This is a dirty calculation I put together several years ago for my personal use. I use it to sort through player data and key in on players (both college and pro) that come out with the highest number. After I calculate this number, I go back and look at usage. If a player is a high usage player and they measure out high in my little homemade metric, I start looking at video of them. Then I start digging down to how they stack up against their top competitors, the top teams and the top individual opponents they play against. My problem is I don't know what to call this calculation. It is really a measure of total offensive productivity per total NET possessions used. As you can see I give the player credit for offensive rebounds, steals, and I count their assists as 2 points. So, they get some credit for the D and for their playmaking. I haven't figured out how to incorporate blocks since that is not necessarily a change of possession. However I'm sure I could and should add some sort of credit for blocks (perhaps adding half the blocks to the denominator) as well as more than 2 points credit for assists since a certain percentage of assists are for 3-point shots or for And-1 plays. That's why I'm saying it's a dirty statistical measurement. Maybe somebody here could help me name it and perhaps the more brilliant among us could torque this calculation somewhat to make it better if you can see what it is? Still, even with no changes to it, I feel like this gives me a good starting point of which players have good skill level combined with an aptitude for knowing what is the best play for them to make (basketball iq). It also tells me potential players who may be in a reserve capacity with limited usage who may perform quite well with increased usage. This calculation is why I tend to view guys like Paul Millsap quite favorably and see low risk in signing him to a deal at the right price, even if it is extended years. Because of the way Millsap (and others) play the game and because they have been productive season over season playing different kinds of systems, I tend to believe that guys like him have high bball iq combined with a high skill level and as long as they stay healthy will be productive and will retain trade value over the next few years. In other words, I think they are good "bets" as far as giving contracts to. Here it is for your perusal: [Total points scored + 2(Assists)] / [Total FGA + 0.44(FTA) - ORB + Assists - Steals + Turnovers]
Your metrics is way too complicated to be useful. For something to catch on, it needs to be simple and useful at the same time.
Well, actually the reason I did my own metric is I got tired of trying to dig through the PER and eff metrics and trying to understand them. My calculation is simple compared to that.
I don't know that this is necessarily new, it's just a points per possession metric expanded to look at assists, steals and offensive boards. Like any PPP stat, it's going to favor guys who are efficient off-ball finishers and look down on guys who create tough shots for themselves. I do think this has the advantage of toning down the alleged efficiency of Tyson Chandler style players who score efficiency but never create plays for others, and giving a bit of a boost to Ricky Rubio style players who are weak scorers with good assist numbers. So I'd consider your stat better than standard PPP, but only by a little. And despite the presence of steals, this is an offense only stat that discounts (just like PPP) the usefulness of defensive rebounding, something that is captured in PER and WS (and vastly OVERcaptured in Wins Produced).
Seriously? It's actually really basic. basketballholic, do you have data on drawn offensive fouls? Those are as good as steals. Out of curiousity, how did you decide to use 2 as the scalar for assists? Obviously some made baskets are worth more (suggesting the scalar should be >2), but it also seems somehow wrong to credit assists as being as valuable as made shots (since a player can sometimes get an assist without doing much, which isn't usually the case with a made shot by the same player).
^Correction: Drawn offensive fouls are almost as good as steals, but they don't usually lead to fast breaks.
I would describe it as "points generated per consumed possession". I expect this would underrate defensive bigs, since there's no apparent credit given for blocks or defensive rebounds. You could estimate the point-value of a blocked shot and defensive rebound and add that to the numerator of your formula. Maybe something like 0.65 for defensive rebounds and 1.0 for blocked shots. Edit: I just ran the formula for all players in the past season. It actually is really slanted towards offensive-rebounders and low-usage guys. On Miami, both Haslem and Battier had better ratings than LeBron. Is it that one should only use this metric to compare similar players?
I haven't added defensive rebounds simply because teams are supposed to defensive rebound. If I had some way to measure how much above average a guy defensively rebounded for his position based on how many possessions his team gets, then I would possibly see a reason to add it. But just to throw raw defensive rebounds in the mix....I feel like it would distort too much. I was thinking more along the lines of 0.5 x blocks, allowing for the fact that not all blocks result in a change of possession. And of course, some guys are better than others at keeping the block in play. I feel like blocks are somewhat overrated especially at the NBA level where the best players play off the blocked shot, rebounding their own block and throwing it down or laying it in quite often. Would be interested in a more sophisticated formula for adding and taking away possessions for defensive rebounding if somebody has one.
By the way, points generated per consumed possession describes what I'm after with this calculation quite perfectly.
I wished I did. That would be a good take away to possessions used or if you want to say it differently it would add to the possessions created by offensive rebounds and steals.
Tyson Chandler actually comes out at the top of the charts on this metric also. He's just such a good finisher and offensive rebounder and finisher that he torques this metric also. Using this metric you see the gap between him and a guy like Asik quite easily.
I use it that way after I look at the initial data. Like I say...it's a starting point. And remember I said I look at usage next. I funnel through the players. Miami's bigs can easily be explained. We know they are just cleaning up messes. So, no...it's not an end all be all. However, when you look at .... say ..... big 5s for example. I think it gives a somewhat true picture of what guys are really bringing to the table. For instance, Tyson Chandler is rated higher than Dwight Howard is this year. You gotta account for all those missed free throws that chew up possessions with Howard. In my opinion he simply hasn't been the top center in the league this year. Then when you take into account style of play and you see Dwight is very rarely involved in pick and roll finishes at the rim....which is the single most highly efficient play in basketball....but rather he's catching....dribbling with back to the basket...and he's not an efficient post up player...especially in light of hte poor free throw shooting. Yes, his defensive presence and his defensive rebounding is understated by this metric...but so is Tyson Chandler's.
maybe he should find a way to factor in minutes played, and maybe use ORB(.4) or something to lessen it's effect
By the way, I want to look down on guys that create tough shots for themselves. That's what I call basketball stoooooopid. That encases guys like Rudy Gay and Brandon Jennings, etc. It shows they really aren't that good but they're just burning up a lot of possessions to generate numbers that are empty numbers.
If that's the case, then I would suggest taking a look at Ortg, Drtg, WS, WS/48 stats at basketball-reference.com. They're based on similar concepts of determining point creation and possession consumption. If you want to discard usage (on both ends), then Ortg - Drtg could be what you're after.
I'm jealous of you stat geeks... it's actually one of the subjects I liked in school, but never expanded my knowledge in it.
Could be. But you know what? I have to buy the book, read it, understand what is going on to trust it. That's just not me. Looking at the results, Durvasa, is my metric that far off Ortg - Drtg? Or is it that different from WS/48?
Yeah, its pretty different. For players who played at 1000 minutes this season (265 in total), r^2 between your metric and Ortg-Drtg is 35%, and between your metric and WS/48 its 22%. Here's the top 20 and bottom 20 according to the three metrics (I'll call yours PG/PC): Player Ortg-Drtg Player WS/48 Player PG/PC 1 Tyson Chandler 29 LeBron James 0.322 Reggie Evans 2.51 2 Chris Paul 25 Kevin Durant 0.291 Tyson Chandler 2.31 3 LeBron James 24 Chris Paul 0.287 Andre Drummond 1.98 4 Kevin Durant 22 Tyson Chandler 0.207 Chuck Hayes 1.89 5 Greg Smith 20 James Harden 0.206 Nick Collison 1.89 6 Kosta Koufos 19 Tony Parker 0.206 Festus Ezeli 1.87 7 Nick Collison 19 Tiago Splitter 0.197 Kosta Koufos 1.86 8 Tiago Splitter 18 Marc Gasol 0.197 Greg Smith 1.76 9 Thabo Sefolosha 18 Blake Griffin 0.196 Zaza Pachulia 1.73 10 Marc Gasol 17 Russell Westbrook 0.195 Amir Johnson 1.71 11 Jimmy Butler 17 Dwyane Wade 0.192 Pablo Prigioni 1.69 12 Serge Ibaka 16 Tim Duncan 0.191 DeAndre Jordan 1.67 13 George Hill 15 Brook Lopez 0.191 Jason Kidd 1.65 14 Andre Drummond 15 Deron Williams 0.184 Thabo Sefolosha 1.65 15 Kawhi Leonard 15 Carmelo Anthony 0.184 Kenneth Faried 1.64 17 Shane Battier 15 Serge Ibaka 0.181 DeMarre Carroll 1.62 17 Steve Novak 15 Stephen Curry 0.18 Jimmy Butler 1.62 18 Brandan Wright 14 David West 0.179 Udonis Haslem 1.61 19 Kenneth Faried 14 George Hill 0.177 Ekpe Udoh 1.60 20 Kyle Korver 14 Chris Bosh 0.175 Chris Paul 1.58 ... 246 Dion Waiters -12 Draymond Green 0.028 Avery Bradley 1.15 247 Alexey Shved -13 Dahntay Jones 0.026 Alan Anderson 1.15 248 Brandon Knight -13 Dion Waiters 0.026 Terrence Ross 1.15 249 Dahntay Jones -13 Marquis Daniels 0.023 Arron Afflalo 1.15 250 Richard Hamilton -13 Brandon Knight 0.022 Dahntay Jones 1.15 251 Norris Cole -14 Alexey Shved 0.02 DeMar DeRozan 1.15 252 Marquis Daniels -14 Norris Cole 0.019 Glen Davis 1.13 253 Draymond Green -15 Luc Mbah a Moute 0.017 Rudy Gay 1.13 254 Luc Mbah a Moute -15 Richard Hamilton 0.016 Derrick Williams 1.12 255 Avery Bradley -15 Stephen Jackson 0.013 Eric Gordon 1.12 256 Andrea Bargnani -15 Avery Bradley 0.012 Dion Waiters 1.11 257 Kevin Seraphin -15 Andrea Bargnani 0.008 Byron Mullens 1.11 258 E'Twaun Moore -16 Brendan Haywood 0.006 Stephen Jackson 1.09 259 Byron Mullens -16 E'Twaun Moore 0.006 Richard Hamilton 1.08 260 Stephen Jackson -16 Byron Mullens 0.005 Ben Gordon 1.08 261 Brendan Haywood -17 Kevin Seraphin 0.004 Austin Rivers 1.06 262 Thomas Robinson -17 Thomas Robinson -0.002 Gerald Green 1.03 263 Ben Gordon -19 Ben Gordon -0.018 Andrea Bargnani 1.03 264 Michael Beasley -22 Austin Rivers -0.038 Kevin Seraphin 1.03 265 Austin Rivers -25 Michael Beasley -0.047 Michael Beasley 0.99
According to your metrics: EDIT: James Harden's FGA should be 1337 and his FTA should be 792. My fault. That makes his actual rating based on the metric 1.314374317 which is worse than Jeremy Lin. Cleverly disguised LOF thread. :grin:
I think the flaw of using any formula based metric is that it doesn't control for team's style of play or player roles.