Stats Guide

Player: On/Off Stats

The Player On/Off section shows all of CTG's team stats broken down by whether the player is on the court, off the court, or what position they're playing. The default view shows you only the off-court stats, but on-court and position stats can be seen by using the links above any table). It's a complicated page, but a very powerful one, since over time you can pick up trends in how a team tends to perform with a player in the game vs. out of it.

This guide tells you how to think about these tables using the first table, Team Efficiency and Four Factors, as an example. For specific explanations of each stat in any of the tables, please see the appropriate guide:

Off Courtclick to show/hide

What we're interested in when looking at off-court stats is the difference in how the team performed with the player on the court compared to off. So, to make it easier to interpret, these stats are shown relative to what the on-court stats show: a +5.2 in the Offense Pts/Poss column, for example, means that the team scored 5.2 points per 100 possessions more with the player on the court than when they were off. That's good - it means the team was better on offense when the player was on the court. The percentile shown is based on this difference. So a 90 for an off court stat means that 90% of players had a worse on/off court differential in this category.

For example, in the Efficiency and Four Factors table, the expected wins (Exp W) stat will tell us the difference in expected wins with a player on vs. off the court. Let's say a player has a 45 for on court expected wins and a +10 for off court. That means when the player was on the court his team played at the level of a 45 win team, and when he was off the court they played like a 35 win team.

Example: Robin Lopez

The 2016-17 Bulls played a little bit worse when Lopez was on the bench than when he was in the game, with an efficiency differential that was 2.5 points better with Lopez in the game. That number ranked on the 68th percentile, meaning it was a pretty good number, but not great.

Look at the ORB% column for offense. We see this past year that the Bulls' offensive rebound rate went up by 3.3 percentage points with Lopez in the game relative to when he sat. That was a big change: 87% of players had a worse differential. And if we go back through the years we find that this is a pattern with Lopez. The year before, in New York, the differential was on the 93rd percentile. The prior 4 years? 64th, 53rd, 83rd and 93rd. There seems to be a pattern there. The fact that there's a multi-year track record across multiple franchises and that this conclusion is supported by Lopez's good individual offensive rebounding rates gives us strong confidence that Lopez is causing his team to get many more offensive rebounds.

Similarly, we can look for streaks of red and blue in this table to see if there is a consistent effect Lopez is having. His offensive rebounding impact might not be surprising, but his defensive rebounding numbers might be. See that bright blue streak in the ORB% column on the defense side of the table? That tells us that over the last 5 seasons Lopez's teams have seen their defensive rebounding take a hit (sometimes a major one) when he's sat.

This is surprising and important because Lopez's individual defensive rebounding rates are abysmal for a big man: he has never had a season rating above the 25th percentile for big men in defensive rebounding percentage. But these numbers show us something different: that his teams consistently defensive rebound well when he plays, and that they defensive rebound worse when he sits. Watching Lopez reveals that he is very good at boxing out opponents, and so even if he doesn't get the rebound himself he will often make sure his team gets the rebound. This might not have been as apparent (or its value could be questioned) without seeing a multi-year streak of Lopez's positive impact on his team's defensive rebounding.

The Gritty Details

Off court stats are included only for games in which the player was on the roster. That means if a player is traded mid-season, the rest of the season's stats for that team are not included in his off-court stats. As with on court stats, players with under 100 minutes are not included in the percentile calculations.

On Courtclick to show/hide

These stats tells us specifically how the team performed when a player was on the court. To give us context for how good any of these stats are, the percentiles compare this number to all other players. So let's say a player's team scored 109.0 points per 100 possessions when they were on the court, which ranked on the 75th percentile. That means that 75% of players' teams scored less than 109.0 points per 100 possessions when those players were on the floor.

Example: Robin Lopez

This past year, the Bulls had a +1.4 efficiency differential with Robin Lopez on the court. That number was better than 68% of other players. Of course that was likely heavily influenced by his teammates. And we can see this in the offensive four factors numbers: with Lopez on the court, the Bulls were a poor shooting team (15th percentile eFG%) but a very good offensive rebounding team (93rd percentile) — which was the general profile of the Bulls that year, as they ranked 30th in eFG% and 4th in ORB%.

By looking through the other years in each of these categories we can see that the low eFG% isn't a general property of teams when Lopez has been playing — but the high ORB% mostly is. This gives us some confidence that Lopez has something to do with the good offensive rebounding of his team but probably isn't the cause of the low eFG%.

The Gritty Details

Players with under 100 minutes are not included in the percentile calculations.

Position Statsclick to show/hide

The deepest layer of these stats is the position stats. The stats and percentiles shown in the position rows can tell us how the team's play changed based on the position the player was playing. The minutes stats for these rows tells us what percentage of a player's minutes were spent at each position.

It's important to be careful when drawing conclusions from these stats. It's possible that the team plays differently when the player plays various positions because another important player is impacting the stats. For example, the 2016-17 Trail Blazers played much worse with CJ McCollum at PG than at SG — but if McCollum was at PG it means Damian Lillard was out of the game, so it's possible those stats are more reflective of Lillard being on the bench than of McCollum's PG play.

Example: Kristaps Porzingis

Kristaps Porzingis provides an interesting example of interpreting these position stats. In each of his two seasons he has spent about 80% of the time at power forward and about 20% of the time at center. We can see by looking at the "Diff" column that his team's overall play hasn't really changed much when he's been at PF vs. C: in his rookie year, 2015, their efficiency differential was +0.2 with Porzingis at PF and +0.4 with him at C. This past season, 2016, the Knicks played slightly better with Porzingis at center: an efficiency differential of -2.5, corresponding to a 34 win team with him at PF, and a differential of 0, corresponding to a 41 win team, with him at C.

But the Knicks' style of play did change. In 2015 their offensive Pts/Poss was on the 39th percentile (104.7) with Porzingis at PF and 84th percentile (111.2) with him at C. In 2016 we saw a similar change: from 37th percentile to 73rd. The Knicks' offense, in other words, got a lot better when Porzingis played center compared to when he played power forward. This makes sense: putting a three-point threat at the 5 position really spaces the floor.

We can feel a little more confident that this is a consistent effect because not only did the offense go up both seasons, it did so in the same way: offensive rebounding went way down, but this was more than canceled out by the fact that turnovers went way down as well, and free throw scoring went way up.

But the reason why the team had a similar differential with Porzingis at either position is because the defense also fell off a lot both years: defensive rebounding fell and opponent free throw rates went way up.

(The fact that both sides' free throws went way up might be due to increased spacing, but it also might be due to the Knicks using Porzingis at center more later in games, when it's more likely to be in the bonus and both teams are more likely to foul. It's also possible that the centers Porzingis played with were impacting these numbers: both Robin Lopez and Joakim Noah, for example, have consistent track records of lowering both opposing and their own team's free throws. These are examples of how you have to be careful interpreting these stats — seeing the changes is different than knowing what is causing them.)

The Gritty Details

Position stats are determined by looking at each 5-man unit the team has played and ordering that lineup in the way that it seems most likely we'd assign the 5 positions: point guard, shooting guard, small forward, power forward, and center. This is a very difficult exercise. Some players we might think of differently on offense or defense (in some lineups Giannis Antetokounmpo, for example, might play point guard on offense but power forward on defense). Some players we could have reasonable disagreements about what position they're playing. This lineup ordering is done with a lot of manual input. Ultimately, then, the position stats should be interpreted with caution.