APBRmetrics

I was thinking if maybe it would be useful to develop an alternative to Wins Produced using RAPM.

The way I figured it could be done is pretty easy. First, figure out what the replacement level RAPM is. Subtract from each player's RAPM to create RAPMAR.

Next, using the RAPMAR and SD of RAPM calculate the positive area under the curve and the negative area under the curve of the graph with RAPMAR as the X-axis and minutes as the Y-axis.

Add up the positive areas and negative areas for each player on the team. Calculate each player's percentage of the sum of all the team's player's positive values and negative values. Multiply times wins and losses, respectively.

What do you guys think of this idea?

Well I thought about this on my way home and realized it wouldn't work using replacement level times actual wins and losses. I guess it would be OK to just use regular RAPM and calculate win shares and loss shares and compare that to a replacement level of wins/losses given the same amount of minutes.

You might define replacement level as zero wins and all losses, against NBA teams.
And therefore, WAR are just wins expected or produced. "Losses above replacement" is then just as meaningless as it sounds.

And no need to consider any production as an alternative to walrus poop (WP).

You can calculated a Points above Replacement statistic for any "efficiency differential denominated" stat. Any of the Plus/Minus stats are denominated in efficiency differential.

The general equation is (PMStat + Replacement Player Level)*%min = Points above Replacement (per 100 possessions played).

The trick is figuring out replacement level. This workbook should help you: http://public.tableausoftware.com/views ... 1?:embed=y .

I'd set RAPM replacement level at -3.5 or so.

Once you have Pts above Replacement calculated, you can attempt to convert to wins. In general, you can't convert directly (Pt. differential does not convert to wins linearly). A quick hack is to use the "average" pt differential to wins conversion, which Dave Berri calculated for us as 3.15 (see the first table at http://wagesofwins.com/how-to-calculate-wins-produced/ ).

So, tying it all together: Dirk had an RAPM of +8.0 last year. He played 64.7% of Dallas's minutes.

So his Points above Replacement would be (8.0+3.5)*.647=7.44, and his Wins above Replacement would be 7.44*3.15=23.4.

Zounds! That formula gives Dirk 23.4 of the Mavs' 35 wins?
He has 7.6 Win Shares (b-r.com), and I give him 7.1 equivalent wins.

If you've scaled to an 82-game season, that's still 18.8 of 35, in 66 games.

Mike G wrote:Zounds! That formula gives Dirk 23.4 of the Mavs' 35 wins?
He has 7.6 Win Shares (b-r.com), and I give him 7.1 equivalent wins.

If you've scaled to an 82-game season, that's still 18.8 of 35, in 66 games.

Yes, that was an 82 game season, and yes, that's what RAPM implies.