APBRmetrics

colts18 wrote:

ryannow wrote: The numbers still just don't add up. I ran the exact same equation, but with the actual total minutes played for this year's Spurs, and I came out with a value of 10.2. Did it for the Warriors and got 7.2. Both of those are more than 2 points higher than the team's actual SRS in that time period. It makes sense why this would happen: if you add one +4 player to a team at 0, you can expect them to go to around +4, but that doesn't mean a second one would lift the team to +8, or a third one to +12. That's what the "effect of being up x" term is attempting to capture; it's harder to take a team that's already very efficient and boost it by n points than it is to take an inefficient team and boost it by n points. As a result, you can take +10 worth of players, and end up with a +8 team, although I'm not sure exactly what the shape of this curve is in J.E.'s model.

You ran the projection and got +10.2 for the Spurs because though are actual minutes. I'm using playoff strength. The Spurs aren't playing many scrubs in the playoffs so their strength is higher than +10.2. Plus J.E. already said that his xRAPM takes into account the leading by X effect.

Let me elaborate, as I think my point wasn't clear the first time. I'm not trying to argue that the Spurs' playoff strength estimate should be 10.2. My point is that summing xRAPM gives systematically higher values than SRS, at least for playoff teams, and the two cannot be treated as equivalent. The values we are seeing from summing xRAPM come in higher than SRS values should come in, and this isn't explained simply by the change in minutes. To illustrate this, I calculated the sum of xRAPM values using the exact minute weightings we had over this past season, and found that the xRAPM values summed to a couple points higher than the actual SRS of that season, at least for the two teams I checked. If the sum of minute-weighted xRAPM could be treated as an SRS estimate, this should've been equal, or pretty close to it.

I recognize that it takes into account the leading by X effect, and in fact I believe that's the reason why this occurs. The difference between +0 and +1 is less than the difference between +7 and +8, and equivalently the difference between +0 and +8 is more than 8 times the difference between +0 and +1. From my calculations before, it seems it's more like it takes 10 times the amount of improvement to go from 0 to +8 as it does to go from 0 to +1. To the best of my understanding, that means if you have a team at +8, xRAPM will have the sum of your team's parts as about 10 times as far above average as a +1 team.

If we want to convert from sum of xRAPM to SRS, then, we'd need to apply the inverse of that function. We'd need to recognize that when xRAPM says the Spurs are 10 times farther above average than a +1 team is, that doesn't mean they're a +10 SRS team, it means they're +8 SRS. This function scales the values to reasonable proportions.

I don't know the exact shape of this function, and it seems it would have to be empirically determined rather than being inferred from the formula, since it depends to a degree on whether a team built their leads early in the game or late, but some normalization function is necessary. Without doing it, you end up with the Warriors finishing 6th in the conference, losing their starting C who ranks as one of their best players, distributing his minutes to significantly lesser players on their bench, and, due to a few extra minutes for their starters and fewer for their reserves, more than doubling their SRS to the point that they are projected to have a higher SRS in the playoffs than every team in NBA history in any season, yet they're still projected to only rank third in their conference. GSW's starters and one-big lineups have been good enough that I could buy a +7 or +8 estimate if I knew xRAPM was high on their players, but there's no way forecasting a Bogut-less GSW at +11.9 SRS passes the laugh test, playoff rotations or not.

I used the pythagorean win% from DSmok1's posts for both teams. Then I used the log 5 method to generate an expected win%. Log 5 formula:

(home win%/Home loss%)/(home win%/home loss%+road win%/road loss%)

Gotcha. That makes sense.

colts18 wrote:[..] his xRAPM takes into account the leading by X effect.

That's exactly his point why you should account for it, too

J.E. wrote:

colts18 wrote:[..] his xRAPM takes into account the leading by X effect.

That's exactly his point why you should account for it, too

If I take that into account, wouldn't I be double counting it because you already account for it in the stat?


I'm not sure that leading by X is the right word to describe what he is trying to say. I think he is talking about diminishing returns. If you have a team of players all with a 3 RAPM, would you get a +15 per 100 team? Some say yes. Others say no. The latter gets argued by people who cite the 2011 Miami Heat as an example. The 2011 Heat acquired the #1, #2, and #9 players according to 2010 RAPM. Of course that didn't lead a dominant team that would be predicted by having 2 +8 guys.

Reran the projections using actual minutes played and got some weird results

O rating, D rating, total per 100
Spurs 6.7 9.7 16.4
Clippers 8.7 2.4 11.1
Heat 6.5 3.6 10.1
Mavs 7.8 2.2 10.0
Rockets 7.0 2.6 9.6
Grizzlies 2.0 5.8 7.7
Warriors 1.4 5.0 6.4
Pacers -1.2 7.4 6.2
Blazers 6.7 -0.9 5.8
Wizards 0.7 4.3 5.0
Nets 4.4 0.7 5.0
Thunder 0.6 3.5 4.1
Bulls -1.7 5.6 3.9
Hawks 1.3 2.2 3.5
Raptors -0.3 0.5 0.2
Bobcats -1.5 -1.0 -2.5

The Thunder's projection is way down. Maybe J.E. can explain it, but why are the numbers on the ESPN today so much different than the ones right after the regular season? For example, Dirk's RAPM improved by 0.99 points, LeBron's by 0.84 points per 100 in just 1 game. Channing Frye improved by 0.99 and he didn't even play in the playoffs. Both Andray Blatche and Udonis Haslem declined by -1.3 points in just 1 game.

Indiana might want to think about not giving Turner any minutes at all.

They were 42-13 before he got there, then went 14-13

In the first 4 playoff games Indiana outscored the Hawks by 7.7 per 48 when he was sitting, but got outscored by 18 per 48 when he played (60 minutes total). That was before he posted a -14 in 4(!) minutes in game 5

Isn't this a little too obvious? What am I missing?

That trade has to be one of the most detrimental in history

I am not sure how much of Turner's on/off numbers are because of him and how much is because of the lineups and the other players in them. In regular season his on/off for team defensive results was terrible in Indiana but they were positive in Philly by about as much last season and have see-sawed year to year throughout his career.

How much is random outcomes and how much do the lineups really matter? Many seem to believe explicitly that lineups do not much matter or matter much. I've held out that they might more more than commonly believed but can't say that I have enough basis to prove it due to mostly low to very low sample sizes. The starting lineup was the most used lineup in league at about 20 minutes per game used. It was +9 in regular season but -2 in playoffs. That is not on Turner.

By RPM Turner is a -2 player per 100 possessions. Bad but that is not bad enough in appearance to me to explain the really terrible raw +/- lineup results for the regular season or playoffs. But it is the best estimate by this method. A lot of these bad results are being assigned to others, for at least part of their overall rating. There must be better producing lineups that don't involve Turner. Involving better players but also perhaps just being better lineups. 5 of Indy's most used lineups in the playoffs are strongly positive (including one with Turner). Unfortunately only one was used even 30 minutes in the regular season and it was strongly negative then. Not an endorsement of using past lineup results as a guide... when the testing was so small. An argument, I think, for more disciplined regular season lineup testing. Only 1 lineup got tested 200+ minutes. It should be at least 6.

Granger is and was the better player by RPM and RAPM. Indiana is just one of many teams that do not appear to use any form of APM much, if any. I am not sure if there are better results / fewer surprises from using APM over pure boxscore metrics. It would be interesting to see a study of the principal players in the last 50-200 trades that checked which metric type was the better guide for post trade performance and which teams tended to do better and worse on post trade performance of their received player and for the trade comparisons.


How good is regular season lineup performance at predicting playoff performance?

Most used playoff lineup is the ATL starting unit and it would good in both regular season and playoffs, slightly better in playoffs. OKC starters are second most used. Good in regular season but a big falloff in playoffs, for 3rd of last 4 times. Portland 3rd, very good in regular season and nearly doubly so in playoffs. LAC with Redick very strong in both periods. Indy the falloff already noted. For Chicago,Toronto and Washington, a horrendous falloff in playoffs. Dallas a big improvement. Memphis not as good but still mildly positive. For teams with a big falloff or improvement it might help to look at regular season performance against that specific team or that type of team. Of these 10, only 3 went about as expected. But it maybe too early to do this kind of analysis.

Spurs starting unit is at -10 per 100 possessions but only used about 10 minutes per game (thankfully?) It was dynamite in regular season. Basketballvalue's site would have allowed me to see how this lineup did against Dallas but it can still be done with a little manual effort. The Spurs starting lineup was good against Dallas in regular season but only got 19 minutes of testing.

Miami's most used lineup plays about as much and are twice as bad so far. Same as regular season. But use that info or not? It was only test 67 minutes. Just two lineups tested over 200 minutes. But conventional wisdom is that the coach really knows his team and how they perform. I guess he is not worried by this most used lineup failing in regular season and playoffs. They know better than the numbers. Probably true in a grand perspective but not on this specific case so far.

The team got lousy after that trade. Is Hibbert lousy because of it? Did Paul George lose his touch because of it?
How does the team just since the trade do with and without him?

Using nbawowy.com and 82games, I see the following:

Pacers as whole (with and without Turner on the court) since trade til right now 1582 minutes, -97, -2.9 pts per 100 possessions

Pacers with Turner since trade til right now 627 minutes, -104, -8.0 pts per 100 possessions

Pacers since trade til right now but without Turner on court 955 minutes, +7, -+0.4 pts per 100 possessions

Pacers without Turner since the trade are just barely positive. So their decline from being very good is not totally due to him but his on court time is basically the entire negative performance time.

Pacers with Turner more than twice as bad in just the playoffs compared to regular season.

Pacers before the trade 2374 minutes, +458, +9.2 per 100 possessions.

If Pacers would have stayed the same as before the trade in just their non-Turner minutes after the trade they would have done almost 4 pts better per game than they did.

In the time turner was on the court, those lineups cost the team about 3 pts. So the non-Turner time drop-off was a little bigger but they were both important.

One could argue that the trade had an effect on the team for both the Turner on and off times. (One could say in general that players could have impact when they are off as well as on, when one considers their locker room / practice impact.)

My amigo Seth, who also claimed widely that the Turner trade was awful the very second it was made, put out this cool article the other day with another interesting theory re IND -> http://www.washingtonpost.com/news/fanc ... bowed-him/

Congratulations to Frank Vogel for not giving Evan Turner any minutes in game 6 and winning the game on the road, although I think Mahinmi playing more minutes also helped. In Turner's 'defense' the Pacers, again, looked generally awful without Paul George on the floor (who Turner is backing up), being +19 with George, -12 without. George had a +50(!) NET rating in the first 5 games. Sample size for his 'OFF' isn't huge (38 minutes) but that number is insane nonetheless

In another How To Shoot Yourself In The Foot - Edition, why is Scott Brooks giving that many minutes to Caron Butler? Granted, the case is not as clear cut as it is with Turner, but it's still confusing to me why he's playing that much. Butler is just an awful defender. I can see giving him minutes when Prince is on the floor, who does nothing on offense. But, to me, it doesn't make sense to leave him out there when Mike Miller comes into the game. The Thunder allowed the most opponent points (by far, p.p.) in the regular season when he was playing, and it's the same so far in the playoffs. Is Sefolosha injured?

That is surprising, given that Turner has his team's best postseason TS% (.623) AND their highest Ast% (28.2)
http://www.basketball-reference.com/teams/IND/2014.html
How many other players have those qualifications?

His WS/48 (.095) and PER (14.3) both are 6th best for the Pacers.

Mike G wrote:That is surprising, given that Turner has his team's best postseason TS% (.623) AND their highest Ast% (28.2)
http://www.basketball-reference.com/teams/IND/2014.html
How many other players have those qualifications?

His WS/48 (.095) and PER (14.3) both are 6th best for the Pacers.

He's also got a 22% TOV% though and his defense is quite awful. From watching the games it also feels like he's making it harder for others to hit shots because he takes a lot of time to get open (when they run plays) or just dribbles around aimlessly, burning time off the shot clock

Any insight as to why Luis Scola also seems to have been dropped from the rotation?
Chris Copeland is thrust into the mix, and he seems to be truly without merit.

Here is the updated ratings for round 2.

Power ranking of playoff teams using playoff minutes distribution (Off, Def, Ovr per 100 possessions):
spurs 6.5 8.8 15.3
clippers 10.0 2.3 12.2
warriors 4.9 6.6 11.5
rockets 6.6 4.6 11.2
thunder 6.2 3.9 10.0
heat 6.1 3.7 9.7
grizzlies 2.6 5.3 8.0
mavs 6.1 1.3 7.4
Pacers -0.4 6.9 6.5
blazers 6.8 -0.6 6.2
bulls -1.3 6.4 5.1
raptors 3.7 0.9 4.6
nets 3.7 0.4 4.1
wizards 0.9 3.0 3.9
hawks 1.5 1.0 2.5
bobcats -2.2 -0.6 -2.9

Win% for each team on a neutral court
Spurs 0.775
Blazers

Thunder 0.438
Clippers

Pacers 0.590
Wizards

Heat 0.676
Nets

win% for each team for games at home:
Spurs 0.833
Blazers 0.331

Thunder 0.553
Clippers 0.664

Pacers 0.691
Wizards 0.532

Heat 0.758
Nets 0.441


This is the series win% based on 10,000 simulations:
Spurs 94.7
Blazers

Thunder 41.8
Clippers

Pacers 71.2
Wizards

Heat 84.6
Nets

Compare that to the odds from this site with gambling odds http://www.sportsgrid.com/nba/nba-playo ... und-upset/
Blazers: 23%
Spurs: 77%

Clippers: 33.5%
Thunder: 66.5%

Wizards: 36.5%
Pacers: 63.5%

Nets: 19%
Heat: 81%

The biggest difference is the Clippers-Thunder series. Vegas has the Thunder as 2-1 favorites but J.E.'s stat has them as an underdog. I think its like 50-50. That should be an interesting series to watch. Vegas has the Spurs at 77% while J.E. has them at 95%. In that case I'm more inclined to think its closer to the vegas number.

Please refrain from saying things like "J.E. has team X at Y%". Obviously it's not just my numbers that go into your calculations. Further, people in this thread have tried to tell you that you need work in other adjustments ('being up X') but you haven't done so