MOST ACCURATE METHOD OF CALCULATING BASKETBALL WAR ?

[BasketDork]

I've seen the old "Basketball on Paper WAR" methodology on the old bbref.com blog, I've downloaded hoopsnerd.com's (Statman's) WAR, I've read VORP*2.7 gives you a pretty good result, I know there's WARP, (Player Win % - Replacement Level (generally around .415)*(MP/48), and I've seen a couple other concepts as well. What's the most accurate method of determining replacement level? I know the NBA is less position-orientated than ever, but is an individual position replacement level more accurate, due to the position scarcity that varies year-to-year? What's the most accurate way to calculate Basketball WAR ? And I know the replacement player concept is null in college, but is there a method for calculating NCAA WAR?

[Dr Positivity]

A way to evaluate replacement level...

Average team payroll in the NBA is around 71 million and average wins is 41. So dividing 41 by 71 averages out to about 1.7 million per win. We know that league MLE was set at 5.15 to represent average salary. Dividing 5.15 by 1.7 million per win gives an estimate of 3 Ws per average player. This helps us because we can divide 5.15 million for average salary by 0.508 million for minimum salary for a number of 10.1x. In other words, divide 3 Ws for an average player by 10 and it leads to 0.3 Ws for a replacement level player.

A way to double check this - 41 Ws divided by our estimate of 3 Ws per average player = 13.67 players per team included in that average. Using the same 13.67 number multiplied by replacement level of 0.3 Ws, would lead to 4.1 Ws, in other words 13.67 replacement level players is a 4 W caliber team, which checks out as a logical floor for the worst possible team you can make while still using NBA players

[ampersand5]

in regards to replacement level discussion, please see here:
http://www.apbr.org/metrics/viewtopic.php?f=2&t=8635
http://tangotiger.com/index.php/site/co ... nba-player

[Mike G]

You can define "replacement player" as one who could be had on short notice, which is to say an available player not currently on the roster.

Or you can define it as the average player whose minutes replace those of a player who is out with injury or whatever.

If Steph Curry goes out, his minutes are largely taken by players notably better than an available/unemployed guy. Players like Iguodala, Barbosa, and Livingston would take the bulk of his minutes, and in fact even better players would likely see more time. A relative few minutes would go to "available" types (Holiday, Rush).

On a team with less depth, there may not be such luxury, and indeed they may have to bring in some guys 'off the street'. But around the league, we might assume the avg player who actually replaces an NBA-avg player is considerably better than that.

... 13.67 replacement level players is a 4 W caliber team, which checks out as a logical floor for the worst possible team you can make while still using NBA players

Sounds about right. Even a team that can be expected to win zero games, in real life is more likely to win a couple. There's a probability distribution, and no team wins a negative # of games.

[BasketDork]

So unlike in baseball (where most of my advanced statistics knowledge lies), there isn't as much of a need, nor a sensibility, to have WAR and WAA, because essentially, replacement level and average are about the same? Baseball's WAR 'replacement player' is defined as a minor leaguer or what's on the scrap heap, readily available to be purchased, and not what's available off of a team's bench, as they aren't readily available, but signed for their considerable value in a particular role (middle relief, closer, 4th outfielder) and they're considered far more valuable than any replacement player, both salary and productionwise. Because of the multitude of minor leaguers, and the multi-leveled structure of the minor league system, there's a huge gap. Could one infer that college and D-League could be considered "replacement level", and free agents and bench player production be considered "average" ?

[DSMok1]

I don't think the need is any different than in other sports, but defining replacement level is a little harder due to the more continuous nature of playing time.

Please read the threads linked above, they're great resources. The end result was that using two different methodologies, players that weren't getting regular playing time at the beginning of the season and looking at veteran minimum salary free agents, gave convergent results, yielding a replacement level for BPM of -2.

[Mike G]

... replacement level and average are about the same?
... D-League could be considered "replacement level", and free agents and bench player production be considered "average" ?

Only 105 players -- 3.5 per team -- played 1500 minutes this season at above-avg (zero) BPM. About 275 played >300 min. and were above the -2.0 BPM cutoff for 'replacement level' -- some 9 per team.

While that does seem to leave over 1/3 of the league at sub-R level, and maybe 1/3 above avg: those >0 guys tend to play more minutes. So the average NBA player-minute is a higher level than the median player level.