Tis the season of NBA Player Projections. There are a bunch of sites posting their fantasy projections, either for free or as part of a subscription. But how can you tell if a set of projections is even worth the paper its printed on? Well, let's look a past performance...namely how well did each do in projecting the actual 2005-06 stats in each major category.
The Candidates2004-05 Stats - the baseline. Any monkey could use the previous season's statistics. A good place to start.
2004-05 Post All-Star Break - a lot of people like to look at 2nd-half stats from the previous year.
NBA.com - free projections from their website
CBS Sportsline - free projections when you sign up for a team
Basketball Monster - Great website for rankings and managing your team. Projections were free last year. You need a $12 membership to get them this year.
Rotowire - pay for projections. You can buy these via the draft kit on yahoo for $15.
Wegoblogger - the projections I used for last year. You can get my current year projections here for free.
Note: if you have any other projections, let me know and I'll add them to the mix.
The projections for each category were compared to the actual 2005-06 statistics for 167 players that were common to each set of projections. 9 rookies were also included (but obviously not for the 2004-05 sets since they weren't in the NBA). For each of the 16 major categories (MIN,FGM,FGA,FG%, 3PM,3PA,3P%, FTM,FTA,FT%, REB,AST,STL,BLK,TO,PTS), the volatility of each of the projections was measured, and converted into a weighted-average standard deviation (essentially a +/- measure). Lower +/- means higher accuracy with the actual results.
Note: NBA.com and CBS did not have minute projections.
FIELD GOALS MADE
Note: NBA.com did not have FGM/FGA projections (only percentages).
FIELD GOALS ATTEMPTED
FIELD GOAL PERCENTAGE
THREE POINTERS MADE
THREE POINTERS ATTEMPTED
Note: NBA.com did not have 3PA or 3P% projections
THREE POINT PERCENTAGE
FREE THROWS MADE
Note: NBA.Com did not have FTM and FTA projections (only percentages)
FREE THROWS ATTEMPTED
FREE THROW PERCENTAGE
So what does this all mean? Interestingly, using just the average BLK's and 3PM from 2004-05 would've been your best bet. NBA.com did well in FT%, 3PM, and STL. And I was surprised to see how poorly the post all-star break projections did across the board.
If we convert each category to an index where the 2004-05 stats equals 100, we can see how each projection did in the 9 roto categories.
Using CBS Sportsline and Rotowire would've left you worse off than just using the previous year's statistics. My projections offered up a 5% improvement on average. Not trying to toot my own horn, but just showing that it's not too difficult to come up with a reasonable set of projections. Do your homework. Use mine, basketballmonster or nba.com as a starting point or just as a sanity check. If you see somebody who you think is ranked too low or too high, make sure there's some number you can point to that's causing you to think that (instead of just perception that player A is better than player B).
Perhaps even more critical is a projection's ability to identify those players whose statistics are most likely to change year over year (sleepers versus busts). This can be the deciding factor between victory and defeat. So if we look at the top 25 players that dropped or gained in each individual category, we can see how well each projection did in predicting that change. Again, 100 represents the baseline of 2004-05 (but since the players are ranked by their difference from the baseline, you'd expect most predictions to come below 100).TOP 25 DROP
|Top 25 Drop||FG%||FT%||3M||REB||AS||ST||BK||TO||PTS||AVG|
TOP 25 GAIN
|Top 25 Gain||FG%||FT%||3M||REB||AS||ST||BK||TO||PTS||AVG|
The post all-star break numbers do show up as doing well in predicting those who might take a big dip in blocks the following year (much better than anything else), so there's one area where you might use it. My projections did a good job of capturing those with large increases and did okay with the decreases (other than blocks). Rotowire is probably not useful in anything other than identifying sleepers.