2014 NBA Playoffs 2nd Round Preview

There are many potential areas of improvement in both the model and the testing of the model, most importantly perhaps a means of adjusting for various series states (i.e. how should the line be adjusted when the home team is down 1-0), but in the still small sample size of the first round, I was 20-14-1 against the spread using my model with subjective adjustments.  While that's a 59% win percentage (which would be amazing if it were indeed my true win percentage), the small sample size means that this likely isn't my true win percentage, as even a coin flip would still finish with the same record or better in 34 trials 20% of the time.  Here are the full results:

ATL 6 times: 3-2-1
MIA 3 times: 2-1-0
BKN 7 times: 4-3-0
CHI 2 tims: 1-1-0
DAL 1 time: 1-0-0
SAS 1 time: 0-1-0
MEM 7 times : 4-3-0
GSW 7 times: 5-2-0
HOU 1 time: 0-1-0

I will now proceed to provide a similar analysis of the second round.  There will be new minutes allocations (changed due to updated information, i.e. that Shane Battier has apparently retired, as well as different matchups) and a slightly smaller home-court-advantage adjustment.  The minutes allocations will be slightly more conservative for the top players, as I failed to account for the potential of foul trouble dragging average minutes down (which is what happened in Clippers-Warriors), and the home-court-advantage adjustment will be 3.2 points.  Of course, as mentioned earlier, one future improvement would be to make these adjustments empirically.  Similar to before, all of the computation was done in R and the script I ran will be included.

Series Previews:

East: (1) Indiana vs. (5) Washington

	Model	Vegas*
Game 1 Probability	IND: 69% / WAS: 31%	IND: -175 / WAS: +155
Game 1 Spread	IND: -5.4 / WAS: +5.4	IND: -4 (-115) / WAS: +4 (-105)
Series Probability	IND: 71% / WAS: 29%	IND: -160 / WAS: +140

Based solely on regular season numbers, it would appear that my model would suggest betting the Pacers.  However, the same disclaimer from Round 1 still applies, given that Indiana has simply been a much worse team over the final few months, including the series against the Hawks.  Given that my model does not have a recency weighting factor, the regular season numbers use Indiana's season-long averages.  One potential proxy for Indiana's struggles (that happens to include a separate but also important factor) would be Nate Silver's study on rest.  If I understand the study correctly, a team's margin of victory relative to SRS in the first round (independent of the number of games played) is not significant, but the recency factor should impact the number of games played as well.  Furthermore, while Atlanta posed specific matchup issues for Indiana that Washington doesn't in terms of three-point shooting bigs, there are other issues for Indiana.  The supposed better matchup for Hibbert might not necessarily be a good thing, as Hibbert has been horrendous in the second half of the year and handing Hibbert more minutes might have a net negative effect.  Furthermore, Indiana is a horrible offensive team that can't pass, screen, or space the floor, and their past offensive adequacy was dependent on David West overpowering his defender and drawing help, and he'll likely be unable to manhandle Nene the way he might other defenders.  As a result, I would adjust the margin of victory for each game towards Washington by approximately 4 points.  This moves Indiana's expected margin of victory in Game 1 to 1.5 and actually makes Washington a 58% favorite in the series.  As a result, I would bet Washington.

East: (2) Miami vs. (6) Brooklyn

	Model	Vegas*
Game 1 Probability	MIA: 79% / BKN: 21%	MIA: OFF / BKN: OFF
Game 1 Spread	MIA: -10.0 / BKN: +10.0	MIA: -7 (-115) / BKN: +7 (-105)
Series Probability	MIA: 90% / BKN: 10%	MIA: -550 / BKN: +375

There are a number of potential adjustments to the model in the Miami-Brooklyn series.  The first is that Silver’s study suggests that a team coming off a sweep is even more likely than expected to win against a team coming off a seven-game series.  Counteracting this, there is some anecdotal evidence that there might be diminishing marginal returns on RAPM, so, for example, a ten-point expected margin using RAPM in reality might only be eight.  Furthermore, Brooklyn had significant success against Miami during their regular season four-game series (yes, the margins of victory were small, but that they were positive is a significant improvement for Brooklyn over the overall numbers of each team), and there are matchup reasons that explain this.  Miami’s bread-and-butter half-court offense involves LeBron pick-and-rolls with a big, and Brooklyn has the big wings and quick bigs to switch these plays. Furthermore, supply of big wings also makes it difficult for LeBron to post up and draw help, as Pierce, Johnson, Livingston, Kirilenko, and even Teletovic could have success defending him in isolations.  On the other side of the ball, Miami’s defense in the past has been predicated on aggressive traps of pick-and-rolls (a little less so this year), which led to both efficient defense and efficient offense (as the traps led to turnovers and fast breaks).  However, Brooklyn does not run many pick-and-rolls, and their offense involves more isolations than normal.  While isolations, even ones by skilled isolation players, are inherently inefficient shots, they’re inefficient against anyone and Brooklyn won’t experience much extra resistance on their isolations because they’re facing Miami.  As a result, I would adjust the expected margin six points towards Miami for the first-round performance/rest, five points towards Brooklyn for diminishing marginal returns (more for the Heat home games), and five points towards Brooklyn for the matchup issues.  This would make the Heat just under a six-point favorite in Game 1 and a 73% favorite in the series.  These numbers are already very close to the Vegas lines, and given the uncertainty in my manual adjustments, I would abstain from this series for now.

West: (1) San Antonio vs. (5) Portland

	Model	Vegas*
Game 1 Probability	SAS: 84% / POR: 16%	SAS: OFF / POR: OFF
Game 1 Spread	SAS: -11.9 / POR: +11.9	SAS: -6.5 / POR: +6.5
Series Probability	SAS: 96% / POR: 4%	SAS: -900 / POR: +550

Having not paid much attention to inter-series rankings beforehand, I was surprised to see that Dallas graded as a significantly better team than Portland.  The reason for the dissonance was threefold: 1) Dirk was a significantly better per-minute player than Aldridge (who already was pretty awesome), 2) Portland’s worst rotation players (Wright and Williams) grade out worse than Dallas’ (Calderon and Marion), and 3) Dallas’ players simply have less name value than Portland’s (specifically, Carter, Harris, and Wright).  However, there are reasons to believe that these are two very similar teams. They’re both built around a two-point shooting bigs creating threes for teammates, and both coaches smartly utilized switching man defenses to slow down elite pick-and-roll attacks in the first round (unsurprising given Stotts’ background as a former assistant under Carlisle).  Furthermore, much of Dallas’ high ranking in my model was dependent on a minutes distribution tilted away from their bad starters, and simply adjusting Wright, Harris, and Crowder’s minutes towards Calderon, Marion, Dalembert, and Blair lowers the Mavs’ efficiency down to 7.9, much closer to Portland’s 6.9.  Given San Antonio’s struggles against Dallas and the clear blueprint established in slowing down their offense, I certainly don’t believe the Spurs should be almost a nine-point favorite on a neutral court.  However, the Spurs also adjusted later in the series to more Ginobili pick-and-rolls in place of Parker pick-and-rolls, as Ginobili was able to hit pull-up threes or get to the basket using his long strides against switches, and they had much more success in the later games of their series.  As a result, I would adjust the line four to five points towards Portland per game (more for the Spurs home games due to diminishing marginal returns on RAPM), which would make the Spurs a seven point favorite in Game 1 and an 83% favorite in the series.  As a result, I would abstain from this series for now.

West: (2) Oklahoma City vs. (3) L.A. Clippers

	Model	Vegas*
Game 1 Probability	OKC: 52% / LAC: 48%	OKC: -225 / LAC: +185
Game 1 Spread	OKC: -0.1 / LAC: +0.1	OKC: -5 (-115) / LAC: +5 (-105)
Series Probability	OKC: 36% / LAC: 64%	OKC: -200 / LAC: +170

The biggest difference between the Thunder and Clippers statistically is that the Thunder give significant minutes to horrible players (Butler, Perkins, Adams).  There is definitely a mainstream misconception that the Grizzlies were better than the Warriors minus Bogut and thus that the Clippers struggled in Round 1 more than the Thunder, and this is simply not true.  In reality, the Warriors were an elite team that also presented matchup problems that Oklahoma City likely will not provide.  Golden State was able to mostly limit the Clippers' Paul-Griffin PNR by switching, which was made possible by the Warriors assigning a bigger defender as Paul's primary defender (mostly Thompson and sometimes Iguodala) and Draymond Green's ability to defend perimeter players after a switch.  Furthermore, the Warriors had no hiding places for perimeter players, so Darren Collison, Redick, and Crawford were enormous liabilities defensively, and this won't be the case with Fisher, Sefolosha, Butler, and even Jackson serving as potential hiding spots.  Furthermore, Oklahoma City had a lot of issues defending the three in the regular season, due largely to inattentive ball-watching or over-helping, and the Clippers have the ball/player movement to take advantage of that.  Durant will definitely be a problem, given that Barnes is at best an above average defender, and the Clippers don't really have big wings behind him as second options.  Still, my raw numbers like the Clippers significantly more than Vegas does, and the net matchup adjustments may actually be in the Clippers favor.  As a result, I would have a lot of confidence betting the Clippers in this series.

*All Vegas numbers from Bovada


List of minutes estimates and xRAPM ratings:

Show/Hide

x
     Spurs:

player	minutes	orpm	drpm	rpm	onet	dnet	net
Boris Diaw	20	0.55	1.45	2.00	0.23	0.60	0.83
Cory Joseph	2	-0.83	-3.38	-4.21	-0.03	-0.14	-0.18
Danny Green	25	0.68	2.61	3.29	0.35	1.36	1.71
Kawhi Leonard	35	0.58	1.84	2.42	0.42	1.34	1.76
Manu Ginobili	30	4.66	-0.11	4.55	2.91	-0.07	2.84
Marco Belinelli	12	1.03	-2.94	-1.91	0.26	-0.74	-0.48
Matt Bonner	5	2.05	0.37	2.42	0.21	0.04	0.25
Patty Mills	15	2.70	0.50	3.20	0.84	0.16	1.00
Tiago Splitter	26	-0.92	4.59	3.67	-0.50	2.49	1.99
Tim Duncan	35	-0.10	5.37	5.27	-0.07	3.92	3.84
Tony Parker	35	3.16	-0.07	3.09	2.30	-0.05	2.25

     Blazers:

player	minutes	orpm	drpm	rpm	onet	dnet	net
C.J. McCollum	3	0.03	-0.66	-0.63	0.00	-0.04	-0.04
Damian Lillard	40	4.31	-2.03	2.28	3.59	-1.69	1.90
Dorell Wright	10	-0.86	-0.94	-1.80	-0.18	-0.20	-0.38
Joel Freeland	2	-2.83	0.96	-1.87	-0.12	0.04	-0.08
LaMarcus Aldridge	40	1.73	3.17	4.90	1.44	2.64	4.08
Meyers Leonard	0	-4.22	-0.98	-5.20	0.00	0.00	0.00
Mo Williams	25	0.28	-2.66	-2.38	0.15	-1.39	-1.24
Nicolas Batum	40	1.47	-1.19	0.28	1.23	-0.99	0.23
Robin Lopez	37	-0.29	3.38	3.09	-0.22	2.61	2.38
Thomas Robinson	5	-2.37	-2.03	-4.40	-0.25	-0.21	-0.46
Wesley Matthews	38	1.54	-0.89	0.65	1.22	-0.70	0.51

     Thunder:

player	minutes	orpm	drpm	rpm	onet	dnet	net
Caron Butler	22	-0.75	-3.61	-4.36	-0.34	-1.65	-2.00
Derek Fisher	19	-1.14	2.00	0.86	-0.45	0.79	0.34
Jeremy Lamb	2	-0.49	-0.79	-1.28	-0.02	-0.03	-0.05
Kendrick Perkins	20	-7.12	2.73	-4.39	-2.97	1.14	-1.83
Kevin Durant	43	6.36	0.17	6.53	5.70	0.15	5.85
Nick Collison	16	2.75	3.40	6.15	0.92	1.13	2.05
Reggie Jackson	23	0.81	0.50	1.31	0.39	0.24	0.63
Russell Westbrook	36	4.30	0.19	4.49	3.23	0.14	3.37
Serge Ibaka	34	0.88	3.03	3.91	0.62	2.15	2.77
Steven Adams	10	-2.68	-0.93	-3.61	-0.56	-0.19	-0.75
Thabo Sefolosha	15	-1.14	2.29	1.15	-0.36	0.72	0.36

     Clippers:

player	minutes	orpm	drpm	rpm	onet	dnet	net
Blake Griffin	41	2.58	1.88	4.46	2.20	1.61	3.81
Chris Paul	38	5.44	1.84	7.28	4.31	1.46	5.76
Danny Granger	13	0.30	0.70	1.00	0.08	0.19	0.27
Darren Collison	15	1.48	-3.00	-1.52	0.46	-0.94	-0.48
DeAndre Jordan	36	0.68	3.60	4.28	0.51	2.70	3.21
Glen Davis	10	-3.08	2.00	-1.08	-0.64	0.42	-0.23
Hedo Turkoglu	2	-0.96	0.24	-0.72	-0.04	0.01	-0.03
J.J. Redick	25	2.98	-2.96	0.02	1.55	-1.54	0.01
Jamal Crawford	25	1.55	-2.31	-0.76	0.81	-1.20	-0.40
Jared Dudley	5	-0.76	0.24	-0.52	-0.08	0.03	-0.05
Matt Barnes	30	2.14	0.58	2.72	1.34	0.36	1.70

     Pacers:

player	minutes	orpm	drpm	rpm	onet	dnet	net
C.J. Watson	18	0.39	-0.40	-0.01	0.15	-0.15	0.00
Chris Copeland	4	0.34	-2.05	-1.71	0.03	-0.17	-0.14
David West	36	0.98	2.33	3.31	0.74	1.75	2.48
Evan Turner	14	-1.14	-1.12	-2.26	-0.33	-0.33	-0.66
George Hill	37	2.00	0.60	2.60	1.54	0.46	2.00
Ian Mahinmi	14	-4.85	4.05	-0.80	-1.41	1.18	-0.23
Lance Stephenson	35	0.79	-0.04	0.75	0.58	-0.03	0.55
Luis Scola	10	-3.27	0.76	-2.51	-0.68	0.16	-0.52
Paul George	40	0.03	2.49	2.52	0.03	2.08	2.10
Roy Hibbert	32	-0.94	3.68	2.74	-0.63	2.45	1.83

     Wizards:

player	minutes	orpm	drpm	rpm	onet	dnet	net
Al Harrington	4	-0.29	0.22	-0.07	-0.02	0.02	-0.01
Andre Miller	8	1.80	0.12	1.92	0.30	0.02	0.32
Bradley Beal	38	-0.86	-0.08	-0.94	-0.68	-0.06	-0.74
Drew Gooden	8	-0.97	-2.19	-3.16	-0.16	-0.37	-0.53
John Wall	40	2.54	-0.46	2.08	2.12	-0.38	1.73
Marcin Gortat	36	-0.16	3.85	3.69	-0.12	2.89	2.77
Martell Webster	17	2.19	-2.87	-0.68	0.78	-1.02	-0.24
Nene Hilario	36	-1.65	4.39	2.74	-1.24	3.29	2.06
Trevor Ariza	38	-0.26	0.94	0.68	-0.21	0.74	0.54
Trevor Booker	15	-0.12	-2.77	-2.89	-0.04	-0.87	-0.90

     Heat:

player	minutes	orpm	drpm	rpm	onet	dnet	net
Chris Andersen	16	0.42	3.99	4.41	0.14	1.33	1.47
Chris Bosh	38	0.29	3.70	3.99	0.23	2.93	3.16
Dwyane Wade	35	1.23	0.87	2.10	0.90	0.63	1.53
James Jones	6	-2.10	0.97	-1.13	-0.26	0.12	-0.14
LeBron James	42	8.72	-0.16	8.56	7.63	-0.14	7.49
Mario Chalmers	33	0.54	1.03	1.57	0.37	0.71	1.08
Michael Beasley	2	-2.94	-2.89	-5.83	-0.12	-0.12	-0.24
Norris Cole	12	-1.60	-1.87	-3.47	-0.40	-0.47	-0.87
Rashard Lewis	10	-2.34	-0.72	-3.06	-0.49	-0.15	-0.64
Ray Allen	20	2.03	-2.77	-0.74	0.85	-1.15	-0.31
Shane Battier	12	-0.29	1.36	1.07	-0.07	0.34	0.27
Toney Douglas	2	0.25	-1.62	-1.37	0.01	-0.07	-0.06
Udonis Haslem	12	-4.21	1.60	-2.61	-1.05	0.40	-0.65

     Nets:

player	minutes	orpm	drpm	rpm	onet	dnet	net
Alan Anderson	15	-0.88	-1.64	-2.52	-0.28	-0.51	-0.79
Andray Blatche	6	-0.54	-0.04	-0.58	-0.07	-0.01	-0.07
Andrei Kirilenko	16	-0.72	1.16	0.44	-0.24	0.39	0.15
Deron Williams	36	4.69	-1.86	2.83	3.52	-1.40	2.12
Joe Johnson	41	3.39	-1.08	2.31	2.90	-0.92	1.97
Jorge Gutierrez	0	-2.11	-1.27	-3.38	0.00	0.00	0.00
Kevin Garnett	25	-2.86	6.57	3.71	-1.49	3.42	1.93
Marcus Thornton	10	-0.12	-1.96	-2.08	-0.03	-0.41	-0.43
Mason Plumlee	17	-0.90	-1.34	-2.24	-0.32	-0.47	-0.79
Mirza Teletovic	15	-0.28	-2.16	-2.44	-0.09	-0.68	-0.76
Paul Pierce	33	-0.62	3.14	2.52	-0.43	2.16	1.73
Shaun Livingston	26	-0.18	0.27	0.09	-0.10	0.15	0.05

R Script:

Show/Hide

library(XML)
library(RCurl)
v.pages <- 1:11            #change if increase in number of pages on http://espn.go.com/nba/statistics/rpm/_/page/1/sort/RPM

v.column.names <- c("player","team","orpm","drpm","rpm")
v.teams.home <- c("sas","okc","ind","mia")
v.teams.away <- c("por","lac","was","bkn")
v.teams <- c("sas","por","okc","lac","ind","was","mia","bkn")

hca <- 3.2                #difference between neutral court and home court; this is total hca adjustment, not the adjustment for each team
avg_eff <- 104            #average league points per 100 possessions

df.rpm <- data.frame(player=character(0),team=character(0),orpm=numeric(0),drpm=numeric(0),rpm=numeric(0))

for(page in v.pages)
{
    v.url.rpm <- paste("http://espn.go.com/nba/statistics/rpm/_/page/",page,"/sort/RPM",sep="")
    table <- readHTMLTable(v.url.rpm)[[1]]
    table <- table[as.character((table[,2]))!="NAME",]
    table2 <- cbind(as.data.frame(matrix(unlist(strsplit(as.character(table[,2]),", ")),ncol=2,byrow=TRUE),stringsAsFactors=FALSE)[,1], as.data.frame(matrix(unlist(table[,3]),ncol=1,byrow=TRUE),stringsAsFactors=FALSE), as.data.frame(matrix(unlist(table[,6]),ncol=1,byrow=TRUE),stringsAsFactors=FALSE),as.data.frame(matrix(unlist(table[,7]),ncol=1,byrow=TRUE),stringsAsFactors=FALSE),as.data.frame(matrix(unlist(table[,8]),ncol=1,byrow=TRUE),stringsAsFactors=FALSE))
    table2[,3] <- as.numeric(table2[,3])
    table2[,4] <- as.numeric(table2[,4])
    table2[,5] <- as.numeric(table2[,5])
    colnames(table2) <- v.column.names
    df.rpm <- rbind(df.rpm,table2)
}

#minutes distributions and individual xrapm:
df.sas.r2 <- data.frame(player=c("Tony Parker","Tim Duncan","Kawhi Leonard","Marco Belinelli","Boris Diaw","Danny Green","Manu Ginobili","Tiago Splitter","Patty Mills","Cory Joseph","Matt Bonner"),minutes=c(35,35,35,12,20,25,30,26,15,2,5))
df.sas.r2 <- merge(df.sas.r2, df.rpm[,c(1,3,4,5)], by="player", all.x=TRUE)
df.por.r2 <- data.frame(player=c("LaMarcus Aldridge","Nicolas Batum","Damian Lillard","Wesley Matthews","Robin Lopez","Mo Williams","Dorell Wright","C.J. McCollum","Thomas Robinson","Meyers Leonard","Joel Freeland"),minutes=c(40,40,40,38,37,25,10,3,5,0,2))
df.por.r2 <- merge(df.por.r2, df.rpm[,c(1,3,4,5)], by="player", all.x=TRUE)
df.okc.r2 <- data.frame(player=c("Kevin Durant","Serge Ibaka","Russell Westbrook","Reggie Jackson","Caron Butler","Thabo Sefolosha","Jeremy Lamb","Kendrick Perkins","Derek Fisher","Nick Collison","Steven Adams"),minutes=c(43,34,36,23,22,15,2,20,19,16,10))
df.okc.r2 <- merge(df.okc.r2, df.rpm[,c(1,3,4,5)], by="player", all.x=TRUE)
df.lac.r2 <- data.frame(player=c("Blake Griffin","DeAndre Jordan","Chris Paul","Jamal Crawford","J.J. Redick","Matt Barnes","Darren Collison","Jared Dudley","Danny Granger","Glen Davis","Hedo Turkoglu"),minutes=c(41,36,38,25,25,30,15,5,13,10,2))
df.lac.r2 <- merge(df.lac.r2, df.rpm[,c(1,3,4,5)], by="player", all.x=TRUE)
df.ind.r2 <- data.frame(player=c("Paul George","Lance Stephenson","George Hill","David West","Roy Hibbert","Evan Turner","C.J. Watson","Luis Scola","Ian Mahinmi","Chris Copeland"),minutes=c(40,35,37,36,32,14,18,10,14,4))
df.ind.r2 <- merge(df.ind.r2, df.rpm[,c(1,3,4,5)], by="player", all.x=TRUE)
df.was.r2 <- data.frame(player=c("John Wall","Trevor Ariza","Bradley Beal","Marcin Gortat","Nene Hilario","Martell Webster","Trevor Booker","Drew Gooden","Al Harrington","Andre Miller"),minutes=c(40,38,38,36,36,17,15,8,4,8))
df.was.r2 <- merge(df.was.r2, df.rpm[,c(1,3,4,5)], by="player", all.x=TRUE)
df.mia.r2 <- data.frame(player=c("LeBron James","Dwyane Wade","Chris Bosh","Mario Chalmers","Ray Allen","Norris Cole","Shane Battier","Chris Andersen","Rashard Lewis","Toney Douglas","Michael Beasley","Udonis Haslem","James Jones"),minutes=c(42,35,38,33,20,12,12,16,10,2,2,12,6))
df.mia.r2 <- merge(df.mia.r2, df.rpm[,c(1,3,4,5)], by="player", all.x=TRUE)
df.bkn.r2 <- data.frame(player=c("Joe Johnson","Deron Williams","Paul Pierce","Shaun Livingston","Marcus Thornton","Alan Anderson","Andray Blatche","Kevin Garnett","Mirza Teletovic","Andrei Kirilenko","Mason Plumlee","Jorge Gutierrez"),minutes=c(41,36,33,26,10,15,6,25,15,16,17,0))
df.bkn.r2 <- merge(df.bkn.r2, df.rpm[,c(1,3,4,5)], by="player", all.x=TRUE)

df.sas.r2$onet <- df.sas.r2$minutes * df.sas.r2$orpm / 48
df.por.r2$onet <- df.por.r2$minutes * df.por.r2$orpm / 48
df.okc.r2$onet <- df.okc.r2$minutes * df.okc.r2$orpm / 48
df.lac.r2$onet <- df.lac.r2$minutes * df.lac.r2$orpm / 48
df.ind.r2$onet <- df.ind.r2$minutes * df.ind.r2$orpm / 48
df.was.r2$onet <- df.was.r2$minutes * df.was.r2$orpm / 48
df.mia.r2$onet <- df.mia.r2$minutes * df.mia.r2$orpm / 48
df.bkn.r2$onet <- df.bkn.r2$minutes * df.bkn.r2$orpm / 48

df.sas.r2$dnet <- df.sas.r2$minutes * df.sas.r2$drpm / 48
df.por.r2$dnet <- df.por.r2$minutes * df.por.r2$drpm / 48
df.okc.r2$dnet <- df.okc.r2$minutes * df.okc.r2$drpm / 48
df.lac.r2$dnet <- df.lac.r2$minutes * df.lac.r2$drpm / 48
df.ind.r2$dnet <- df.ind.r2$minutes * df.ind.r2$drpm / 48
df.was.r2$dnet <- df.was.r2$minutes * df.was.r2$drpm / 48
df.mia.r2$dnet <- df.mia.r2$minutes * df.mia.r2$drpm / 48
df.bkn.r2$dnet <- df.bkn.r2$minutes * df.bkn.r2$drpm / 48

df.sas.r2$net <- df.sas.r2$minutes * df.sas.r2$rpm / 48
df.por.r2$net <- df.por.r2$minutes * df.por.r2$rpm / 48
df.okc.r2$net <- df.okc.r2$minutes * df.okc.r2$rpm / 48
df.lac.r2$net <- df.lac.r2$minutes * df.lac.r2$rpm / 48
df.ind.r2$net <- df.ind.r2$minutes * df.ind.r2$rpm / 48
df.was.r2$net <- df.was.r2$minutes * df.was.r2$rpm / 48
df.mia.r2$net <- df.mia.r2$minutes * df.mia.r2$rpm / 48
df.bkn.r2$net <- df.bkn.r2$minutes * df.bkn.r2$rpm / 48

#team-level xrapm
df.all.r2 <- data.frame(team=v.teams)
df.all.r2$onet <- as.vector(t(data.frame(sas=sum(df.sas.r2$onet),por=sum(df.por.r2$onet),okc=sum(df.okc.r2$onet),lac=sum(df.lac.r2$onet),ind=sum(df.ind.r2$onet),was=sum(df.was.r2$onet),mia=sum(df.mia.r2$onet),bkn=sum(df.bkn.r2$onet))))
df.all.r2$dnet <- as.vector(t(data.frame(sas=sum(df.sas.r2$dnet),por=sum(df.por.r2$dnet),okc=sum(df.okc.r2$dnet),lac=sum(df.lac.r2$dnet),ind=sum(df.ind.r2$dnet),was=sum(df.was.r2$dnet),mia=sum(df.mia.r2$dnet),bkn=sum(df.bkn.r2$dnet))))
df.all.r2$net <- as.vector(t(data.frame(sas=sum(df.sas.r2$net),por=sum(df.por.r2$net),okc=sum(df.okc.r2$net),lac=sum(df.lac.r2$net),ind=sum(df.ind.r2$net),was=sum(df.was.r2$net),mia=sum(df.mia.r2$net),bkn=sum(df.bkn.r2$net))))
df.all.r2$home_perc <- (avg_eff+df.all.r2$onet+hca/4)^14/((avg_eff+df.all.r2$onet+hca/4)^14+(avg_eff-df.all.r2$dnet-hca/4)^14)
df.all.r2$away_perc <- (avg_eff+df.all.r2$onet-hca/4)^14/((avg_eff+df.all.r2$onet-hca/4)^14+(avg_eff-df.all.r2$dnet+hca/4)^14)

#specific round 2 matchups
df.r2.matchups.1 <- data.frame(team=v.teams.home, opponent=v.teams.away, stringsAsFactors = FALSE)
df.r2.matchups.1$team_home_perc <- merge(data.frame(team=df.r2.matchups.1$team),df.all.r2[,c(1,5)], by="team", all.x=TRUE, sort=FALSE)[,2]
df.r2.matchups.1$team_away_perc <- merge(data.frame(team=df.r2.matchups.1$team),df.all.r2[,c(1,6)], by="team", all.x=TRUE, sort=FALSE)[,2]
df.r2.matchups.1$opponent_home_perc <- merge(data.frame(team=df.r2.matchups.1$opponent),df.all.r2[,c(1,5)], by="team", all.x=TRUE, sort=FALSE)[,2]
df.r2.matchups.1$opponent_away_perc <- merge(data.frame(team=df.r2.matchups.1$opponent),df.all.r2[,c(1,6)], by="team", all.x=TRUE, sort=FALSE)[,2]
df.r2.matchups.1$team_matchup_home_perc <- (df.r2.matchups.1$team_home_perc - (df.r2.matchups.1$team_home_perc*df.r2.matchups.1$opponent_away_perc))/(df.r2.matchups.1$team_home_perc + df.r2.matchups.1$opponent_away_perc - (2*df.r2.matchups.1$team_home_perc*df.r2.matchups.1$opponent_away_perc))
df.r2.matchups.1$team_matchup_away_perc <- (df.r2.matchups.1$team_away_perc - (df.r2.matchups.1$team_away_perc*df.r2.matchups.1$opponent_home_perc))/(df.r2.matchups.1$team_away_perc + df.r2.matchups.1$opponent_home_perc - (2*df.r2.matchups.1$team_away_perc*df.r2.matchups.1$opponent_home_perc))
df.r2.matchups.1$series_perc <- df.r2.matchups.1$team_matchup_home_perc^4 + 4*df.r2.matchups.1$team_matchup_home_perc^3*(1-df.r2.matchups.1$team_matchup_home_perc)*(1-(1-df.r2.matchups.1$team_matchup_away_perc)^3) + 6*df.r2.matchups.1$team_matchup_home_perc^2*(1-df.r2.matchups.1$team_matchup_home_perc)^2*((3*df.r2.matchups.1$team_matchup_away_perc^2*(1-df.r2.matchups.1$team_matchup_away_perc))+df.r2.matchups.1$team_matchup_away_perc^3)+4*
df.r2.matchups.1$team_matchup_home_perc*(1-df.r2.matchups.1$team_matchup_home_perc)^3*df.r2.matchups.1$team_matchup_away_perc^3
df.r2.matchups.2 <- df.r2.matchups.1[,c(2,1)]
colnames(df.r2.matchups.2) <- c("team","opponent")
df.r2.matchups.2$team_home_perc <- df.r2.matchups.1$opponent_home_perc
df.r2.matchups.2$team_away_perc <- df.r2.matchups.1$opponent_away_perc
df.r2.matchups.2$opponent_home_perc <- df.r2.matchups.1$team_home_perc
df.r2.matchups.2$opponent_away_perc <- df.r2.matchups.1$team_away_perc
df.r2.matchups.2$team_matchup_home_perc <- 1-df.r2.matchups.1$team_matchup_away_perc
df.r2.matchups.2$team_matchup_away_perc <- 1-df.r2.matchups.1$team_matchup_home_perc
df.r2.matchups.2$series_perc <- 1-df.r2.matchups.1$series_perc
df.r2.matchups <- rbind(df.r2.matchups.1,df.r2.matchups.2)
df.r2.matchups <- df.r2.matchups[order(df.r2.matchups$series_perc, decreasing=TRUE),]



#print results
df.r2.matchups
df.all.r2[order(df.all.r2$net,decreasing=TRUE),c(1:4)]
df.sas.r2
df.por.r2
df.okc.r2
df.lac.r2
df.ind.r2
df.was.r2
df.mia.r2
df.bkn.r2