers for simplicity’s sake), it was like Virginia shot 58.8% and their opponent’s shot 46.2%. Over the long-term, this is a dominant, dominant team. But for individual games, there is still the chance that on a hot shooting night for opponents combined with a cold shooting night from the Cavaliers could end up with an upset.

In fact, using these numbers, we can find out the percent chance of how often a cold-shooting night from the Cavaliers would result in their loss against an average offensive output for their opponent’s. Running a simple binomial distribution can tell us the chances that a team that shoots *X* percent on shots worth *Y* points in a game that has *Z* shot attempts will end up with any number of points at the end of the game.

In their actual 60-possessions games last year, if their opponents scored their median 56 points, the Cavaliers had roughly a 1.7% chance of going to overtime, and a 2.1% chance of losing outright in regulation, which is a shockingly high number for the rare team that is elite on both ends. If the opponents scored at least 62 points in those 60 possessions, which they had a 23.6% chance of doing on any given night, the chances that the Cavaliers wouldn’t be able to even force overtime is 10.6%.

How would an extra 20 possessions mitigate that risk? Substantially, it turns out.

In a hypothetical 80-possession games that assumes Virginia would still be able to score and defend at the same rate as they did while playing at Florida International’s pace, if their opponents scored their now median 74 points, the Cavaliers had roughly a only a 0.7% chance of going to overtime, and only a 0.9% chance of losing outright in regulation. If the opponents scored at least 82 points in those 80 possessions, which they had a 21.4% chance of doing on any given night, the chances that the Cavaliers wouldn’t be able to even force overtime is only 7.0%. Obviously this example makes a lot of assumptions about both Virginia’s offense and defense, but the point remains the same: If your advantage is reasonably static, increase that cumulative advantage with more trials.

I like to illustrate this concept of advantage using something everybody is familiar with, a 6-sided dice. If I gave you, Eric, a bet that you got to roll a dice for 5 minutes, and every time it came up a 1, 2, 3, or 4, I had to give you a dollar, but every time it came up a 5 or a 6, you had to give me a dollar, how often would you roll the dice in that 5-minute period? To extrapolate it out a bit, if it was a 25-sided dice, and I gave you the numbers 1-13, and I got 14-25, how often would you roll it then? You would still roll it with the same vigor, but obviously you’d understand that sometimes, due to variance, I would actually come out ahead with the 25-sided dice (as well as I would with the 6-sided dice a very small amount of the time), but that doesn’t mean you should switch up your strategy.

Tempo is actually one of my favorite things to talk about in basketball with people, because many people see it as purely a psychological warfare against the opponent to play at an outlier pace, especially when it’s slow. But, efficiency is like a less strong version of basketball, eventually, it will win out over all other forces. One of the smartest men I know who is entirely uninterested in basketball once told me that if he was a coach, he would play with the slowest tempo of whatever league he was in.

Generally, I’ve found that people who argue that slower offenses wear defenses down are looking over the fact that as a defensive player, whenever a team was burning clock intentionally on offense, it actually meant I had to defend less. Those offenses are all artificially filled out with more unnatural, non-threatening action happening all over the perimeter that is never actually going to have a shot result from it. Knowing this was invaluable, as I knew the first 20 seconds in the shot clock, all I had to do was communicate and rotate and I didn’t have to wear myself out chasing red herrings around. Serious offensive actions will invariably have shot attempts result from it at some rate, and if there are long periods of inactivity early in the shot clock from an offense, it means that the offense is not actually stressing out the defense any more than playing fast would, and arguably is stressing the defense considerably less than a fast offense would.

Anyways, as a general rule: Play fast when the favorite, play slow when the underdog for as long as you realistically can.

EF: Generally speaking, what do you think of Florida’s shot selection last year?

JG: There was a little bit of everything, from good, to fine, to bad.

The good was very easy to identify with just a cursory glance at the team’s Synergy page. The Gators had less than 100 possessions used in the post all year, and only 5% of the total possessions the team used came from the post. While a lot of this undoubtedly had to do with the perception that Kevarrius Hayes was not much a post-up threat, there’s no reason to look a gift horse in the mouth, or get away from a good thing now that Hayes has graduated. The fact of the matter with post-ups are that they have a two-fold negating effect to efficient offense:

1. The actual efficiencies on post-up possessions are much, much lower than many would anticipate. Of the 181 players last year to have at least 100 total possessions in the post, only 28 had a PPP above 1, and only 6 of those 28 were Power Five Conference players. Also worth noting, the post-entry is the hardest pass in all of basketball, it has a significantly high turnover rate and often takes longer than any other pass to throw, wasting precious seconds off the shot clock.

2. The passive effect of having a back-to-the-basket big man is that he spends much of his time in the most valuable real estate on the court, right next to the basket, chasing these inefficient possessions. If this player is not a good enough shooter to draw the opposing big out to respect his jumper, this post-up player has effectively pre-collapsed the most important rim defender to the spot that the offense wishes he was at the least.

Also, this might surprise some people, but by the time that I got to college, knee injuries had sapped me of the little athleticism I had, so I was forced to play center almost exclusively, and I had a lot of back-to-the-basket moves and posting up was a big part of my game. My advice to all big men has been for years now to invest in their shooting, pick-and-diving, and catch-and-finishing instead of their back-to-the-basket games for these reasons. All three of those things are both actively and passively more useful for teams in the current basketball ecosystem than post-ups.

As far as the fine, Florida was almost dead average in the proportionality of their jump-shot possessions that ended with mid-range jumpers instead of a three-pointer. So while I’m not going to bend over backwards to congratulate them when they used 230 possessions on those jumpers which only returned 152 points, there are more important things to focus on. In a perfect world, a team would take close to zero mid-ranges, but the aforementioned St. Thomas team I worked with took ~5 mid-ranges a game, while Florida took ~6.5. The biggest difference in shot selection with the two teams was that St. Thomas had a historically great (for level of competition, obviously) PPP 1.03 on those shots, while Florida only mustered a PPP of 0.66. To put that in terms of three-point shooting, St. Thomas would only be right to pass up a mid-range if there was a way to manufacture a three-pointer of at least 35%. With Florida, they only needed to manufacture a three-pointer of 23% or better, so even a three-pointer from Dontay Bassett was a better, more efficient option than a Gator mid-range. From experience, it would be hard to draw up an action that couldn’t create at least a 25-30% three-point attempt even if the shot clock was at 3 seconds remaining. At least the amount of times they were taking these shots wasn’t a big concern, but still, the Gators can always do better on this front. Every team can.

As far as bad, the overall efficiency of the team offensively was not great, and the culprit was the guard play during pick and roll. As a team, whenever the pick and roll guard called their own number, the team’s PPP was 0.67, about the exact same as a mid-range jumper. The problem was that between Andrew Nembhard, KeVaughn Allen, and Jalen Hudson had 302 tracked individual possessions running pick and rolls, with 222 field goal attempts coming between the three of them. With such a high volume of low-efficiency shots, this was just pure negligence offensively.

As far as the main culprits go, Andrew Nembhard was only a Freshman, and clearly was making a strong effort to get other players involved in PnR actions. Of the 399 tracked possessions of PnR’s he ran last year, exactly 2/3rds of them involved him passing the ball. Among the other 133 possessions, he did only attempt 93 field goals, but I still cannot possibly ignore that his PPP was 0.58 on those individual PnR possessions, which ranked 205th out of 209 D1 players with at least 133 possessions. Coupled with the fact that only 22 of his 354 total individual possessions last year came with 4 or fewer seconds on the shot clock, he clearly needs to make better decisions with his shot selection in his remaining three college seasons. I don’t think it would be fair of me to really get after Andrew for these numbers because, like I said, he was only a Freshman and he wasn’t out here just chucking like Larry Hughes, but that doesn’t change the simple fact that they need to improve if this team wants to win games, or Andrew wants to continue playing basketball after Florida.

Not helping matters were KeVaughn Allen and Jalen Hudson combining to attempt 129 field goals when running a pick and roll, yet only managed to hit the roll man on 23 tracked possessions. Even with Steph Curry, James Harden, and Chris Paul picking apart NBA defenses with their pick and roll mastery, those three players all understand that hitting the roll man is the best threat they have towards manipulating the defense into giving the offensive team a high-efficiency shot. If a player is almost 6 times as likely to call their own number as they are to hit the pick-setter in a position to do something with the ball, there’s no way the defense will ever respect rolls or pops.

Giving too much freedom to guards running pick and rolls is an institutional problem in Division I basketball, but that only means that reigning them in and having them play more efficiently will provide an even greater payoff.

EF: Jalen Hudson had a major decline from 2017-18 to 2018-19. Looking at the numbers could you have predicted such a decline?

JG: As much as I would like to say that there was something in the data that screamed, “this is unlikely to continue at this rate!” there really isn’t any one thing that I could confidently point to as a signal in all the statistical noise of Jalen Hudson’s career.

The main culprit in his efficiency drop from his Junior to Senior season was obviously going from a 40.4% three-point shooter all the way to a 28.0% three-point shooter. I’m sure cases like this one exist, but frankly, I don’t know any off the top of my head.

The only thing that I’m coming up with is this summer I was watching college film with a friend of mine who after his Junior year was a projected 1st round pick, but went back to school for his Senior year, shot it worse from three, and then went undrafted (don’t worry, he’s had a great overseas career). We were watching some film from his Senior year, and he pointed out that an ankle surgery completely changed his shooting form, leading to the drop-off in performance. Going back to his Junior year’s film, I saw he was completely correct: He wasn’t getting anywhere near the same lift, and to compensate, was cranking the ball from behind his head in a flinging motion like a catapult instead of the smooth stroke he had a Junior. The only reason I bring this up is because the drop in my friend’s three-point percentage due to this clearly frustrating injury that affected his form was… 6%, less than half of Hudson’s drop.

The only thing that leaped out to me looking through Hudson’s numbers was that he shot 45% from three on contested catch-and-shoot jumpers in 2018-19, which dropped to 32% last year. He actually shot considerably better on contested catch-and-shoot threes his Junior year than unguarded ones (37%). This might seem initially like an indicator of variance having a greater impact on the results than actual shooting skill, and I have noticed a few correlations with this phenomenon affecting other aspects of player performance in games, but I’ve found 10-15% of players consistently shoot better contested than uncontested, year-in, year-out. His Sophomore year at Virginia Tech, Hudson was once again significantly better shooting when contested versus uncontested, with 45% to 32% splits, albeit on low attempts.

One of the few things that I can say that I remembered from watching the Gators last year was that Hudson took many questionable shots early to the middle of the shot clock that I was not a fan of. To a point that he was enabled or pressing for shots due to frustration, this could be evidence to either of these things. One of the things that is clear in the numbers/tendencies is that Hudson at no point made a meaningful adjustment to his inputs to try to curb his outputs.

Basketball is an emotional game, and also a social game. Sometimes things happen on the court or off of it that have drastic effects on performance. I don’t know what that could’ve been for Jalen, and I won’t speculate, but this would be an all-time hot streak followed by a cold one if this is just the result of variance. There’s nothing from 2018-19 that I can point to that I could in good faith argue was a signal to what was to come. Crashes always seem so obvious in hindsight, as do bull markets, but the truth is that forecasting is difficult.

EF: Florida got Kerry Blackshear and suddenly have someone who is great at post-ups so I want to talk to you about how you use a player like that….

JG: As you might guess from my earlier answers: Don’t use him exclusively, or the majority, in the post.

Anybody who follows me on Twitter knows that I loved Virginia Tech’s team last year. Though Virginia was my pre-tournament pick to win the National Championship, I had Virginia Tech upsetting Duke in the Sweet Sixteen. I loved the way their entire starting five worked together, maintaining flexibility and efficiency, and to me, the two players who uniquely tied it all together were Nickeil Alexander-Walker and Blackshear. Talking with members of the former Virginia Tech staff that are now at Texas A&M, I know they valued Kerry for many of the same reasons that I did.

Blackshear’s versatility in taking jumpers, finishing around the rim, crashing the boards, diving or popping after setting screens, and his high-level secondary creation ability to pass the ball would have me using him in whichever way most effectively leverages his ability to be flexible and efficient for every single match-up, this of course would have him punishing smaller players in the post for short spurts. In fact, Kerry Blackshear was one of the 6 players last year with over 100 possessions in the post who scored a PPP above 1 (he had 163 possessions and a PPP of 1.04), but to use him only in the post would be such a waste of his abilities. The average PPP in Division 1 basketball is roughly 0.90, and last year Kerry scored at 0.96 or higher in all seven of his most used Play Types tracked by Synergy (Post-Up, Offensive Rebounds, PnR Roll Man, Cuts, Spot Up, Transition, and Isolation).

I would love to see a Kerry Blackshear who exclusively pops and shoot six threes against a plodding big man on a Wednesday night, grabs four offensive boards and scores seven times close to the basket then on Saturday against a smaller, quicker team, then ties it together with five assists against a zone defense on Tuesday. You don’t force Kerry Blackshear Jr. into a box, you gameplan so that Kerry Blackshear Jr. forces the other team’s coaching staff into a box.

One of my rules of thumb are that with bigs, you ride their dominant aspects and incorporate it heavily into every level of an offense. Kerry might not be dominant at any one thing, but his versatility can double as dominance if utilized correctly and enabled by his teammates and coaches.

EF: Is there a returning Gator that you really like from an efficiency standpoint?

JG: Another question with an easy answer: Noah Locke.

It might seem bizarre that I’m picking a player with a FG% of 37.5% on over 8 shots a game as a Freshman for the answer to this question, but pretty much any player who can get off 6 threes a game and still shoot 38% on them is going to curry favor with an analytics guy like myself.

He needs to work on being less one-dimensional though. There’s being averse to penetrating, then there’s only attempting 5 total shots at the rim in the half-court in almost 1,000 minutes played—which is what Noah did his Freshman year. Don’t get me wrong, I’m happy he’s focusing on his strengths and that’s shooting the ball, but he needs to be a threat once he’s over the three-point line, and currently, the best thing that a defense can do on any given possession is run Locke off the line.

EF: Florida’s defense has been it’s calling card over the last few years under Mike White and last season saw the team be quite effective defensively by some of the traditional metrics as well as KenPom. How good do you think Florida was defensively looking at the numbers and did anything jump out at you?

JG: Forecasting analytically for defenses is generally harder than offense, due to a few aspects of how control over situations varies when a team is playing on offense versus playing on defense.

I think a good place to start is by noting that when adjusting for pace of play and strength of opponent, under Mike White the Gators have consistently been good but not great, save for the 2016-17 season. While always in the top 25, the only year inside the top 10 (#5), the Gators went all the way to the Elite Eight. A slow pace of play can make a good defense appear great, though defense only truly become elite when they can defend regardless of pace. As mentioned earlier, a faster pace of play favors the superior team, so a good defensive team that can only exist in a slower system is a bit like owning a supercar with an engine that will only allow you to hit 60 MPH, and nothing faster.

As far as forecasting for next year, there are few things that are obvious. First, the Gators are going to miss KeVaughn Allen immensely. He’s not going to be a player easily replaced, and he was one of the few players with a notable and measurable team defensive impact when he’s on the floor. The great news is that Andrew Nembhard has already emerged a player who is a fantastic defender on a high volume of defensive possessions. Second, Kevarrius Hayes has also moved on, but unfortunately there is no ideal replacement that has already made his presence known from within the program.

As far as broader predictions, that’s nothing I would feel confident signing my name to as far as a forecast goes. Even when doing my consulting with Mercenary Analytics, it takes me about 20 hours or so of work and an 8 to 10 game sample size before I’m confident with my advisings when it goes to defense. Defense involves the weighing of so many different variables and is so prone to being susceptible to emotional biases that the “loudest” players defensively, the ones who leap off the screen with their movement and hustle, can sometimes be the worst defenders in five-man units. To give an example, the defender who falls asleep twice a game and leaves an elite catch-and-shoot player wide open is probably far more destructive defensively than the player who consistently gives ground but isn’t completely burnt on closeouts. But, one is far more noticeable on film and far more unlikely to accumulative steals than the other.

Without solid data to work off of, the amount of film and tracking I would need to invest before I was confident with a prediction would be large. So, based on the loss of two key players, and only one who figures to be more or less seamlessly replaced, the only logical forecast would be that the Gators should be good but not great defensively given White’s track record.

EF: One area where Florida has struggled the last two years is when it comes to rebounding, particularly defensively. Are there any numbers you like to look at when it comes to rebounding from either a team or an individual player standpoint?

JG: Oh man, you’ve really buried the lead and you’re really setting me up to look like a lunatic with this question buried so late in this piece, Eric.

Well, here’s my professional take: I don’t care about almost any rebounding numbers.

The only two that I find any amount of value worth acting on are defensive and offensive team rebound percentages as part of five-man lineups.

Before we get off course and I turn this into a 8,000-word TED Talk on the reasons why I hate individual rebounds as a stat (and there are many), let’s see just how poor Florida is on the defensive glass in absolute terms. By team defensive rebounding percentage, they were 309th out of 353 tracked teams in Division I last year. That sounds very bad, and relatively speaking obviously it is bad, but how does this affect the final score?

Another, and more practical, way of conceptualizing this is in absolute terms of just how much this number hurt Florida against the average Division I team. Florida rebounded 69.85% of their opponent’s missed field goals last year. Comparing that number against North Carolina’s mark of 78.90%, and you would think that the Gators just hemorrhage points on the glass. But the Tar Heels had the second highest defensive rebound percentage of all P5 teams, so that’s not exactly a far measuring stick to compare them against. A better metric would be the media defensive rebounding number, which was Virginia Tech’s 74.09% last year, so the Gators rebounded 4.24% fewer of their opponent’s missed shots than average. A mark of 4.24% means that Florida missed out on a defensive rebound on one out of roughly every 24 misses than an average defensive team.

Let’s even take it one step further: We already know that the Gators played 36 games last year, grabbed 767 defensive boards, and conceded 331 offensive boards. So, Florida’s defensive rebounding opportunities per game was 30.5, of which 21.3 went to the Gators. Considering the average team in Division I had a defensive rebounding rate of 73.8%, the average team would’ve grabbed 22.5 of those same 30.5 rebounds.

With Synergy, we can even take it one step deeper. On the year, Florida’s opponents had 177 possessions tracked as immediately put back up after an offensive rebound, and scored 179 points. On the other 152 offensive rebounds the Gators gave up, we can assume that the opponents had a normal offensive possession following the rebound, which after factoring out offensive rebound possessions, had a points per possession of 0.827 on the season. So, on the 331 extra possessions Florida gave up, we can make an estimation that they conceded 304.7 points, which comes out to a total expected points per possession of 0.92.

How are these numbers valuable? By combining the expected points per possession on offensive rebounds of 0.92 multiplied by the marginal offensive rebounds the Gators gave up when compared to the average team of 1.2, we can expect that the Gators gave up a whopping total of… 1.1 more points per game defensively than the average team due directly to the offensive rebounds they gave up.

Staggering. I hope everybody was sitting down when they saw the extent of the defensive devastation.

At the core of my basketball philosophy is the belief that basketball is a game of markets, and therefore also opportunity costs. Every decision made or not made has an opportunity cost built into it, and there is no decision that can be made in basketball that offers only pure surplus value and no drawbacks. Even when adding one of the top players to a team, like a LeBron James, Steph Curry, or Giannis Antetokounmpo, there is some sacrifice being made to facilitate those players being on the floor. In this example, it’s not unrealistic to expect that with LeBron you’re sacrificing offensive scheme flexibility to build a system around James, with Curry the perimeter defense of a team is likely to take a step back, and with Giannis you’re probably losing some amount of spacing—whether or not the opposing teams can take advantage of these drawbacks is obviously the most important aspect of these tradeoffs, but whether any given team is able or unable to exploit a drawback does not mean that the drawback doesn’t exist.

When making a decision to build five-man units, a whole slew of variables are either knowingly or unknowingly placed into a hierarchy by the coaching staff. In the coaching staffs I work with, I like to unpack this hierarchy. At the top of my hierarchies are the two efficiencies, offensive and defensive. If we sacrifice a little bit of free throw shooting or defensive rebounding to bolster our efficiencies, it’s better than bolstering our free throw shooting or defensive rebounding at the expense of making our team worse on the scoreboard.

To make it more tangible, in an attempt to be better on the glass, how much worse would the defense have to be to lose more than 1.1 points per game, or how much worse would the offense have to be to lose more than the same 1.1 points per game? The margins of basketball are so small that those changes would be basically undetectable visually when watching the game. Or, in the worst case scenario, in an effort to grab more rebounds, the offense loses 4 points per game due to diminished spacing (a number that would be conversative when comparing the offensive ratings of the same 4-man groupings with a stretch big versus a non-stretching big), and also loses 1.5 points directly on defense when the rebounding big jumps at a pump fake and fouls a 75% free throw shooter when a non-rebounding big wouldn’t have. So, our team is now 5.5 points per game worse because we prioritized rebounding over winning, and we don’t even know it.

Mike White has run a consistently very good and occasionally great defense. The opportunity cost of this is that the defense allows 1.1 more points than the average team on opponent’s offensive rebounds, which just happen to look very “loud,” happen at inopportune times (because, obviously, there’s never a good time to give up offensive rebounds) and are very frustrating when watching the team.

My advice to fans who are bothered by the offensive rebounds given up would be to remember that every team in basketball gives up offensive rebounds, and Florida only gives up roughly 1 more offensive rebound a game than everybody else. So get that frustration out of the way early, every game when the Gators give up their first offensive rebound, get mad because that’s the “extra” rebound that Florida has given up for the night, then settle in and enjoy the rest of the game as normal.

You can follow Joseph on Twitter at @JosephGillMA where you’ll get one of the best basketball follows on the entire app.

Eric Fawcett
Eric hails from Edmonton, Alberta, Canada. His blend of sports and comedy has landed his words on ESPN, Bleacher Report, CBS Sports, Lindy's and others. He loves zone defenses, the extra pass, and a 30 second shot clock. Growing up in Canada, an American channel showing SEC basketball games was his first exposure to Gator hoops, and he has been hooked ever since. You can follow him on Twitter at @Efawcett7.