With Beverley's return imminent, we are undoubtedly facing a barrage of controversy over who should be in the starting lineup and I think it is a good time to review what we have seen so far. With the NBAwowy website, which is just absolutely fantastic, I thought I would go in depth and summarize how well Beverley and Lin have fared thus far with their time in the starting lineup. I get fairly frustrated with blatantly misleading statistics which get thrown around here so often so I figured I would make this thread a reference point for further discussion and delve into the specifics of where the differences are. Here are my baseline query stats, if anyone wants to look at something I may have overlooked: http://nbawowy.com/query/llvyhyj21t7w4s4i http://nbawowy.com/query/0xyuna2cf905p14i Qualifiers: -The sample size isn't the greatest, but we are only 41 games in and this gives us the best available data. While history is not always useful in predictive performance, I think we can establish a pretty good baseline of what to expect moving forward unless some significant change occurs with coaching or other factor. -We are measuring performance in the starting lineup, not comparing the two as players overall, only theorizing why what has occurred, did occur. -Each game has the starters together for ~22 minutes and roughly 45 possessions. This was a haphazard sample selection of 6 games and the average minutes/possessions for those six games. Since the stats are shown per possession, I will normalize for 45 possessions to illustrate the difference one might estimate the data shows per game, where appropriate. -We are looking at data with ~220-230 minutes of playing time and # of possessions in mid 400s, which is a fairly hefty amount. -Quality of opponent is an area of concern, as the numbers would be skewed if one player played predominantly against weaker teams, so I will account for that below. Strength of bench is not accounted for and is an admitted weakness of this approach. -Data is through the OKC game. Quality of competition reference - Below are the teams faced, including offense PPP, defensive PPP, offensive rebounding % and defensive rebounding % as well as the average of all those metrics for each player. The point is to put a number on strength of competition faced: Lets start with defense: Beverley has traditionally been touted as the better defender of the two. Our quality of competition variable above shows that Lin, on average, has played against slightly more potent offenses (1.058 ppp vs. 1.069 ppp on average), although keep in mind he has had the benefit of quite a few opposing injuries which cannot really be quantified. With that in hand lets look at the lineup breakdown defensively: Using the PPP metric and accounting for .011 difference per possession in competition, we will favorably adjust Lin's number from 1.075 ppp to 1.064 ppp, in an attempt to normalize the competition, while keeping Beverley's at .991. Over the course of one game (45 possessions noted above) that comes out to: 47.9 points per game against Lin's starting lineups 44.6 points per game against Bev's starting lineups or 3.3 points per game. Looking through the breakout of types of shots given up, I didn't find anything I considered to be mind blowing. Anyone can look through the links above and theorize why the differences of shot types allowed are distributed. I don't think anyone is really surprise by the fact that Beverley lineups outperform Lin lineups defensively, but so far, the measurable difference per game has been about 3.3. points give or take, which is fairly significant. A quick note, we have lost two games by 3-4 points in the past few weeks (Hawks and Kings). On offense Traditionally Lin has been seen as the superior offensive player and often as the best "overall" player of both of them. Starting with our quality of competition variable, our spreadsheet above shows Bev on average has played against slightly more potent defenses (1.049 vs. 1.060 ppp on average), which is (interestingly identical to the defensive difference) .11. Lineup Breakdowns: Using the PPP metric and accounting for .011 difference per possession in competition, we will unfavorably adjust Lin's number from 1.086 to 1.075 and keep Beverley's at 1.162. Over the course of one game (45 possessions estimated above) that comes out to: 48.38 points scored per game for Lin's starting lineups 52.29 points scored per game for Bev's starting lineups or 3.9 points. This one is a little curious as most people would say that Lin is a better offensive player so something else must be going on to account for this difference. On first glance, you see a somewhat different shot selection allocation between the two groups. Lin's group has more dunks and layups, but they are hitting them at a lower percentage than Beverley's group. Beverley's group is shooting more 3 pointers and hitting them at a higher clip.This data does not seem to support our conclusion and explain why Beverley's group is out performing Lin's. With a dead end there, we can next look to other items which such as turnovers which might affect PPP. Since this is a measure of points per possession, rather than points per shot, extra turnovers might be what is causing the difference Unit Turnover Data: A traditional criticism of Jeremy has been his turnovers, but in this case, the turnovers are only .8 apart (14.8 to 14.0) per 100 possessions so that is a not factor for the most part. So that is another dead end. Another theory that might explain the difference in offensive production is that Jeremy is taking more shots than Beverley which is taking the ball out of Howard/Parsons/Harden's hands which results in less efficient shots/possessions. To test that, we can look the amount of shots taken by each player in each lineup. Below are the shot distributions by player for the lineups: Looking at the above data, it looks as the main difference in offensive production is that Lin is taking shots away from Parsons more than anyone. Harden looks like he consistently takes the most shots in either lineup. Howard/Jones take about the same amount of shots in both lineups, but Beverley takes significantly less shots than Lin and does less canibalizing of Parsons opportunities. There also seems to be a large personal production modifier, as noted below: Individually speaking we also see Parson's and Howard's TS% plummet with Lin in the lineup, while Jones's plummets with Beverley in the lineup. We can guess all day as to why that happens, but ultimately it is just going to be a guess. The most notable difference is the usage rates of each player with the different lineups. Just about everyone is consistent with the exception of Parsons whose possessions are eaten up by Lin, who takes about a 4% usage drop alongside Lin, vs. Beverley. A side note on Terrence Jones. He appears to play much better with Lin, which leads me to wonder if that is a Lin bonus, or if he has just come in into his own in general as a part of his own personal growth as a player. It will be interesting to see if his emergence continues when Beverley returns to the starting lineup. Rebounding It is unclear to me how NBAwowy accounts for rebounding in terms of PPP. Does an offensive rebound count a possession or does it restart a new possession? Regardless, we will still take a look to see what the data shows. Beverley has been touted as the superior rebounder, so we would expect the data to show as much. Keep in mind our quality of competition modifiers from above. It looks like Beverley faced weaker offensive rebounding teams by about 1.3% (making it easier for the Rockets to get defensive rebounds) and the defensive rebounding caliber was just about equal. Here are the rebounding rates: This shows a comfortable advantage for Beverley's squad: Offensive rebounds: +2.8% Defensive rebounds: +2.8% (4.1% - 1.3% modifier for competition) My best estimate is about 20 (give or take) offensive and defensive rebounding opportunities per game for the starting unit. Which makes for about one extra possession per game, which as we saw above, adds about 1.1 extra points. Takeaways -Up to this point there is not much data to argue Lin has performed well enough to keep the starting lineup spot. -Beverley's squads appear to score better, turn the ball over less, offensively rebound better, and defensively rebound the ball better. -You could claim that Lin is perhaps the catalyst for Jones's emergence, but there is an equally convincing argument that he has just developed as a player recently. Either way, we will be able to see if he continues his surge when Beverley returns. Thanks for reading! Go Rockets!
Appreciate the analysis. Just a quick thought about what you refer to as Lin "cannibalizing" Parson's usage. I may be wrong, but could the numbers seem to show this because Lin started for the first part of the season, (because of Bev's injury in game 1), and that's the same time Parsons was having major problems with his jumper?
Usage is an estimate of plays used, either make or miss. A shooting slump will not affect usage, only the TS%.
Maybe Terence Jones has a higher FG% with Lin because he is more willing to set screens, and Lin creates opportunities with those. Whereas Howard has a higher FG% with Bev because bev throws the rock down low to him more frequently. Just a thought.
Not to feed a potential Lin debate, but I don't understand why people don't want to see Lin in the 6th man role. He was thriving in that role. He was shooting at a high percentage, playing off the ball instead of having to create offense off the dribble. He was averaging 15 pts and playing extremely efficiently; making an early case for 6th man of the year. Beverly is a better on the ball defender and a better perimeter defender. He's also better at taking care of the ball. They compliment each other very well and they've played virtually the same amount of minutes. I, for one, like the combo of Bev as the starter, with Lin playing big minutes off the bench. It's best for the TEAM!
Nice effort Nitpicking here: -Bev played GSW without Iggy -McHale tells Bev to go after rebounds, and Lin/Brooks to get back on D so you would expect see higher rebound numbers] The other concern is these lineups have only been together 14 games (Bev) and 17 games (Lin). That's basically like writing this up at the end of November had we been healthy. And across those games you have Chandler's cold start, Harden's lingering foot injury, Harden turning it on and off on D, Dwight going from awful to awesome in the post. The main reason I'm harping on sample size is Parsons shot 26.7% from 3 with Lin and 51.6% from 3 with Bev, which makes me really question whether this isn't all just random variation.
Nice try good effort, but shouldn't this thread be named Lin vs Beverly? Oh wait, theres already one or two or 100 threads on that.
Bev and Lin really are dangerous when they play together, thats really good lineup. With all the talk about their individual defense, so far in this season it was awesome seeing plays where the both of them really shut players down during a double team and then transition for an easy bucket.
I think Bev will start for the rest of the season as long as he is healthy, and that is probably the best thing for the Rockets, at this point.
Nice effort Nitpicking here: - the whole team performance is so inconsistent this season, which game is included or excluded could have big impact - Lin's line up has faster pace, I don't think you consider that when you calculate offense total and defense total, and used the PPP*45 To be honest, just look at the past 5 games, @WAS, @BOS, @NOA, OKC, and MIL, even Lin started in all these 5 games, the way he was used was dramatically different, with this big of variation even within the same line up, I don't know how significant that could be to compare that to another line up, which one you would think is the normal performance of this line up then? Basically the variance within the Lin line up could be bigger than the difference between these two line up.
Thanks for the analysis, OP. You did a good job trying to eliminate (or at least contain) variables. As Morey has alluded to, you have to see if the stats confirm what you see with your eyes. My eyes tell me that Bev is a better rebounder and defender than Lin. He's also a better "fit" with the starting unit. I'd like to see TJones and Lin get a lot if time on the court together, and maybe give DMo and Casspi more time with Harden.
The title is misleading. It's more about analyzing the Lin vs Beverly debate ad nauseum. And I already know you have an agenda based off of your post history on this topic. But I appreciate the statistical hard work you put in honestly.
Nice analysis. Regarding Jones: Feel like he needed big minutes and confidence to adjust to the NBA, he only became a big part of our team and starting unit after Bev went out due to his injury. Pretty sure the last 15+ games were important and the reason for the big difference in the Lin/Bev numbers. Jones should shoot a much higher fg% now, when paired with Bev.
My theory is that Parsons is distracted when Lin is on the floor. That's the only possible explanation why Parsons's performance dropped so much. Conclusion: Parsons should be the 6th man. Now the Iggy-Howard thread makes sense.
