Reb and TmMinorPossessions are both at the league level.
The rest should be right.
LgFTScore is a bit of a mystery. It's .117 in my spreadsheet. Not sure if I accidentally pasted the value over the formula at some point. But using that will match things up.
Haha, well that is an interesting response. Do you use that same value for LgFTScore for all seasons? Just curious how you calculate WARP for other seasons/players since there seems to be some mystery involved although I can't imagine that specific factor has much of an overall impact (whether it's .117 or something else).
What type of non-scoring possessions is it supposed to represent?
Reb and TmMinorPossessions are both at the league level.
The rest should be right.
LgFTScore is a bit of a mystery. It's .117 in my spreadsheet. Not sure if I accidentally pasted the value over the formula at some point. But using that will match things up.
Ok, I think I have it. Posting for any others trying to calculate:
Ok going to see if I can work my way back on one of these guys to see where I went wrong. Let me know which input is bad in this calculation for Chalmers:
I think we're good on OppRebOpp% (.4686), RebFactor3 (.8861) and RebFactor2 (1.121).
So the only inputs that could be bad would be TmDefRatMinor (91.138) or TmDReb% (.7152), and then obviously something within those calculations is bad.
I have figured the correct formulas to get defensive rating. The results correspond to what Kevin posted and also match the attachment at the beginning of this post. My offensive ratings match what jkeen39 has at the moment. Any help figuring out what we have done wrong would be appreciated. This has been a fun math puzzle. Looking forward to getting the offensive ratings figured out. The following is an example using Lebron James.
These are the inputs I have for Mario Chalmer's Offensive Rating. Any help in determining where I went wrong would be appreciated!
AstFGM% - An estimate of what percentage of Field Goals Made were Assisted
=(TmAst/TmFGM)*(1.53-1.442*((Ast/MP)/(TmAst/(TmMP/5)))-(.041*((ORB/MP)*48))-(.787*Usage1)+(.014*(3PA/FGA)^2*((1/Usage1)^2)))
=(1317/2446)*(1.53-1.442*((222/1825)/(1317/(16040/5)))-(.041*((16/1825)*48))-(.787*.174715843)+(.014*(260/498)^2*((1/.174715843)^2)))=0.577727594
AstFGM - An estimate of how many Field Goals Made were Assisted
=AstFGM%*FGM
=.577727594*223=128.8332535
Points Created (PC) - An estimate of offensive points created
=Points+(.75*Ast)-(.75*AstFGM)
=627+(.75*222)-(.75*128.8332535)=696.8750599
Offensive Possessions (OPos) - An estimate of the number of individual possessions
=FGA+(.44*FTA)+TO+(.375*Ast)-(.375*AstFGM)
=498+(.44*101)+142+(.375*222)-(.375*128.8332535)=719.3775299
TmOffRatMinor - An estimate of the Offensive Rating for the player and four league average players
=(OffRat*Usage1)+((1-Usage1)*(LgOffRat+(.25*(Usage1-.2)*100)))
=(96.87195262*0.174715843)+((1-0.174715843)*(91.45028988+(.25*(0.174715843-.2)*100)))=91.87587488
TmOReb% - An estimate of the imaginary team's Offensive Rebound %
=100*((OReb%/100)+ ((4*(LgOReb%)/100))/(5-(LgOReb%)/100))*(1-(OReb%/100))))
=100*((1.088847523/100)+((4*(26.95628226/100))/(5-(26.95628226/100))*(1-(1.088847523/100))))=23.63456028
PlayerRebOpp% - An estimate of the number of available rebounds for the player
=((FGA-FGM)+(0.56*(FTA-FTM)))/(FGA+(0.44*FTA)+TO)
=((498-223)+(.56*(101-80)))/(498+(.44*101)+142)=0.418970253
TeammateOffRat - An estimate of the Offensive Rating for the four league average players
=LgOffRat+(0.25*(Usage2-0.2)*100)
=91.45028988+(.25*(.183634287-.2)*100)=91.04114706
Teammate FGA% - An estimate of the percentage of Field Goal Attempts for the four league average players
=(1-((LgTO/LgPos)*(1+(0.5*(0.2-Usage1))))-(LgFTA/LgPoss)*0.44)
=(1-((27620/208412.68)*(1+(.5*(.2-.174715843))))-(44472/208412.68)*.44)=0.771909971
Teammate eFG% - An estimate of the eFG% for the four league average players
=((TeammateOffRat/100)-(LgFTM/LgPos))/(Teammate FGA%*2)
=((91.04114706/100)-(33465/208412.68))/(.771909971*2)=0.485704716
Teammate FG% - An estimate of the FG% for the four league average players
=Teammate eFG%/(Lg eFG%/Lg FG%)
=.485704716/(0.487297255/0.447933013)=0.446469121
TeammateRebOpp% - An estimate of the number of available rebounds for the four leauge average players
=(1-Teammate FG%)*Teammate FGA% +(0.56*((LgFTA-LgFTM)/LgPos))
=(1-.446469121)*.771909971+(.56*((44472-33465)/208412.68))=0.456851556
TmRebOpp% - An estimate of the imaginary team's ability to rebound
=(PlayerRebOpp%*Usage1)+(TeammateRebOpp%*(1-Usage1))
=(.418970253*.174715843)+(.456851556*(1-.174715843))=0.450233092
RebFactor1 - Adjustment to take account for the fact that every missed shot does not result in a rebound
=LgReb/((LgFGA-LgFGM)+0.56*(LgFTA-LgFTM))
=83513/((161225-72218)+.56*(44472-33465))=0.87750544
RebFactor2 - Adjustment to take account for possessions that include more than one second chance
=(LgOffRatMajor-LgOffRatMinor)/(LgOffRatMinor*(LgOReb/LgPos))
=(102.5246384-91.45028988)/(91.45028988*(22512/208412.68))=1.121096918
TmOffRat - An estimate of the Offensive Rating for the imaginary team
=TmOffRatMinor+(TmOReb%*TmRebOpp%*RebFactor1*RebFactor2)
=91.87587488+(23.63456028*.450233092*.87750544*1.121096918)=102.3442172
TmOffRat3 - An estimate of the offensive rating for the imaginary team with an adjustment for floor spacing
=(TmOffRat+(7*(3A%-Tm3A%)))/0.96
=(102.3442172+(7*(0.379872597-0.149576249)))/.96=108.2878038
First, there are a couple of places in the teammate section where the formula calls for Usage1 and you're using Usage2. That one's pretty minor. The bigger difference is because in the TmOffRat that whole section needs to be multiplied by TmOffRatMinor.
I've hand calculated Kubatko's Win Shares countless times. I notice many similarities, the obvious, ORtg, DRtg, Accounting for Usage, % of Tm MP, Accounting for Lg FT's, Defensive Stops, Attempting to account for individual defensive impact on Opponent FG Att, without me looking too much into the nuts and bolts at this very moment, what are the major differences between WS and WARP ?
Yeah, WARP and Win Shares both have a common Dean Oliver lineage. Three main differences I see:
1. WARP is the only non-SPM system that attempts to account for floor spacing with the adjustment for 3-point attempts. In fact, this is something of an SPM element since it came from comparing WARP results to adjusted plus-minus.
2. Win Shares credits all team defense to individuals (and, generally, is constrained by team performance). So players on good defenses look better by Win Shares and those on poor defenses look better by WARP -- particularly big men.
3. Lastly, while I couldn't articulate exactly how Win Shares handles usage, it's my experience that there isn't nearly as big an adjustment for it as in WARP. WARP falls somewhere between PER and Win Shares on this continuum (and also in terms of team defense).
When accounting for replacement level, how is this calculated? It says, generally around .415, I believe? How greatly does this vary year-to-year?, would .500 make for 'Wins Above Average' ?
It's set at whatever creates 930 total WARP around the league, matching up with a replacement level of 10 wins per team per season. Matches up pretty well with what I've found when I look at actual replacement-level players, though Daniel Myers' studies have shown slightly higher replacement level.
Yeah, there's an ancient wins above average column in my spreadsheet but I've never used it.
I clicked on the Basketball Prospectus link and saw it was updated in 2012. I didn't recall that thorough of an explanation of how exactly marginal wins were arrived it. Knew it was the Pythagorean theorem, the constant ^14 exponent, and that it was Individual WPct minus .415, I just didn't recall the detailed explanation that followed, I had an older version saved. I'm just curious now that you've better explained it, how much of a difference would come from your calculations and J. Kubatko's, if you simply plugged in his O & DRtg's, and compared them to WARP, by your explanation, this would give me a sense of how many wins came solely from a player's ability to space the floor, and say a big's rim protection or a wing's ability to disrupt an opponent's shot ? Given that, like you said, Win Shares allocates all team defense to individual players. I didn't remember seeing the floor spacing component either.
OK, this is driving me insane and I can't take it anymore. I've been trying to replicate this today and I have hit a snag. I'm hoping someone can quickly help me.
I can't figure out where LgOffRatMajor comes from. I've read this whole thread and KP's original article and I just can't figure it out.
I'm stuck on RebFactor2. Any help would be greatly appreciated.