One of the major success stories of the NBA's statistical revolution, has been the league's gradual relinquishing of the Mid-Range shot, in favor of the 3-Pointer. An average 3-Point attempt will yield more points-per-possession then its Mid-Range counterpart, and intuitively, team's are shooting more 3's then ever. However, I believe its still not enough, and that team's could significantly increase their Offensive-Prowess by trading in even more Mid-Ranger's for 3's.
The first reason teams could stand to shoot more 3's, has been well known in the analytic community for years. Relative to Mid-Range jumpers, 3-Point attempts have a significantly higher chance of being rebounded by the offense (as seen by CounttheBasket's heat-map). Similar results have been found by Kirk Goldsberry. Now that's nice, but one of the counter-arguments has always gone something like this. The "average" 3-Pointer may be more valuable then the "average" Mid-Range shot, but if a drastic change in shot-selection occurred, the conditions that produced the above results may change. In short, teams would be trying to force up increasingly contested 3-Point attempts, while discarding open Mid-Range shots. And, as we all know, the best shot in basketball is an open shot. Right? Wrong. Apparently, the worst shot in basketball is an open Mid-Range shot, with both contested 3-Point and Close attempts being more effective. True, this seems to change when the shot-clock winds down to 4 seconds or less, and these league averages are obviously dependent on the quality of player taking the shots, but wow. This graph (via Krishna Narsu) is a somewhat radical shift in what Basketball-Heads consider a "good shot". The league-average FG% on long-Mid-Range jumpers (16-23) in 2013, was 38.3%. When taking into account the value of a 3-Pointer (3 points vs. 2), that's equivalent to 25.5% shooting from the 3. Meaning, even if shooting more 3's resulted in a significantly lower 3-Point%, as long as you were shooting above 25% (which 289 NBA players did last year), the offense wouldn't suffer. When taking into account the benefits of Floor-Spacing and Offensive Rebounding differences, the break-even point could be even lower. That's all for now. Teams are shooting more 3's then ever, and on the surface, it doesn't seem to be enough.
What does "3 and D" mean? "3 and D", refers to a player's ability to play adequate defense, while posing a legitimate 3-Point-Threat. Such player's are often referred to as the cogs of a championship team, the guys who do the little things that help win games. I set out to determine how many NBA players could be considered "3 and D" guys, and of course, who the best of them were.The first step was to tabulate 3-Point-Prowess. This was done in a similar way to my Net Points piece, although I didn't account for assists, as I'm unsure of their effect on 3-Pointers. The general formula for "Net 3's" is below.Net 3's = (3's Made x 3)-(3's Made x 1.08)-(3's Missed x .708)1.08 is the average amount of points scored on a 3-Point Attempt during the 2013 season. .708 accounts for the 27% chance of an Offensive Rebound after a missed 3. Next, I had to tabulate defense. After contemplating a blend of XRAPM, ASPM, and DRTG, I just decided to used D-XRAPM. Its arguably the best publicly available metric for evaluating defense, and allows me to be lazy.To qualify as a "3 and D" player, your standardized Net 3's and Defense (D-XRAPM) score,both had to be above 0. Otherwise 1-dimensional players like Curry and Sanders end up at the top of the list. To qualify for the Top-30 list below, you had to record at least 15 minutes per game, and 41 games played. - Despite weighting Net-3's and Defense equally, and removing one-dimensional players, it looks like 3-Point Prowess still over-influenced the "3 and D" score, relative to Defensive-aptitude. - Only 42 players during the 2013 season qualified as legitimate "3 and D" players, after filtering using the above minute/game restrictions. - The most balanced "3 and D" players (+1 or greater in both Defense and Net 3's) were Paul George, Sefalosha, Durant, and Pierce. That's all folks. The results jive pretty well with what we'd expect, though the amount of "star" players present is somewhat surprising (could be due to my minute restriction). *All stats are minute and pace adjusted, special thanks to NBA.com/Stats and StatsfortheNBA.
One thing NBA fanalysts love to do, is break the game up into parts. Who's the best defender, shot blocker, passer, or scorer, are all questions bandied back and forth, as the more boring question of who the best player/team is, is shunted to the side. I generally agree with this popular sentiment, and as such, set out to rank the best scorers of the 2013 season. To do this, I employed a variation of Dre's Net Points, a formula that determines scoring prowess by subtracting field-goal/free-throw attempts from point totals, to determine the "net" amount of points a player contributed, through scoring alone. I tweaked this a bit, essentially in the spirit of Rosenbaum's Alternate Win Score. I treated all made field-goals/free-throws as "lost/used" possessions (-1), but gave missed field-goals a value of -.72, and missed free-throws a value of -.86, to reflect the possibilities of an Offensive Rebound. The value for missed free-throws was determined by using NBAWOWY to calculate the ORB% on missed free-throws. The value for missed field-goals was simply derived from the league average ORB% rate, though adjusted a bit higher, to compensate for the already counted and lower free-throw ORB%. The next step was to account for assisted-shots. Assists are nebulous, and their effect on shooting-percentages is somewhat controversial. However, after reading a post on Vantage about a 10% increase in FG% for "open" shots, I felt comfortable deducting about .2 points (10%FG) from every assisted shot made. The following is a "correlation table". Observe the legend.ANP-48 = Alternate Net Points per 48 minutes%AST = Percentage of Player's Shots that were AssistedUSG% = Percentage of a Teams Possessions a Player Used
ORTG = Points Produced by a Player per 100 Possessions
TS% = A Player's Raw Shooting Efficiency
Next is a table ranking the Top-30 Scorers of 2013, in terms of Alternate Net Points Per Game (ANP-PG).
- The only non-perimeter player in the Top-30 was Stoudemire. This kind of makes sense, as big-men primarily contribute through rebounding and defense, not direct scoring.- TS% has the largest correlation with Alternate Net Points, and none of the Top-30 scorers have a TS% less then 52%.- %AST and USG% don't seem to have a significant impact on a player's Alternate Net Points rating, at least on the surface.That's all for now. Keep in mind these values reflect scoring prowess relative to the league-average in 2013. We could also tabulate these relative to team-averages, thereby determining the most "valuable" scorers as opposed to the "best" scorers. * All values are pace-adjusted, and yes, I did use the .44 free-throw multiplier during calculations. Special thanks to NBA.com's awesome stats.
In a recent Dwyer article I read, Gary Payton, one of the greatest defenders ever, was complaining about today's NBA. This isn't unusual, as many retired greats spend their free-time denigrating the current-generation of players, and by extension, elevating their own "glory days". Payton's complaint in question wasn't even a new one. Essentially, he was alleging that his "generation" endured physical perimeter-defense without the escape of Free-Throws; an escape today's player's apparently have. I decided to test that claim, or at least a portion of it."Peyton's Postulate" says that the guard of yester-year endured a more physical defense then the guard of today, but was rewarded with equal or less amounts of Free-Throws. To test this, I tabulated the average Free-Throw-Rate (FTR) among perimeter players (PG/SG/SF), for every season dating back to 1980. I plotted out the results, and tried to analyze any obvious trends. * The table is messy. Stare at it long enough, and you will understand. * I Identified 4 significant trends, which I labeled, and addressed in turn below. 1.) 1980-1996 = Steady FTR. This "era" saw the inclusion of the 3-Point Line in 1980, and a medley of rules preventing "zone-defenses". "Physical Defense" was all the rage, with the first check against hand-checking (!) not introduced until 1995, though this seemed to have little effect. 2.) 1996-2004 = Plummeting FTR. This era is infamous for offensive ineptitude and the resulting lack of aesthetic appeal. The falling FTR captures this, as perimeter players struggled to get to the line. This is somewhat strange, as this era was subject to many rule-changes aimed at curtailing hand-checking. However, zone-defense restrictions were also relaxed, which likely led to more advanced defensive-schemes, which likely made it harder for guards/wings to draw fouls for Free-Throw-Attempts. 3.) 2005-2007 = Rising FTR. The FTR for perimeter players soared during this period, which is usually attributed to the NBA's increased emphasis on eliminating the hand-check. I agree with this sentiment, though it also helped that guys like LeBron/Wade/Melo/CP3/Derron/Roy entered the league. 4.) 2008-2013 = Plummeting FTR (Again). This era saw no major rule changes or flux in perimeter talent, so why did the average FTR drop? If I had to guess, I would blame two things. The first, would be the emergence of Thibodeau's "revolutionary" strong-side-zone defensive system, which discourages direct drives to the basket. The second factor, would be the increased use of the 3-Point-Shot, a play that rarely leads to Free-Throws. To recap, I think we've undermined Payton's Postulate. His era (the 90's) seemed to face a more physical perimeter defense, but was rewarded with more Free-Throw's, relative to the player of today. The low FTR for perimeter players post-2007 seems to be partly self-inflicted (higher 3-Point Rate), and partly due to rule-changes (advanced zone defenses). I could probably take this farther by charting 3-Point-Rate and general Offensive Rating for perimeter players, to really flesh-out the changes across eras/seasons. Special thanks to Basketball Reference's Play Index and this list of NBA Rule Changes.
Today I decided to modify David Berri's Win Score (WS) and Dan Rosenbaum's Alternate Win Score (AWS), with a couple SportVu derived metrics. Specifically, I replaced traditional rebounds and assists, with contested rebounds and potential assists, so as to get a different Vu (I had to) of player production. To this point in the season, about half of all potential assists become an assist. Though I would rather try to derive the value of a potential assist from "scratch", I decided to just make the potential assist value +.25 (half of the WS/AWS value of an assist), since this study is Win Score focused (WS/AWS places the assist value at +.5). Since possession based linear metrics like WS/AWS are all about gaining possessions, it made sense to count all contested rebounds as +1. Non-contested rebounds were worth 0, and there was no penalty for missed rebounds. That's because a player trying for and missing a rebound, likely didn't prevent a teammate from trying for and grabbing the rebound (which is the case for missed shots). To be "safe" I added a position adjustment to both WS and AWS, as well as their SportVu derivatives, so I could compare metrics and players without fear of an overt position bias (which I suspect of the SportVu WS/AWS derivatives). I made all scores per-48 minutes, and filtered out players with less then 10-games played, and 20 minutes per game, which left me with a pool of about 210 players. Observe the key.WS48 - Position Adjusted Win ScoreAWS48 - Position Adjusted Alternate Win ScoreWS-VU - SportVu modified Win ScoreAWS-VU - SportsVu modified Alternate Win ScoreBelow are a few correlations (R2 value displayed).
Here are the top 30 players (who qualify) in terms of WS-VU, with their AWS-VU shown on the left.
- The correlation between the WS-VU ad AWS-VU was 91%, which didn't seem low enough to warrant a separate table sorting by AWS-VU.
- The correlations between WS/AWS and their SportVu derivatives were all pretty high. The position adjustment in everything probably helped.
- Derrick Rose ranked last among both WS-VU and AWS-VU.
That's it. Questions and recommendations are always welcome, as I fine tune and try to get this in optimal shape by the All-Star break. Then I'll start seeing how well the Sport-Vu WS/AWS derivatives explain wins, and what type of defensive adjustment I could install to get a Wins-Produced like correlation.
Sorry for the mini-hiatus everyone, I was busy wrestling my Final Exams into submission these last couple of weeks, and simply couldn't find the time to post anything. Things have just settled down, so I thought I'd do a little something on SportVu Rebounding.
Specifically for today, I thought I'd look at the correlations between Offensive Rebounding Rate (ORB%) and Defensive Rebounding Rate (DRB%), vs. Contested Rebound Percentage. I shouldn't have to do this, but the SportVu data at NBA.com doesn't have Offensive/Defensive Rebounding Splits, so I decided to run a couple regressions. Observe the key below, and note that correlation numbers are the R2 term *100.
ORB% = Offensive Rebounding Percentage
DRB% = Defensive Rebounding Percentage
CRB% = Contested Rebounding Percentage
ORB/DRB = Offensive/Defensive Rebounding Ratio
All = All Players
Rebounds = Top 100 Players in Total Rebounds
Minutes = Top 100 Players in Total Minutes
- Predictably, we see the strongest Contested Rebounding correlations with Offensive Rebounding, and visa-versa with Defensive Rebounding. This seems to jive with many NBA metrics, which place a larger importance on Offensive Rebounding then Defensive.
- The correlations when including "All" players in the dataset seem relatively weak, although the ratios still seem alright. This will probably correct itself when we have a larger sample to work with.
- The "Minutes" group showed quite a strong DRB%:CRB% correlation. This could be due to more perimeter players making their way into this dataset (compared to the Big-Men dominated "Rebounds" category), who snag a greater proportion of Defensive Rebounds (many of them contested).
- Jordan Hill leads all players in Contested Rebounds per 48 minutes. (Minimum 12 games and 20 minutes per game to qualify).
- Thus far, about 33% of all Rebounds collected in this NBA season have been "contested". This is a remarkably low number, which we should keep an eye on.
- That's all for now. There are clearly many more layers that can be added to this study, such as factoring in position, height, consistency after changing teams, and so on. It just seems like a waste of time to tabulate all that data, when the season/sample is barely at an respectable level.
Okay, so I finally cracked and decided to do something with the SportsVu data over at NBA.com. I'm starting out small, so as to not waste my efforts on an extremely small sample of data. Today, I decided to look at play-making ability. This isn't a new concept, and is usually done using Assist:Turnover ratios. However, with SportsVu at or disposal, we can take it a little further.
The first step was to determine what percentage of a player's Touches ended in a Turnover (bad) or a Potential Assist (good). This was done by simply dividing total Turnovers/Potential-Assists, by total Touches. Next, I needed to establish the "value" of a Turnover and Potential Assist. A Turnover is is usually counted as -1, since you lost your team a possession, and the average team scores about 1-point-per-possession.
The Potential Assist was a bit trickier, since I don't have access to enough data to come up with an exact value. I decided to start with the Win Score value of an assist (+.5). As of today, 50% of all Potential Assists become Assists, so I divided .5 by 2, to get a value of +.25 for a Potential Assist. Not perfect, but it should suffice for now.I weighted Turnovers and Potential-Assists as described above, added them together, and standardized. Below is a table ranking the top 30 play-makers in the NBA.PAPT% = Potential Assist per Touch Percentage
TPT% = Turnover per Touch Percentage
PME = Play-Making-Efficiency
- To qualify for the table, a player must rank at least average in terms of minutes per game, and touches per game. PME is pace and minute adjusted
- Wade, Harden, Westbrook, and Durant all ranked below average.
- Questions will be answered in the comments, and hopefully I can clean the values/methodology up as the season progresses.
I'm completely swamped with work from my actual job (lab-chemist), and kind of in-between major projects for the site. So, I thought I would cobble together a fun little piece ranking the most "clutch" NBA players from the 2013 season. Clutch, being defined as the last 5 minutes of a game, with neither team ahead or behind by more then 5 points.
We all know about the clutch stats available at 82games and NBA.com, which provide a great record of everything a player does in crunch-time. Unfortunately, neither site really attempts to rank overall clutch play, so I decided to take a crack at it.
I'll be calling this the HDR (Hoop-Don-Rating), which is essentially Alternate Win Score (the best linear-boxscore metric) mixed with Defensive Rating (a defensive adjustment).
The final HDR rating is per-100 possessions (pace adjusted).
To qualify for the top 30, players had to play more then 75 total "clutch minutes" (basically the average or above).
DRTG = Defensive Rating
AWS48 = Alternate Win Score per 48 minutes
HDR per 100 = Hoop-Don-Score adjusted for pace
- LeBron James is the "clutchest" player in the NBA. In fact, the Heat as a team were pretty clutch last year, as evidenced by Chris Bosh, Ray Allen, and Dwyane Wade ranking in the top 30.
- To be fair, multiple teams had multiple players appear in the top 30 (CHI, SAS, MEM, NYK).
- Larry Sanders was apparently #2 in clutch-effectiveness. For the record, Carmelo ranked #57.
- Chris Paul was the 2nd best clutch player via AWS48, but his negative Defensive Rating pulled him down to #17. Sorry CP3.
That's all folks. Questions about methodology will be answered in the comments.
This is the second part of my skilled rebounding study, this time with a focus on Defensive Rebounding. The full methodology is explained here, on the Offensive Rebounding article I last published. I know, I'm lazy, I don't want to go through it again.
The only change was converting the expected Offensive Rebounding Percentages into expected Defensive Rebounding Percentages, which was done by subtracting the ORB% from 100, to get DRB%. The breakdown is below.
Remember, the process is actual rebounds (DRR) minus expected rebounds (XDRR), equals skilled rebounding (XDRD).
DRR=Defensive Rebounding Rate
XDRR=Expected Defensive Rebounding Rate
XDRD=Expected Defensive Rebounding Difference
- Normal DRR had a +8.51% correlation with wins in 2013. The calculated XDRR had a -1.61% correlation, and the "skilled" XDRD had a +15.4% correlation.
- The Thunder, Heat, Bucks, and Blazers were all below average in terms of traditional DRR, but based on where their opponents shot from, they were actually above average rebounding teams.
- The Wizards, Bulls, Wolves, and Clippers were all above average in terms of traditional DRR, but based on where their opponents shot from, they were actually below average rebounding teams.
That's all for now. I'm working on some cool ideas for rating a players "value" to his team, as well as messing around with the new SportsVu stats at NBA.com. I'll tabulate the "skilled" Total Rebounding Rates for 2012-2013, and post them on a tab at the left, along with the previous Offensive/Defensive Rebounding components.
This post, and hopefully future ones, will focus on measuring "skilled" rebounding at the team level. This will be done by isolating and removing the effect that shot locations have on rebounding rates, which ideally, should leave us with a measure of each team's rebounding "skill".
The first step in this process was to pull shot location data for each team, from Hoopdata. I then calculated the percentage of each team's missed shots for each of the 6 major shot locations. This includes misses for free-throws, which were tabulated using team FT% and the .44 multiplier. An example of this data is shown below.
Next, I determined the league average offensive rebounding rate for each major shot location shown above. This was done by looking at the above Grantland chart, this Count the Basket chart, and using NBAWOWY to look at rebound rates of missed 3-pointers and Free-Throws. The tentative values are displayed below.
The expected Offensive Rebounding Rate was multiplied by the percentage of each team's misses by shot location, summed, and standardized. I then subtracted this number from each team's actual Offensive Rebounding Rate (standardized), to arrive at a "skilled" rebounding values.
This process is essentially Actual Rebounds (ORR) minus Expected Rebounds (XORR), equals Skilled Rebounding (XORD).
ORR=Offensive Rebounding Rate
XORR=Expected Offensive Rebounding Rate
XORD=Expected Offensive Rebounding Difference
- Normal ORR had a -.487% correlation with wins in the 2013 season. The calculated XORR had a -3.79% correlation. The "skilled" XORD had a +1.48% correlation.
- The Nuggets were the best Offensive Rebounding team in the league during the 2013 Regular Season, but based on where they took their shots from, they were actually a below-average rebounding team.
- Atlanta and Miami are two teams who were ranked quite low in traditional Offensive Rebounding Rate (ORR), but actually seemed to rebound at a league-average rate, based on where they took their shots.
That's all for now. I plan on doing Defensive Rebounding once I monkey around with the weights a little more, just to make the results as tight as possible.