Cavalier Combos – Searching for Synergy with Plus Minus Data

April 15th, 2014 by Tom Pestak

The value of individual basketball players is coming into focus, and while individual player metrics are the holy grail of baseball, there needs to be a somewhat more holistic approach to team building in the NBA than the “moneyball” approach of maximizing the value of contracts by more rigorously modeling that which leads to winning.  I’m not sure how much GMs worry about chemistry in major league baseball.  Does it really matter if your center fielder and left-handed relief pitcher (that throws two pitches) get along?  There is very little on-field chemistry in baseball.  Obviously, the chemistry between pitchers and catchers is supremely important, but after that, your middle infielders should probably practice together?  (Right?)

Basketball is different, because players aren’t confined to specific roles in space and time.  In baseball, team-building from a positional standpoint and managing player roles is fairly obvious.  (I’m not implying that team-building or managing is simple – I’m saying that it’s easy to fill a hole at catcher.  You sign the best catcher you can afford and position him behind the plate.  You don’t really have to take the other guys on the field into consideration.)  Even if that lefty with two pitches is absolutely dominant, most managers know not to immediately promote him to starter and no GM is going to give him a 10 year 300 million dollar deal, even if he has a mind blowing BABIP.  He’ll probably mostly face lefties and bottom of the lineup righties.  Teams have figured this stuff out.

Basketball, in some cases, is losing the structure of traditional position roles.  Rule changes have morphed the game from an inside-outside attack to hyper-athletic guards initiating offense with dribble penetration and kicking out to spot-up shooters.  Stretch 4s and 5s are more common than back to the basket big men.  The most successful franchise of the past 15 years, the Spurs, more closely resembles a shape-shifting amoeba than a traditional box-and-1, house-shaped offense.  These changes necessitate a better understanding of chemistry and fit.  Assembling the greatest collection of individual talent does not guarantee greatness in the NBA, the 2012-2013 Lakers being the most recent example of this.

Question remain about the chemistry and fit of the Cavs “core”, as we at CtB have roughly categorized Irving, Waiters, Thompson, and Zeller/Bennett.  These are the young pups drafted early, making a lot of money, and struggling to maximize each other’s talents.  Last season, the Cavs added Shaun Livingston, Wayne Ellington, and Marresse Speights.  Together, with a rejuvenated Luke Walton, the four of them more than maximized each other’s talent.  Offensively, they had a multiplicative effect, and “The Herculoids” as we called them began dominating second units.  The Cavs looked like a competitive team for a little over a month, and the Herculoids had a lot to do with that.  For whatever reason, the Cavs front office was unwilling or unable to bring any of them back, and for much of this season, the Cavs’ offensive chemistry has been a quagmire.

So long, farewell.

Fare thee well.

Looking ahead, the Cavs front office has some difficult choices to make.  Brian Windhorst discussed some of these in his interview with Robert.  I decided to pore over the data and investigate which combinations of players seem to fit, in the sense that the Plus/Minus (PM) of lineups with them is better than the average of their individual PM.  Conversely, there exists enough PM data to make a case for certain players making each other worse.


Kyrie Irving + Tristan Thompson = This one really smarts.  Who’da thought Buddy Ball would be so ineffective.  Kyrie and Tristan have logged over 2000 minutes together this season.  Individually, neither has had a great PM, with Kyrie posting a -4.48 plus/minus per 48 minutes (PM/48) and Tristan -4.6 PM/48.  Again, this is an oversimplification, but it we expect their combined PM to be about the average, that would put them around -4.53 PM/48.  Instead, the Irving/Thompson duo is posting a -4.61 PM/48 together, not significantly worse, but there seems to exist little synergy between the two.  Without an outside shot, TT is ill-suited to be a great pick and roll partner with Kyrie.  Worse, he provides little to no floor spacing for the dribbling- ah intensive Irving.

Kyrie Irving + Jarrett Jack = They don’t work together, it’s that simple.  Lineups with these two flailing about in the backcourt have been outscored by 168 points in 1123 minutes.  That’s -7.18 points per 48 minutes, or essentially the scoring margin of the 2014 Utah Jazz.  The average of Jack’s -5.58 PM/48 and Irving’s -4.48 PM/48 is -5.02 PM/48.  Quite obviously, these two are crimpin’ each other’s style.

Kyrie Irving + Dion Waiters = Oy.  Yeah it’s not Kyrie and Jack bad, but they’re not making each other better.  It should be stated that they haven’t logged nearly as many minutes as Kyrie and Tristan and as we’ve seen with LeBron/Wade and Durant/Westbrook, having two ball-dominant players on the court at the same time creates challenges offensively (that can be overcome).  Individually, Waiters has posted a -1.62 PM/48 this season.  Averaging with Irving’s -4.48 PM/48 gives us -3.04 PM/48.  Together they have posted a -4.54 PM/48.  As Jason Lloyd noted, Kyrie and Dion have just recently embraced the need to work together to bring out the best in each other.  This goes along with Windhorst’s advice – they need to figure this out.  The best thing for the Cavs going forward would be for these two, their most talented players, to start “vibin”.  They’re running out of time to prove they can.  You really can’t offer big time money to both of these guys if they can’t run the floor together.

STOP!  Don’t go any further.  I can’t take this.  Does ANYONE play well with Kyrie?

Oh, sorry, didn’t mean to get you so upset.  I wasn’t planning on taking questions until the end but I’ll gladly answer yours as long as you put down that knife and step away from those terrified people you’re sitting next to…Thaaaaaaaat’s better.

Kyrie Irving + C.J. Miles = MASFRESCO!  How badly do you think Uncle Drew wishes your ankle would heal up? Doesn’t anyone have a SENZU bean in that locker room?!  GEEzus! comma Alonzo, offer up some stem cells or something.  Get that man back on his feet.  They’ve only logged 637 minutes together but they’ve been DEVAHstating.  Individually, Calvin Jr has posted a team high +2.29 PM/48.  Again if we average that with Irving’s -4.48 PM/48 we get a slightly less than average  -1.08 PM/48.  And yet, Irving plus Miles = POSITIVE FIVE POINT FOUR THREE!  (put yer HANDS UP!)  Yes, +5.43 PM/48, or better than any Eastern Conference Team’s point differential!  My theory is that Irving likes to play with 3-point snipers that know how to move without the ball.  C.J. fits the bill, and together, clearly, they have been outstanding.

Dion Waiters + Anderson Varejao = The Cavs best two-man unit that has logged over 1000 minutes just so happens to be the Wild Thing and Freon aka Threeion aka Primary Protaganist of “The Dion Waiters Zone”.  A la carte, Dion posts a -1.62 PM/48 while Andy is third on the Cavs with a +1.68 PM/48.  Averaging gives us an expected value of +0.03 PM/48, essentially breaking even with opponents.  Instead, the Cavaliers are outscoring opponents by +7.38 points per 48 minutes with those two in tow.  That’s a better point differential than Lob City 2014.  I can’t find a single player that looks better with Andy on the bench.  I guess what I’m saying is….#NeverTradeAndy.  Sign up for club membership by tweeting #NeverTradeAndy to @tompestak  Act now!  Members receive exclusive offers such as a free 18-month subscription to Cavs:TheBlog.  I’d tell you how much savings that is but you can’t divide by 0, it blows up computers.

Dion Waiters + Matthew Dellavedova =Delly is second on the team in PM/48 with +2.23.  Waiters, as noted above, is -1.62 PM/48.  Teams wielding Delly and Dion should post somewhere in the vicinity of +0.31 PM/48, or roughly the equivalent a .500 team.  That’s pretty good.  What we have here, other than a failure to communicate WITHOUT SCREAMING is a WORKING EXAMPLE OF TERRESTRIAL THERMONUCELAR FUSION!  Move over Taylor Wilson, please direct all federal funding to the Auss Boss and his colleague from South Philly.  Who would have thought that pouring FREON on a SUPERDOVA could create fusion?!  Yes, together, Dion and Delly have posted an astonishing +11.04 PM/48.  Or, with the raw numbers, they’ve outscored opponents by 158 points in just 687 minutes.  That’s better than the point differential of the 69-win ‘96-‘97 Chicago Bulls.

OK, so Dion looks pretty darned good with Varejao and with Dellavedova.  Not so much with Kyrie.  Who else struggles with Dion?

Dion Waiters + Tristan Thompson = Let’s get through this one as quickly as possible.  Tristan, overall, is posting a -4.6 PM/48.  Waiters, -1.62 PM/48.  Combined they should land at or near -3.11PM/48.  Instead, when they take the court, the Cavs are outscored by 6.06 points per 48 minutes.  Yeeeouch.  Not good at all.

Hey, I see you reaching for that knife again.  I can venture a guess as to what you’re thinking: Does Tristan Thompson, everyone’s lovable Canadian Dynamite, play well with anyone?!

Tristan Thompson + C.J. Miles = C.J. to the rescue again!  In 695 minutes together, Miles and Thompson have a +2.62 PM/48 while their individual averages indicate it should be more like -1.15 PM/48.  Of those 695 minutes, Kyrie appeared in 585 of them.  Those teams (obviously) did very well too.  It appears Kyrie+Tristan+C.J. works very well for the Cavs and Kyrie+Tristan+NOT C.J. falls flat.  Thompson+Varejao has outscored opponents by 14 points over 1058 minutes.  So those two are holding serve with opponents.

Anderson Varejao + Matthew Dellavedova = As you’ve probably guessed, the Cavs are outscoring opponents when these two play together.  Averaging their individual PM gives us an expected value of 1.96PM/48.  Care to venture a guess as whether they are exceeding that as a combo?  You mostly likely guessed “well yeah!  It’s Wild Thing and the guard version of Wild Thing, and their MO is making other players better.  Combined, their PM/48 is probably OVER 9000!”  Good guess, it’s over 9!  +9.27 PM/48 to be exact.

So what about the new guys?  Hawes and Deng?

There’s no easy way to put this.  The Cavs are in a tough spot with Spencer Hawes, True American Patriot, and 1st quarter 3-point specialist.  The sample sizes are small, as Hawes has played less than 600 minutes with any other Cavalier.  However, the Cavs have only outscored opponents with Hawes on the court when he’s accompanied by Delly, and they have gotten shredded when he shares the floor with Jack or Waiters.  He seems to have some synergy with Kyrie and Deng, but nothing to write home about.

Likewise, Deng hasn’t moved the needle too much when paired alongside the Cavaliers’ core.  He’s definitely helped, but not 10-million-dollars-a-year helped.  In very small samples, like under 200 minutes, he’s done some amazing things with Delly and/or Varejao.

What about Zeller?

Zeller is coming along nicely as a backup center.  He’s played well with the following front-court mates: Varejao (+5.9 PM/48) and Thompson (+3.26 PM/48) and not so well with Bennett and the 70 minutes he’s shared with Spencer Hawes.

Did you look at any other combos with Bennett? 

No.  I don’t see the point.  Outscoring the other team wasn’t the motivation for playing Bennett this season or he never would have seen the light of day.  And he’s had such a uneven season I don’t think any numbers would be instructive going forward.

Do you think you could wrap all this up for us?  Put a nice bow on it?  Make us feel better?

Zeller has improved greatly from last season.  Dellavedova is a legitimate NBA player and looks like a diamond in the rough, especially if he keeps up the torrid 3-point shooting he’s shown recently.

Good find, Cavs.

Good find, Cavs.

Mostly, I have concerns going forward.  I really don’t want the Cavs to make the same mistakes they made last season where they jettisoned (or, allowed the escape of) Livingston and Ellington.  Those two had skillsets the team sorely needed and they’ve missed it this season.  C.J. Miles and Matthew Dellavedova are leading the Cavs in PM and I believe it’s partially because they know how to play without the ball and they are excellent spot up 3-point shooters.  Additionally, Dellavedova is a supremely chaotic defender and Miles, at least at the beginning of the season, was leading the Cavs in percentage of points scored in transition.

Miles about to Posterize

Miles about to Posterize

And Varejao, well, Varejao’s just awesome, and it is high time each and every one of you starts shouting it from the mountaintops.  I was watching a game with my Dad last week and he audibly grimaced when Varejao took one of his safety-valve 17-footers from the top of the key.  Of course he made it.  (and the next two) My dad said something like “that’s not the shot you want…”  I spun around, incredulous…”DAD!  This isn’t your Mother’s Anderson Varejao!  It’s not your 2007 NBA Finals Varejao either.  He’s one of the best shooters in the league from that spot these days.”  (He’s actually shooting 50% from 15-17 feet this season, 60/121.  In 2006-2007 he was 5/24 for 21%)  My cousin still refers to him as an “annoying flopper.”  It shouldn’t be THIS HARD for someone’s reputation to change.  Varejao is a stud (when he’s playing).  He leads the Cavs in RAPM and has double the SWAgR of any other player.  Yet we’re always questioning whether he is in the Cavs plans.  Obviously, he’ll be 32 to start next season and he’s not going to be able to play at this level forever.  But right now, he’s the Cavs best player, hands down.


So I’m concerned that the Cavs core is really not playing well together, and the three players that seem to really make the team tick are not guaranteed to be back next season.  Varejao has a team option for 9.8 million dollars while Miles is not under contract.  Scary stuff if you ask me. (Editors Correction: This article originally noted that Dellavedova was not under contract next season.  In fact, he is at an non-guaranteed rate of  $816,482 and is a restricted free agent the year after). I think the Cavs need to commit to Delly, Miles, and add a reasonable 3-year-deal for Varejao, and then make their top priority a rim-protecting center.  These guys might not be the franchise pieces, but they certainly make the franchise pieces better.


There are also some lineup optimizations that Mike Brown can make.  The Delly – Jack – Waiters – Varejao foursome has been an astronomically high outlier this year, and has only logged 220 minutes together. What they’ve done in that 220 minutes is so crazy, I might write an entire post about it.  Also, despite being the top 2 Cavaliers by PM, C.J. Miles and Matthew Dellavedova have only played 163 minutes together.  And they dominated opponents by 43 points during that 163 minutes, so it’s unclear why C.J. Miles spent 821 minutes on the court without Delly.

I just don’t trust these numbers at all.

Well there are some limitations to this approach.  This is really a quick and dirty look at player combinations.  Obviously, PM is noisy in the first place and some of these sample sizes are way too small to put too much stock into the results.  Also, much like the analogy of the left-handed reliever that throws two pitches, comparing raw PM between players can be an apples to oranges situation if the bench guys are going into favorable situations and the starters are overwhelmed for some reason not completely their fault.  For that reason, I only trust RAPM for individual player value rankings (see chart above).  But these raw PM combinations can be instructive, especially the ones with lots of minutes.  Chemistry and fit has plagued the Cavs’ rebuild.  They’ve had plenty of time to see these guys play together.  There are hundreds of hours of game film, countless practices, and even bloggers willing to spend copious amounts of time toiling away in Excel.  And the Cavs have a beefed up analytics department to boot.  There is plenty of information to absorb and digest before making any decisions this offseason.  Let’s hope the Cavs maximize their assets and pinpoint the skill-sets needed to “enhance” the core.

Player Min PM RAPM PM per 48
Tristan Thompson 2567 -246 -2.19 -4.60
Kyrie Irving 2469 -229 -1.12 -4.45
Jarrett Jack 2227 -259 -3.05 -5.58
Dion Waiters 2045 -69 -0.12 -1.62
Anderson Varejao 1800 63 3.64 1.68
Luol Deng 1353 -40 2.86* -1.42
Matthew Dellavedova 1245 58 -0.03 2.24
C.J. Miles 984 47 1.10 2.29
Tyler Zeller 1023 -20 -1.14 -0.94
Alonzo Gee 989 -249 -3.81 -12.08
Spencer Hawes 780 -97 -2.80* -5.97
Earl Clark 647 -144 -5.69 -10.68
Anthony Bennett 480 -92 -3.34 -9.20
Duo Min PM E(PM) PM per 48
Irving+Thompson 2022 -194 -4.53 -4.61
Irving+Jack 1123 -168 -5.02 -7.18
Irving+Waiters 1036 -98 -3.04 -4.54
Irving+Miles 637 72 -1.08 5.43
Waiters+Varejao 1067 164 0.03 7.38
Waiters+Dellavedova 687 158 0.31 11.04
Waiters+Thompson 1203 -152 -3.11 -6.06
Thompson+Miles 695 38 -1.15 2.62
Varejao+Dellavedova 590 114 1.96 9.27
Hawes+Irving 499 -33 -5.21 -3.17
Hawes+Waiters 403 -101 -3.79 -12.03
Hawes+Thompson 587 -81 -5.28 -6.62
Hawes+Deng 552 -18 -3.69 -1.57
Hawes+Jack 540 -108 -5.78 -9.60
Hawes+Dellavedova 192 17 -1.87 4.25
Hawes+Varejao 87 -4 -2.14 -2.21
Deng+Irving 472 -20 -2.94 -2.03
Deng+Waiters 319 -8 -1.52 -1.20
Deng+Thompson 553 -24 -3.01 -2.08
Deng+Jack 524 -13 -3.50 -1.19
Deng+Dellavedova 181 54 0.41 14.32
Deng+Varejao 136 36 1.96 12.71
Zeller+Irving 446 -19 -2.70 -2.04
Zeller+Waiters 554 44 -1.28 3.81
Zeller+Thompson 383 26 -2.77 3.26
Zeller+Jack 630 -4 -3.26 -0.30
Zeller+Dellavedova 555 40 0.65 3.46
Zeller+Varejao 244 30 0.37 5.90
Zeller+Hawes 70 -15 -3.45 -10.29
Zeller+Bennett 219 -37 -5.07 -8.11
Miles+Dellavedova 163 43 2.26 12.66

Green means that the combo’s PM/48 was better than both individual PM/48.  Red means worse than both individual PM/48.  Black means that the combo PM fell somewhere between each player’s individual PM/48.  While I must stress again that it’s an oversimplification, you can think of the green combos as being better than the sum of their parts and the red combos being worse.  The black are about equal to the sum of their parts.

*Note that Spencer Hawes and Luol Deng’s RAPM values in the first table are italicized.  These values are for their entire season, not just with the Cavs.

Stats from,, and  gB RAPM courtesy of @talkingpractice

Tags: , ,