NBA Week 4 in Review: Small Sample Size Theater Is Winding Down

As we move further and further into the season, the odd kinks at the beginning of the season start to solidify and the patterns are emerging. For some fans, like in Portland and Toronto, this is encouraging, but for others in Cleveland or Charlotte this is distressing. Yet, as everyone has said repeatedly in November, the sample sizes are too small and we should still be a bit cautious, looking to past seasons for more information.

Klay Thompson’s Ascendance

A funny thing happened after a summer of criticism for valuing Klay Thompson more than Kevin Love: Klay has played remarkably better and Golden State looks like the wise club. But rather than retroactively giving the franchise a bump for the decision, we should take a step back and look at the big picture. The past two seasons his usage rate has been 21.8 and 22.6, respectively, with a TS% of 53.3 and 55.5 while assisting on a little over 10% of his teammates’ field goals. That is essentially a slightly above average player on offense, ignoring his spacing, and he was a blackhole, passing and assisting too little for how much he had the ball.

Yet now Klay has an outstanding shooting efficiency of 60.5 (TS%) with a usage rate of 27.4 — he’s shooting more than ever and hitting more of his shots by a huge margin. However, look at his gamelogs for the past season for the first 19 games. He was about as efficient as he is now with a TS% of 61.3 and his usage rate was a little higher at 23. Looking at his shooting splits this season, and you see a shooting percentage of 81.3% inside of 3 feet and 45% from behind the line — the latter will probably come down a little but the former is completely unsustainable and something even young Shaq could hardly manage. The question here is if we’re looking at legitimate improvement or not, but the truth is somewhere in between. He is not *this* good, but his free throws have exponentially increased — perhaps thanks to stealing some of Harden’s dark arts during the summer — and his assists have nearly doubled. He has improved, but the level to which he’s improved is unclear and we’ll have to wait until the dust has settled. If I had to guess, I’d project that he’ll be an Eddie Jones type player who can score pretty well but isn’t a good primary playmaker/scorer yet can hit a lot of three’s and defend at a high level.

The philosophy and fallout from the Love/Klay discussions have a lot to do with an uneasy reliance on probability — something conventional managers and analysts like to eschew. From the information we had last season, Kevin Love was the right choice. If things change and one player is better suddenly, that doesn’t make the choice a poor one. Weird things happen. It’s the NBA. In 2011, if one had tried to trade Conley for Rose, one would have been laughed out of the room. Obviously, we’re only a few games into the season and shouldn’t judge them solely on this season so far, but it’s important to keep in mind in future discussions. Nailing down a trade in the NBA is like trying to hit a moving target, and if that target suddenly accelerates, well, too bad — it’s a tough league.

You can see Thompson use a Harden-esque move by attacking the defender at the basket by tangling their arms together as he shoots in the video below.

Tristan’s Potion

Tristan Thompson currently has a higher rebound rate on offense than defense, which is quite rare: it’s easier to grab defensive rebounds. In fact, the list of players who have done this with at least 1000 minutes and a defensive rebound rate of 10 is extremely short. There’s George Lynch in 1994, Maxiell in 2009, Pekovic in 2012, Mark Madsen in 2004, Michael Ansley in 1990, and, of course, Dennis Rodman in 1987. But no guy played more than 1762 minutes, and no guy had a higher offensive rebound rate: 16. There are few things going well for Cleveland now, but his activity on the glass is a delight. For whatever reason, Thompson is now jumping quickly and looks more effectively athletic. He may have taken a magic potion or he may have undergone a new training program, but it’s working for now, whatever it is.

You can see him grab an offensive rebound against Denver surrounded by four players and then tap it again right in the middle of the swarm.

Philadelphia’s Wretched Offense

While all the talk of their team is about how long they’ll go without a win, few have taken the time to notice how decent their defense is. As odd as this sounds, they have nearly the same defensive rating, per b-ref, as the New Orleans Pelicans with Anthony Davis. They’re doing poorly in shot defense, but they’re creating the most turnovers in the league, per possession, and rebound near the league average rate just like highly regarded teams like the Warriors and Grizzlies. The team wanted to emphasize defense, and perhaps it shouldn’t be a surprise they’re doing so well in that respect — they brought in Mbah a Moute and have Nerlens Noel along with a bunch of athletes to collect steals.

Yet if you’re on track to be one of the worst teams ever and your defense is competent, that means your offense must be ugly — and it is. Using basketball-reference’s rating system, they’re 15.2 points under the league average, which is completely unprecedented. In fact, the two worst net ratings (difference of offensive rating and defensive) in the shot clock era are actually 15.2 as well from the 1993 Mavericks and 2012 Bobcats. So if your offensive rating is 15.2 by itself, something is terribly, terribly wrong.

Relative offensive rating

Fear the Deer’s Defense?

Last season, the Bucks sported the worst defense in the league, and lo and behold, they’ve been above average so far this season. The recovery of Larry Sanders is certainly one cause, but even when he’s been on the bench the team has still defended better. As I mentioned in the team preview, Knight had a down year on defense plus you have a handful of young guys who will only naturally improve like Giannis. They might be getting lucky in a few respects, however. They’re creating a lot of turnovers and that may not hold. Opponents are shooting a low percentage on three’s — that might be unsustainable as well. Also, opponents are shooting only 71.3% on free throws. If their opponents were at the league average, they’d be giving up roughly another point per game.

But we can resurrect the Larry Sanders! love. They’re holding opponents to 100.4 points per 100 possessions when he’s on the floor. He doesn’t have quite the same eye-popping stats — his blocks are down and opponents are shooting 46.7% near the rim when he’s there too, which is good but not extraordinary — but he’s still an effective interior defender for a team that needs it. In the video below, he swats a Pierce attempt at a layup out of bounds and then comes out of the weakside to stuff Pierce again at the rim, thwarting a dunk.

An Old Dog Learning a New Trick

While what’s going on in Oklahoma City is a nightmare, it at least affords them the opportunity to develop their peripheral offensive players against NBA competition. The team’s experimenting with zones, for instance, and guys are trying new things. Nick Collison is 34 years-old and has nearly doubled his career totals for three point attempts already: he’s shooting two a game with a conversion rate of 37%. The consequences here are refreshing. When Durant and Westbrook come back, the Thunder can trot out a lineup where every player can hit a shot behind the arc with two plus defenders in the frontcourt. That’s a rare weapon, and the Thunder could be the most dangerous 8 seed ever. Collison’s three-point shot looks legitimate too; you can see it in the video below.

We Have Hockey Assists … Now What?

Before the SportVU era, one of the most requested statistics was the hockey assist or as it’s known now the secondary assist. It was about highlighting players who created a shot opportunity by throwing the pass that led to the assist and we all wondered what the results were and what they’d tell us.

Unfortunately, what’s been released publicly is disappointing and few people have even attempted to decode the numbers and use them. The leaders in secondary assists are, surprise, the leaders in total assists and starters average anywhere from 0 to 2 a game — hardly earth-shattering data. Yet can’t we do something better, mine some truth from this new type of assist to find some sort of hidden value?

I collected all the box score player tracking data over the weekend and have secondary assists totals, so I thought I’d play around with them and try to extract some sort of meaning. After playing around with the numbers and trying different functional forms, there’s a clear, simple relationship with secondary assists: they decrease per minute or possession as a player’s assists go up. On some level, this makes sense because the more you have the ball and the more you assist, the fewer opportunities you have to collect them. But it’s also an indictment of secondary assists because the guys with the highest ratios of secondary assist to assists are seemingly the least talented playmakers like DeAndre Jordan and Biyombo. You can see how the relationship is linear in the plot below.

secondary assists 2014

I listed the names of a few outliers in that graph. The guys on top get more secondary assists than expected, so perhaps their normal assist numbers are underrating them, while the guys on the bottom are perhaps “assist-padding” their totals. It’s just unclear how much of this is noise.

Testing the value of secondary assists, I threw it into the mix of a statistical plus/minus model built on 14 years of data and tried to find significance after even cross-validation. Basically, I’m trying to filter out the noise, and there are some encouraging signs though I’d caution this is with only one year of data. Secondary assists did appear to correlate with offensive value and there might be even more promise with interaction variables and ratios. But even plain old secondary assists per minute or possessions did the trick. Again, this isn’t perfect with one year, but there might be something here and hopefully it can tell us about players who don’t “hunt” for assists and who help ball movement, not stall it.

Though how can we be sure secondary assist are actually mostly skill and not luck or just a fluke in the stats? So far, there is a decent year-to-year correlation between the ratio of secondary assists to normal assists among the top 40 guys in assists per game with at least 10 games (the R is around 0.6.) A lot of that, of course, is about how the relationship between those types of assists has a lot to do with how many assists you’re getting — you have a higher ratio when you don’t collect a lot of assists. What these hockey assists mean could be unlocked by the end of the season once we have two seasons of data — and this elusive stat could finally prove worthwhile.

Quantcast