DFDepressed FanDepressed Fan



, all the time

Projecting With Advanced Stats (Part 2)

http://www.depressedfan.com/img/projections092210.jpg In part one of the series we took a look at Wins Produced and Win Shares, then put them through their paces a bit in a retroactive way. Now it's time to test them out in a forward-looking way. Warning, math talk ahead.

The Math
Here's the basic idea: WP and WS are both broken down into per-48-minute figures for every player, this is the derivative we're going to be using from here on out, because once you break them down to the minute (or 48), you can then apply their production to any sample size.

So what does this mean? Well, it means we can use WP P48 and WS P48 figures from the previous season to set an expected win total for a team. This is easy to do for past seasons, because we have all the data in front of us. In plain English, if every player on the Sixers '09-10 roster produced to the exact same level as he did in '08-09, we can use the actual minutes played from '09-10 to calculate the team's expected wins. Then we can look at the actual P48 numbers to see how close the prediction was.

To illustrate how we're looking at the stats, this chart displays Andre Iguodala's expected vs. actual production in '09-10, using both metrics:

For the next set of charts, I pulled together WP48 and WS48 stats for every player on the Sixers' roster in each of the past four seasons, then generated expected vs. actual WS and WP for the entire team in each season. One caveat: since there is no way to calculate a rookie's previous season P48 numbers, I simply duplicated their numbers from their rookie season when calculating the expected win totals.

A couple things to keep in mind as you digest those numbers:
  1. The Sixers have gotten a decent amount of minutes from young players over the past four seasons. Young players are typically harder to predict, but you'd hope they'd outperform their expected WP and WS numbers, meaning they're improving. Generally speaking, that hasn't been the case. Only once did the Sixers perform better than their expected WS or WP, in '07-08.
  2. The more minutes a team gets from rookies, the more stability inherent in the expected total (because we're using actual P48 numbers for rookies). Jrue, Speights, Thad and Carney all played over 1,000 minutes in their rookie seasons.
Looking back at the past four seasons of data, the expected WP and WS for the team has been decently predictive for the Sixers. The next step is to apply this math to the upcoming season, and that will be the topic of part three.
by Brian on Sep 22 2010
Tags: Advanced Stats | Basketball | Offseason | Projections | Sixers |