The hot topic of the last week or so has been when Nate McMillan will finally pull the trigger and start Nicolas Batum. Dwight Jaynes summed up the arguments for Batum nicely. Since Batum's first game back I've been pushing for McMillan to ramp up Batum's minutes as quickly as possible and since last weekend Batum's play has made not starting him seem like a criminal act.
Of course, there are a few extenuating circumstances to consider here that make this decision more difficult for McMillan than it might appear.
First, Brandon Roy's health obviously overshadows this entire discussion. It's been touch and go for the last week or two. McMillan has already been juggling his lineups like crazy this season. Any consistency he can find in a starting lineup will be valued more than it might usually. If he decided to start Batum now then he would have a second lineup change to deal with once Roy returns. Perhaps he would prefer to make those switches together.
Second, applying the brakes on Batum to ensure he doesn't feel pain in his recently repaired shoulder is a consideration. When push comes to shove down the stretch, McMIllan has shown no hesitancy in playing Batum for entire fourth quarters at a time. But if he can save five minutes here, five minutes there at this point in the season and Batum's recovery that's a good thing.
Third, and perhaps most important, there's the role and development of Martell Webster. For nearly his entire career, Webster has left the coaching staff and fans scratching their heads, not knowing what he'll produce on any given night. That started to change this January when Webster strung together the best month of his career, stepping up in Roy's absence and getting red hot from the three point line. McMillan relied on Webster so heavily that Webster actually played every single minute of the second half of a back to back. As I noted at the time, that's a pretty big deal.
The question facing Nate McMillan: How do I reincorporate a ridiculously promising Nicolas Batum into my rotation while still hanging on to as much of the career-best production I've been receiving from Martell Webster? While sliding Webster to the 2 might work in the immediate short-term, Roy will be coming back sooner rather than later. While fitting too many productive players into a rotation is a relatively good problem to have, it's not necessarily an easy one to solve. Certainly Webster is going to take a smaller share of the minutes but exactly how much smaller should that share be?
To start answering that question, it's important to get a sense for what situations have led Webster to play his most productive basketball this season and also to determine whether/when his personal success has translated to team success.
Click through for an analysis of Martell Webster's production based on his minutes played and how that might explain Nate McMillan's current decision-making process.
-- Ben Golliver | benjamin.golliver@gmail.com | Twitter
To address these questions, I turned to Basketball-Reference's game logs for Webster this season. I split his 50 games into 5 categories based on minutes played. Those categories....
- Less than 20 minutes played: 10 games
- 21-25 minutes played: 12 games
- 26-32 minutes played: 10 games
- 32-37 minutes played: 9 games
- 38 and more minutes played: 9 games
Note: these categories breakdowns were not 5 clean groups of 10 games because there were a large clusters of games around the 20-22 and 25-27 minute mark.
What jumps out from those categories first is a pretty striking inconsistency in playing time for Webster. This shouldn't really come as a surprise. He's been yanked up and down this season as guys have been injured and then returned from injury. Credit to Webster for maintaining a solid work ethic and team-first attitude throughout.
Now let's take a deeper look at his statistical production when broken down into these categories. The goal is to answer the question, "How many minutes have been ideal for maximizing Martell Webster's production?"
A note on the graphs below. I divvied up Webster's game logs into the 5 categories listed above and looked at Webster's points, rebounds, assists, steals and blocks. As Webster's contributions come on both ends of the court, I thought it worthwhile to look at all of these indicators.
I averaged his production within each of the 5 categories and then standardized that production to "Per 48" minute numbers so that we can easily compare how productive he is when he plays lesser minutes to how productive he is when he plays larger minutes. Obviously, he's likely to score more points (or grab more rebounds, or block more shots) when he plays 35 minutes instead of 20 minutes. But is he scoring more points per minute when he plays more minutes or does his production taper off during games where he stays on the court longer?
Let's look first at Webster's scoring.
On the X axis you'll see the 5 categories I've created based on how many minutes Webster played. On the Y axis you'll see his average points per 48 minutes in each category. As you can see, in games that Webster has played less than 20 minutes, he is averaging less than 10 points per 48 minutes. By comparison, in games that Webster played more than 38 minutes, he is averaging more than 20 points per 48 minutes.
Although it's not a perfectly linear relationship, generally speaking, the more minutes that Webster has played the more efficiently he has produced points. It's not just that he's scored more points while playing more minutes, it's that he's scored more points per minute too.
We have a chicken and egg dilemma when interpreting that information. Are the increased points causing the increased minutes or are the increased minutes causing the increased points? On the other side of it, is he playing less minutes because McMillan senses he isn't "on" that night or is he not playing enough minutes to get himself going?
There's not a really solid answer to those questions but the range of variability in his scoring rate is pretty big, much larger than any of the other statistics we'll look at below. The difference between 20 points per 48 minutes and 10 points per 48 minutes is huge. The high end portion of these numbers reflects an increased scoring burden in Roy's absence, McMillan's tendency to stick with Webster when his outside shot is falling and Webster's ability to catch fire. The low end portion of these numbers suggests Webster is a slow-starter and/or prone to enduring some serious stinkers.
What's most interesting about this graph is that you might have expected bench wing shooters to see the reverse trend. When they play few minutes and come in to quickly knock down a few shots, their points per 48 minutes would be sky high. When they are forced to play heavier minutes and are asked to provide more than just instant offense, their points per 48 minutes might taper off. Not so in Webster's case. Indeed, it seems like he needs a relatively significant chunk of minutes (more than 20) to get things rolling.
Here's a scatterplot that helps reinforce this point.
(Thanks to The Invisible Ninja, who you can follow on Twitter here, for the data points to make this scatterplot.)
Each dot represents one of Webster's games this season. On the X axis, you have Webster's minutes played in each game. On the Y axis, you have Webster's Points Per Possession in that game. PPP is a nice baseline measurement of a player's offensive efficiency; it takes into account points scored, fouls drawn, free throws made and turnovers committed. Anything over 1.0 PPP is considered a very good game. You'll see by the trendline and the arrangement of dots that there is a decently strong correlation between Webster's playing time and his scoring efficiency. He plays more, he scores more efficiently.
But you'll also notice the confirmation that when Webster plays his fewest minutes he has his worst games from an efficiency standpoint. He doesn't have a single game yet this season where he played less than 20 minutes and produced a PPP greater than 1.0.
Now, let's compare that to Webster's rebounding.
Again, the X axis represents Webster's playing time and the Y axis represents his average rebounds per 48 minutes in games played in each of these categories. Look at the range here: at the low end he's averaging 6 rebounds per 48 minutes while on the high end he's averaging 8 rebounds per 48 minutes. That's a difference, sure, but not a gigantic one and not nearly as big as the range in his scoring that we discussed above. Just look how much more stable the overall graph is.
While Webster is not an outstanding rebounder at any time, he's solid on a per-minute no matter how many minutes he plays. That's exactly what you want from a role player.
Also note on this graph that his performance hits a high-water mark in the middle and then falls off slightly when he plays heavier minutes. This is to be expected when he has been tasked with playing heavy minutes guarding an opponent's star player and when his offensive responsibilities have increased in Roy's absence.
Now let's look at Webster's passing.
The X axis represents Webster's playing time and the Y axis represents his average assists per 48 minutes in games played in each of these categories. Again, the range here isn't huge because Webster isn't really known as or expected to be a big-time play maker.
This chart shows that Webster averages between 1 and 2 assists per 48 minutes (roughly) no matter how many minutes he plays with one exception. If he plays less than 20 minutes he's more likely than not to fail to record an assist. But his most productive passing games are those when he plays 21-25 minutes a night. These games would seem to find Webster in a supporting role and playing stretches alongside Brandon Roy and/or LaMarcus Aldridge (which helps any Blazers' assist numbers).
Let's turn to Webster's defense.
Here I combined Webster's steals and blocks into a single chart. The X axis represents Webster's playing time and the Y axis represents his combined steals and blocks per 48 minutes in games played in each of these categories. Again, the range here is amazingly small: no matter how much time he's playing, Webster contributes between roughly 1.5 and 2.5 combined steals and blocks per 48 minutes.
Interestingly, his per-minute defensive statistical production is at its highest when Webster plays fewer minutes and tails off a little bit as his minutes increase. This indicates that, unlike his scoring, Webster is able to make an instant impact defensively right off the bat. You definitely like to see this from a role player.
Now, let's put it all together into a crude aggregate production statistic.
In this graph, I've simply added together Webster's average points, rebounds, assists, steals, and blocks in each of the bands. You can see that, generally speaking, Webster's per-minute production increases as his playing time increases. So not only does he produce more when he plays more but he produces more efficiently when he plays more. This chart is probably influenced too strongly by his scoring numbers but his rebounding and defensive production soften it a little bit.
In any case, increased production resulting from increased minutes for a role player is actually exactly what we should have expected. Our dear friend Kevin Pelton has explored this phenomenon before.
Webster's Minutes and Blazers Winning
Webster doesn't exist in a vacuum, though. Instead, he is part of a five man unit and a 12 person team (if they're lucky). So do the Blazers perform as well as Webster when Webster's minutes increase?
Not exactly. Let's take a look at the Blazers' winning percentage in each of the 5 bands.
Interestingly, the Blazers have won their highest percentage of games when Webster has played 21-25 minutes per game. They've falled off sharply when they've had to rely on Webster more. This is almost certainly due in large part to the absence of Roy. Interestingly, though, in Webster's truly big minute nights -- 38 or more -- the Blazers fared incredibly well. Credit Webster where credit is due here.
With Batum's return, though, there simply are not unlimited minutes available for Martell Webster. The days of Webster playing 38 minutes are almost certainly behind us. Thanks to Batum's injury, there isn't a large enough sample size to isolate the exact point Batum's production outpaces Webster's but if we are approaching this like McMillan we are trying to find a role for Webster between, say, 10 and 30 minutes a night.
What we are able to notice, though, is that Webster's offensive production (especially scoring and assisting) has dropped off the table pretty quickly if he doesn't play at least 20 minutes a night. When it comes to rebounding and defense, he's been more consistent no matter how many minutes he plays. Taken together, the per 48 graphs above, according to Pelton, "imply (but not certainly) that you're going to get a worse Webster when you cut his minutes." This is a key consideration and concern for McMillan.
While I still personally believe that Nicolas Batum should be starting immediately and that a decrease in offensive production from Webster (even a large decrease) is acceptable collateral damage, this data should help provide some perspective into McMillan's decision-making process. An inconsistent Webster is useless to him. A consistently productive Webster is a huge asset. A consistently productive Webster and the emerging Nicolas Batum is practically a dream world. For the time being, I think McMIllan is happy to take the heat for not starting Batum if it means Webster sees that sweet spot of 20-25 minutes of playing time, where he has been near his most productive and the team has enjoyed the most success.
So Batum Will Never Start?
I'm definitely not saying that.
Looking at the final chart above, you can see that the Blazers have also played some of their best basketball without major minutes from Webster. You'll also recall that Webster's rebounding and defensive statistics and impact suggest he can make decent contributions in those facets regardless of his playing time.
What we'll likely see, I believe, is a transition process that eventually finds Batum starting and playing heavier minutes while Webster returns to a lesser-minute, rebounding and defense-oriented role player.
For the time being though, McMillan's theory of starting Webster over Batum is, if not preferred by me personally, understandable and reasonable given Webster's numberical body of work. At some point down the road, once we have access to more data about Batum's play, we'll be able to say more authoritatively whether 22 good minutes from Webster and 26 good minutes from Batum is noticeably worse than 15 less-productive minutes from Webster and 33 good minutes from Batum. But, at this point, the sample size is too small and it's fair to say that McMillan is on solid statistical ground when he starts Webster to ensure he can play at least 20-25 minutes a night.
-- Ben Golliver | benjamin.golliver@gmail.com | Twitter