Joined: 14 Jan 2005 Posts: 971 Location: Delphi, Indiana
Posted: Sun Sep 10, 2006 5:56 am Post subject:
Bob wrote:
Quote:
how many eWins do you have for players like:
derek harper in either 86-87 or 89-90?
maurice cheeks in 80-81, 81-82, 84-85, 85-86, or 86-87?
michael williams in 91-92?
Derek Harper's eWins chart like this:
yr tm eW
84 Dal 1.6
85 Dal 4.4
86 Dal 5.5
87 Dal 8.0
88 Dal 8.6
89 Dal 8.3
90 Dal 9.4
91 Dal 8.6
92 Dal 5.6
93 Dal 4.5
94 Dal 1.0
94 NY 2.9 -- 3.9 for the year
95 NY 5.0
96 NY 5.6
97 Dal 3.5
98 Orl 2.4
99 LA 1.5
Double the eW number to find the Wins added to the team, compared to all his minutes being taken by 'replacement-level' players.
Joined: 27 Jan 2005 Posts: 423 Location: cleveland, ohio
Posted: Mon Sep 11, 2006 9:47 am Post subject:
your first list had rod strickland of 97-98 with 10.1 eWins playing 3020 minutes. i didn't see him on your second list but am assuming his eWins were still something just above or below 10. you have derek harper in 86-87 with 8.0 eWins playing 2556 minutes. i don't know if eWins can be normalized to minutes but if so 8.0 x (3020/2556) = 9.5 eWins for harper had he played strickland's minutes, still less than strickland's eWins...
here are their stats, with the second column simply harper's 86-87 stats normalized to strickland's 97-98 minutes played (multiplied by 3020/2556):
assume for arguement that both played in 77 games (what strickland played in 97-98 )...
in 86-87 dallas opponents got def rebs at 67.2%, so:
(586 + 87/2) x (.672) + 163 = (629) x (.672) + 163 = 423 +163 = 586
or 586/77 = 7.6 zero point team possessions per game
this is a calculation to approximate how many zero point team possessions harper was responsible for by missing FGAs (586) and FTAs (87) that were rebounded by dallas opponents plus his turnovers (163). doing the same for strickland:
(640 + 135/2) x (.676) + 266 = (707) x (.676) + 266 = 478 + 266 = 744
or 744/77 = 9.7 zero point team possessions per game
so off the bat strickland is personally responsible for his team not scoring on an additonal (9.7-7.6)x(77) = 2.1 x 77 = 162 team possessions or slightly more than 2 team possessions per game. since each team possession can be shown to have an average value of about 1 point, that'd be like 162 less points for strickland compared to harper...
on top of this harper was all-D 2nd team in 86-87. i would say he was a better defender in 86-87 than strickland was in 97-98 in terms of forcing missed shots (forced misses of which about 67.2% would be rebounded by dallas), or in terms of forcing more turnovers, which would mean a few more overall zero point team possessions he would force than would strickland (this is an assumption of course)...
knowing this could strickland in 97-98 really produce more wins than harper in 86-87, on a per minute basis? in 86-87 he did play 15% less minutes than strickland did in 97-98, but if they did play the same minutes would eWins still show strickland with more? strickland did get (normalized to his 97-98 minutes) about 2.2 more reb/g - would that be enough to produce more eWins than harper?...
I don't see their assists as being this close (801-720, if minutes are equal). I've got Strick at 10.4, Harper at (a career-best) 7.6 that year, per-36-min/standardized. So my short answer would be that Strick is twice the rebounder and 1/3 again the playmaker -- as well as being equal the scorer that Harper is.
I see the '87 Mavs gave up 110.4 ppg! -- while the '98 Bullets gave up a mere 96.6. So Harper's offensive numbers (as I figure it) are inflated by 14.3% relative to Strickland's. If we accept this adjustment, we can imagine 'equivalent assists' and points, as if they both played in the same 'scoring milieu' (as well as playing the same minutes).
So if the players could be interchanged thru time and team, Strickland might get 801*1.143 = 915 assists running with the '87 Mavs. Then, Harper would only have gotten 609/1.143 = 532 assists, playing 2556 minutes with the '98 Wiz (or was it Bullets?)
Most APBRmetricians accept that [Rebounds/100 Available Reb] is a better measure of 'Rebounding' than 'Reb/G' or 'Reb/48'. It never fails to surprise me when a colleague treats Pts/Min (and Ast/Min) the same, whether they were accrued in 95-pt games or 115-pt games. _________________ 40% of all statistics are wrong.
Last edited by Mike G on Tue Sep 12, 2006 9:01 am; edited 1 time in total
Joined: 14 Jan 2005 Posts: 971 Location: Delphi, Indiana
Posted: Tue Sep 12, 2006 9:00 am Post subject:
Oh yeah,
Derek Harper's career year (in my book) is 1990, when he had 2.88 eW/1000 minutes. Rod Strickland had 6 seasons better ('94-'99), topping at 3.63 in Portland/'95.
Please don't hold me to these figures. I've actually referred to 2 different (eWins) files in this thread. One of them allows weights of statistical categories to vary year by year. The other applies equally to all seasons. The differences generally aren't great. _________________ 40% of all statistics are wrong.
Joined: 27 Jan 2005 Posts: 423 Location: cleveland, ohio
Posted: Tue Sep 12, 2006 1:42 pm Post subject:
I don't see their assists as being this close (801-720, if minutes are equal).
correct, they are not. that was a back of the envelope calculation. when i simulate both strickland 9798 and harper 8687 on both the mavs of 8687 and the wizards of 9798 harper gets approximately 1/3 less assists playing the same minutes than does strickland...
I've got Strick at 10.4, Harper at (a career-best) 7.6 that year, per-36-min/standardized. So my short answer would be that Strick is twice the rebounder and 1/3 again the playmaker...
but the question is how do those rebounds and assists contribute to winning? evidently your eWins are saying alot...
...as well as being equal the scorer that Harper is.
depends on what you mean by equal - equal total points or pts/g scored perhaps? because with significantly better overall shooting (higher ScFG% of 55.4% to 49.0%) and similar turnovers per touch harper was the more efficient scorer. more missed shots means more defensive rebounds for the opposition and more zero point team possessions for your team...
the league average ScFG% in 86-87 was 52.8%, in 97-98 it was 51.5%, a difference of 1.3%. but the difference in overall shooting (ScFG%) for harper vs. strickland in this case is 6.4%....
I see the '87 Mavs gave up 110.4 ppg! -- while the '98 Bullets gave up a mere 96.6. So Harper's offensive numbers (as I figure it) are inflated by 14.3% relative to Strickland's.
not sure why you are looking at points given up on defense as a parameter to measure offensive contributions by, but the team possessions per game for dallas in 86-87 was 102.1 poss/g and for the wizards in 97-98 was 93.0 poss/g, for a difference of 9%...
So if the players could be interchanged thru time and team, Strickland might get 801*1.143 = 915 assists running with the '87 Mavs. Then, Harper would only have gotten 609/1.143 = 532 assists, playing 2556 minutes with the '98 Wiz (or was it Bullets?)
yes, when adjusted for pace strickland would get more assists on the 86-87 mavs playing similar minutes and harper less on the 97-98 wizards...
Most APBRmetricians accept that [Rebounds/100 Available Reb] is a better measure of 'Rebounding' than 'Reb/G' or 'Reb/48'. It never fails to surprise me when a colleague treats Pts/Min (and Ast/Min) the same, whether they were accrued in 95-pt games or 115-pt games.
correct - its always better to frame their context, such as:
1) how a player's rebounding rate helps/hurts a team when their shooting is also considered...
2) how a player's points scored affects his team based on his scoring efficiency, and looking at all of their missed scoring opportunities...
3) how a player's assists actually contribute to winning...
1) in this example strickland is clearly the better rebounder, but also clearly the worse overall shooter. simulating both harper 86-87 and stickland 97-98 on the 86-87 mavs and 97-98 wizards, in both cases playing 40 min/g and 82 games the team's ScFG% was higher on offense and lower on defense (allowed) with harper. the differences were a combined total of +1.7% (more on offense and less on defense) in favor of harper, because harper was the better overall shooter, and better defender...
2) again in both cases harper increased the team's offensive points per team possession scored and decreased the team's defensive points per possession allowed more than strickland did, increasing the team's offensive points per possession scored by an average of 2.5% more than strickland did, and decreasing the team's defensive points per possession allowed by 2.0% more than strickland did...
3) assists - i noticed andre miller's 01-02 season (when he got 882 assists) also rated high on your eWins list. the question is how did those assists actually contribute to winning? in 01-02 the cavs went just 29-53, so my question is just what is the eWins distribution for the 01-02 cavs players?...
that 01-02 season miller got a high number of total assists not because he was a much better "passer", or to be more accurate i should say "win generator", that season than in other seasons. he did get his highest 0.292 ast/min that season, but his next best ast/min in a season was about 0.230 ast/min, and he did that 3 times (99-00, 00-01, and 05-06) - those rates are all almost 1/4 less, no where near that one outlier season...
the reason miller got a very high number of total assists that season was because the cavs had 3 starters (chris mihm, wesley person, and jumaine jones) who all had significantly lower touches/min than the average player at their position in the league that year (all on the same team and starting), who all 3 also happened to shoot the ball with a high percentage of their touches (close to 50% for all 3). thus more touches for miller and more opportunities for assists. regardless of the FG% they shot (unless it was truly awful), the PG would get a relatively high number of total assists, versus being on the vast majority of other teams. as it was miller got 47% of the team's total assists...
Double the eW number to find the Wins added to the team, compared to all his minutes being taken by 'replacement-level' players.
so in this case what eWins did miller generate versus the eWins of his 01-02 cavs teammates? he played only 15% (3023/19855) of the team's total minutes...
when you calculate eWins, can you calculate them for team totals? i'd be curious if whatever credit you are assigning to assists, when applied to team total stats (since assists and FGM on the same FGM would be counted), rates total team eWins in the same order as team winning percentage...
Derek Harper's career year (in my book) is 1990, when he had 2.88 eW/1000 minutes. Rod Strickland had 6 seasons better ('94-'99), topping at 3.63 in Portland/'95.
so are you saying those 6 seasons were each better than any of harper's?...
here are some stats for strickland's (93-94 to 98-99) and harper's (85-86 to 90-91) best 6 year straight runs, normalized to 82 games:
again the 0ptposs is zero point possessions each player is responsible for, by missing FGAs and FTAs rebounded by the defense and by committing turnovers. strickland commits 2 more per game, or 164 for an 82 game season. that's 164 zero point possessions versus 190 more assists. since a team possession can be shown to be worth about 1 point, that's like 164 more points scored (or saved from being scored). can you show 190 assists is worth 190 points? (again the big question of what is the true value of an assist?)...
plus not to mention 30 more steals for harper per 82 games (30 more zero point opponent team possessions he causes than does strickland), plus whatever more zero point team possessions he forces with better defense than strickland (forcing more misses or turnovers)...
simulation shows on a 40 min/g and 82 game basis harper's 86-87 season generating about 7-8 more wins per average 82 game season than strickland's 97-98 season (harper rated as a better defender than strickland in terms of FG% allowed) - that's a huge difference. plus for the above stats for each player's six year runs, and with each player rated a 0.0% on defense (neither increases or decreases the FG% of the players they guard), simulation shows the stats of harper on a 40 min/g and 82 game basis generating 5-6 more wins per average 82 game season...
Maybe I missed this but did Strickland have more or less touches than Harper. Is it possible he had "took" turnovers from other players in the same way Miller took assists, thus the 2 extra ZPP would have been turnovers anyway?
Joined: 14 Jan 2005 Posts: 971 Location: Delphi, Indiana
Posted: Wed Sep 13, 2006 6:37 am Post subject:
...when i simulate both ... harper gets approximately 1/3 less assists playing the same minutes than does strickland...
Same result I get when scaling Ast rate to (Tm+Opp)PPG.
...with significantly better overall shooting (higher ScFG% of 55.4% to 49.0%) and similar turnovers per touch harper was the more efficient scorer...
With Strick hitting FT at almost twice the rate of Harper, their TS% are closer -- .561 to .501
Here's where they place on their respective teams, listed in order of PPG:
Code:
'87 Mavs PPG TS% '98 Wiz PPG TS%
Aguirre 25.7 .560 Webber 21.9 .523
Blackman 21.0 .569 Howard 18.5 .515
HARPER 16.0 .561 STRICKLAND 17.8 .501
Perkins 14.8 .545 Murray 15.1 .563
Donaldson 10.8 .658 Cheaney 12.8 .488
Schrempf 9.3 .560 Eackles 5.2 .542
Tarpley 7.5 .501 Whitney 5.1 .512
B Davis 7.0 .564 T Davis 4.4 .524
A Wood 6.6 .482 B Wallace 3.1 .495
Both are their teams' #3 scorer. Strick is a bigger part of a lesser offense. But both their TS% are very close to the team average (slight edge to Harper).
not sure why you are looking at points given up on defense as a parameter to measure offensive contributions by, but the team possessions per game for dallas in 86-87 was 102.1 poss/g and for the wizards in 97-98 was 93.0 poss/g, for a difference of 9%...
Well, a team and its opponents play the same number of possessions. Since
Poss/G * Pts/Poss = Pts/G
I just use the more straightforward Pts/G
Opponent PPG represents what the NBA-at-large scores in games your team plays.
...i noticed andre miller's 01-02 season (when he got 882 assists) also rated high on your eWins list. the question is how did those assists actually contribute to winning? in 01-02 the cavs went just 29-53, so my question is just what is the eWins distribution for the 01-02 cavs players?...
Code:
10.3 Miller
5.4 Murray
4.9 Person
3.9 Davis
3.6 Ilgauskas
3.5 Jones
2.1 Mihm
.8 Skinner
1.1 Hill
.7 Langdon
.5 Stith
.5 Doleac
.3 Coles
These add up to 38 eW. That being 3 games below .500, eWins 'predicts' this team to finish 6 below: that is, 35-47.
...
the reason miller got a very high number of total assists that season was because the cavs had 3 starters (chris mihm, wesley person, and jumaine jones) who all had significantly lower touches/min than the average player at their position in the league that year (all on the same team and starting), who all 3 also happened to shoot the ball with a high percentage of their touches (close to 50% for all 3).
What you're describing indeed inflates the team's eWins (perhaps to just about what I got for the '02 Cavs, 5-6 wins too many) by the fact that I fail to estimate _Unassisted % of FG_. These are now available (thru 82games.com); and the effect is to reduce the value placed on these 'catch-and-shoot' players, and allow more credit to the assister.
The 3 players you mention have never, as far as I know, been starters on any good team. That all 3 could play major minutes provided they had an Andre Miller on the floor suggests the Cavs only fielded a credible team due to this one player.
Subtract Miller from that team -- putting 'replacement level players' in his minutes -- and eWins supposes the Cavs win about 21 fewer games. The rest of that lineup just might be one of the worst teams ever. (Z only available 1300 minutes)
What does the sim say? _________________ 40% of all statistics are wrong.
Last edited by Mike G on Wed Sep 13, 2006 7:33 am; edited 1 time in total
Joined: 14 Jan 2005 Posts: 971 Location: Delphi, Indiana
Posted: Wed Sep 13, 2006 6:48 am Post subject:
While using basketball-reference.com for my previous post (and neglecting to credit, again), I noticed Rod Strickland was named 2nd-team All-NBA for that 1998 season. He joins a handful of players who were all-League without making the Allstar team that year.
Neither Strick nor Harper were ever named as Allstars. _________________ 40% of all statistics are wrong.
Joined: 27 Jan 2005 Posts: 423 Location: cleveland, ohio
Posted: Thu Sep 14, 2006 6:38 am Post subject:
Maybe I missed this but did Strickland have more or less touches than Harper.
about 25% more touches per minute...
Is it possible he had "took" turnovers from other players in the same way Miller took assists, thus the 2 extra ZPP would have been turnovers anyway?
not sure i understand your reasoning here - in the past 30 seasons about 20 players passed for similar total assists in a season (say 30 more or less) as strickland did that year, but some with as many as 25% more turnovers, some with 25% less turnovers, and with an overall shooting (ScFG%) range of between 46% to 62%. in either case that's a big difference in zero point possessions personally responsible for yet with essentially the same number of total assists...
Subtract Miller from that team -- putting 'replacement level players' in his minutes -- and eWins supposes the Cavs win about 21 fewer games. The rest of that lineup just might be one of the worst teams ever. (Z only available 1300 minutes).
not sure what a replacement level player is, but the 01-02 cavs went just 29-53. is eWins saying that if miller and his minutes were replaced on those cavs by one of the worst PGs statistically that season (say someone like jamaal tinsley of indiana or chris childs of toronto) that the team would have won 21 fewer games to go just 8-74?...
Joined: 14 Jan 2005 Posts: 971 Location: Delphi, Indiana
Posted: Thu Sep 14, 2006 7:25 am Post subject:
One of the great things (to me) about eWins is that it 'defines' the value of the replacement player as one who adds no (or negative) wins. This is a player who could be got on the waiver wire, or from the CBA, or wherever.
The '02 Cavs had Bimbo Coles. By my shorthand, he might only add 1.5 wins playing 35-40 mpg. You can give him 36 and the other 12 to Jeff Trepagnier. Since I estimated 35-47 for the team's 'shoulda won', I now predict they drop to somewhere between 9 and 14 wins. In the sim, that is.
I don't know that anyone defines 'replacement value' as that of a starting NBA player. The idea is that one of these players can be had at any time, on just a couple days notice. Generally, if your starter goes down, his backup moves into the role, the benchwarmer becomes the backup, and you get the 'emergency' guy on a short contract. He's the RP.
Also: How do you deal with players missing games? A lot of Cavs missed 20-40 games that year. Do you allot Tyrone Hill (26G @ 31mpg) or (82G @ 10mpg)? _________________ 40% of all statistics are wrong.
Joined: 27 Jan 2005 Posts: 423 Location: cleveland, ohio
Posted: Thu Sep 14, 2006 8:44 am Post subject:
One of the great things (to me) about eWins is that it 'defines' the value of the replacement player as one who adds no (or negative) wins.
the 5 starters on that 01-02 cavs team that went 29-53 were andre miller, wesley person, lamond murray, chris mihm, and jumaine jones. the bench consisted of ricky davis, zydrunas ilgauskas, brian skinner, tyrone hill, michael doleac, bimbo coles, bryant stith, and trajan langdon. if you replaced each of these 5 starters on the 01-02 cavs with replacement players, and each replacement player played say 36 min/g, what does eWins say the team's W-L record would be?...
Maybe I missed this but did Strickland have more or less touches than Harper.
about 25% more touches per minute...
Is it possible he had "took" turnovers from other players in the same way Miller took assists, thus the 2 extra ZPP would have been turnovers anyway?
not sure i understand your reasoning here - in the past 30 seasons about 20 players passed for similar total assists in a season (say 30 more or less) as strickland did that year, but some with as many as 25% more turnovers, some with 25% less turnovers, and with an overall shooting (ScFG%) range of between 46% to 62%. in either case that's a big difference in zero point possessions personally responsible for yet with essentially the same number of total assists...
The point being, that basketball is a team game and result of entire team possession is what is important. Your logic, which is sound, is that Miller got higher assits on the Cavs primarily due to the lack of other possession users on the team. Likewise, my question, was did Strickland get more turnovers than Harper due to his higher relative possession use rather than his personal worse ballhanding/passing skills. Maybe you already adjusted for this in an earlier post and thus, it seemed obvious that this was a silly question.
If a PG handled the ball 25% more often, wouldn't it make sense - all else equal - that he would turn the ball over more often but not neccessarily account for more overall ZPP, if he was less turnover prone than the remainder of his teamates. He might even make the team more efficient if the turnover rate of the other players was higher than his, no?
Joined: 14 Jan 2005 Posts: 971 Location: Delphi, Indiana
Posted: Sat Sep 16, 2006 9:22 am Post subject:
bchaikin wrote:
the 5 starters on that 01-02 cavs team that went 29-53 were andre miller, wesley person, lamond murray, chris mihm, and jumaine jones. the bench consisted of ricky davis, zydrunas ilgauskas, brian skinner, tyrone hill, michael doleac, bimbo coles, bryant stith, and trajan langdon. if you replaced each of these 5 starters on the 01-02 cavs with replacement players, and each replacement player played say 36 min/g, what does eWins say the team's W-L record would be?...
Well, they would be really bad.
For one thing, those 5 starters only averaged 29 mpg. So the subs averaged 19 mpg. Now you want to give RP's 36 mpg, reducing the subs to 12. If I reduce the eWins of the subs mentioned above (RDavis et al) by a factor of [*12/19], they produce a total of about 9 eWins.
This on top of Zero added by your 'starting 5' (which is what RP's do) suggest a team W-L of [9*2 - 41 =] negative-23 wins.
Since you've deliberately asked for an implausible scenario -- getting rid of your best players, and playing your worst players 3/4 of the game -- you might expect an unrealistic answer. A team 'projected' to win -23 G has some nonzero chance of winning 1 or more.
I wouldn't want to be a member of the team that loses to this one.
Meanwhile, your lineup suggestion brought to mind the 1993 Dallas Mavericks, who finished (with a flourish) 11-71. They started some 'CBA types' like Walter Bond, Mike Iuzzolino, Brian Howard.
Derek Harper missed 20 games but still led the team with some 4.5 eWins. The team totalled 24, suggesting a season of (24*2 - 41 =) 7 wins. Without Harper, might they not have won a game? In fact, they were 8-54 with Harper and 3-17 without him.
Most likely there is a better eW/W conversion formula, that doesn't break down at these extremes. _________________ 40% of all statistics are wrong.
Joined: 27 Jan 2005 Posts: 423 Location: cleveland, ohio
Posted: Mon Sep 18, 2006 9:08 pm Post subject:
Subtract Miller from that team -- putting 'replacement level players' in his minutes -- and eWins supposes the Cavs win about 21 fewer games. The rest of that lineup just might be one of the worst teams ever. (Z only available 1300 minutes)...
again that takes the team from 29-53 to 8-74. miller played 3023/19855 minutes, just 15% of the team's total minutes played. is this saying that the other 14 players on the team, with all the stats they produced (the vast majority of the team's total rebounds, points, steals, blocked shots, etc), playing 16832/19855 or 85% of the team's total minutes played, eWins is saying produced only 8 total wins compared to 21 for just one player?...
Joined: 14 Jan 2005 Posts: 971 Location: Delphi, Indiana
Posted: Tue Sep 19, 2006 6:00 am Post subject:
Player eWins do not add up to the team Wins . When you sum player eW, you can expect
W = eW*2 - 41
to be their season win total.
Upthread, I wrote:These add up to 38 eW. That being 3 games below .500, eWins 'predicts' this team to finish 6 below: that is, 35-47. (This is another way of figuring 38*2 - 41 = 35.)
The Cavs other than Miller totalled 28 eWins. 28*2 - 41 = 15 wins from a Millerless '02 Cavs. This is apparently an inflated estimate, as discussed above regarding credit to assisted-scoring players like Person, Mihm, Jones... I don't think it was Andre getting too much (eW) credit that year.
When Andre left the team, in 2003 the Cavs featured Davis for 40 mpg, had Ilgauskas for almost twice his '02 minutes, added Carlos Boozer and Darius Miles (replacing Person and Murray)-- and dropped to 17-65. Those guys played full seasons that year, and somehow the loss of Miller was coincident with 12 fewer wins.
The Pythagorean-expected records w/wo Miller were 31-51 and 16-66, respectively. Change out those '02 Miller minutes with replacement players', and it's not much further to an 8-74 record.
I asked what your sim would indicate with these scenarios. If you don't have '02, you might try the Timberwolves without Garnett, '05 or '06. Give his minutes to Mark Madsen, (or to anyone else on the team) and they drop to historic wretchedness. _________________ 40% of all statistics are wrong.
All times are GMT - 5 Hours Goto page Previous1, 2, 3, 4Next
Page 3 of 4
You cannot post new topics in this forum You cannot reply to topics in this forum You cannot edit your posts in this forum You cannot delete your posts in this forum You cannot vote in polls in this forum
FAQ
Search
Memberlist
Usergroups
Register
Profile
Log in to check your private messages
Log in
