Power of the Preseason

I have to admit, I haven't really been following the NBA preseason. I glance at the boxscores, but haven't used much, if any, of that information in my projections for the upcoming season. Maybe it's because I got caught up in the hype around James Jones last season and got burned:

 Jones,James GP Min  FGA FG% 3M FTA  FT% REB AS ST  BK PTS
 2005-06 Pre 7  31.29   12.43  0.460  3.14  1.57  0.909   4.00  1.86  0.71  0.71  16.00 
 2005-06 Reg 75   23.56  7.75  0.418  1.47  1.61   0.851  3.37  0.76  0.51  0.65   9.32 

Call it the James Jones corollary. We all have fantasy basketball baggage. It's like the girl who dumped you in 7th grade and broke your heart -- you're scarred for life. Even if that player's name hits the top of your draft queue, you can't help but think of how your fantasy hopes went down the drain because of them three years ago. Corey Maggette, Stromile Swift, Steve Francis. You get the picture. That's how I've been treating the preseason stats this year.

But maybe I'm missing the boat. Maybe there's a sleeper out there posting good preseason numbers and I'm not paying attention. The problem is, how do you know? Baron Davis looks pretty darn good this year, to the point where he's looking like he might get back to top 20 status. Heck, he's shooting 91% from the free-throw line! Then I checked out and it shows that Baron was the best preseason player last year too, shooting 85% from the stripe in 5 games. He ended up shooting 67.5% for the season and ranking about 40th overall.

So how do you know what preseason stats to rely on, and what to dismiss? I don't know. Gut check, I guess. But can we do an analysis to determine if preseason stats provide information in general? If you've learned anything from reading this blog, than you already know the answer to that question. Have spreadsheet, will travel.

With all of the hoopla about the new ball, one thing I realized is that the aggregate preseason data can be a decent predictor of what's going to happen globally in the regular season. You know that shooting is going to be down and turnovers are going to be up in the preseason, but can year over year changes in the preseason data tell us what's going to happen during the regular season? I think so. Basketballmonster has full preseason stats for 2003-04 and 2005-06. Here is how they compare to the regular season on a per 48 minutes basis:

 MIN48.0048.00 48.0048.00--
 FGM6.587.04 6.737.127.1%5.8%
 FGA15.2416.01 15.2315.705.0%3.1%
 3PM0.841.06 1.001.1425.9%13.9%
 3PA2.503.03 2.893.1821.2%10.0%
 FTM4.083.68 4.433.90-9.7%-11.9%
 FTA5.594.90 6.065.23-12.3%-13.7%
 STL1.851.58 1.781.42-14.3%-19.8%
 BLK0.981.02 0.940.934.6%-0.7%
 TO3.562.84 3.482.74-20.1%-21.3%
 PTS18.0718.82 18.8919.284.2%2.1%
 2PM5.745.99 5.735.994.3%4.4%
 2PA12.7412.98 12.3512.521.9%1.4%
 FG%0.4320.440 0.4420.4541.9%2.7%
 FT%0.7290.752 0.7310.7453.0%2.0%
 3P%0.3350.348 0.3460.3583.9%3.5%
 2P%0.4500.461 0.4640.4782.4%2.9%

You can see the change from preseason to regular season is very consistent for many categories. You can also see clearly how shooting improves and turnovers decrease (by 20%!) as the real season starts as expected.

So what does that mean for 2006-07. From the initial preseason data, it looks like turnovers could be an issue with the new ball, as many players have predicted. Any downturn in shooting and scoring will likely be more than offset by an increase in pace as more teams aim to be more like the Suns. Get ready for some high-scoring affairs.

I used the average regular season to preseason change from the two years above to come up with predicted numbers for the upcoming season (06PRED). Because we don't yet have a full preseason of data, I normalized the numbers to bring them closer to the 2005-06 regular season data (06ADJ). Here's what it shows:

 MIN48.0048.00 48.0048.0048.00-
 FGM6.737.07 7.547.337.122.9%
 FGA15.2315.70 16.3416.0215.702.0%
 3PA2.893.16 3.653.413.187.5%
 FTA6.066.50 5.665.445.234.1%
 REB8.088.24 8.398.278.141.5%
 AST4.024.18 4.404.254.103.6%
 STL1.781.76 1.461.441.421.3%
 BLK0.940.90 0.920.930.93-0.8%
 TO3.483.93 3.112.932.746.9%
 PTS18.8919.97 20.6019.9419.283.4%
 2PA12.3512.54 12.7512.6412.520.9%
 FG%0.4420.450 0.4600.4570.4540.7%
 FT%0.7310.737 0.7550.7500.7450.7%
 3P%0.3460.333 0.3450.3520.358-1.8%
 2P%0.4640.479 0.4920.4850.4781.5%

Turnovers, scoring and three-point attempts are up significantly. If you noticed anything from the playoffs last year, it was the inordinate amount of free throws (led by Dwyane Wade). That trend seems to continue, as officials are calling the game more closely (not shown above, but fouls/48 are up 5.5% this preseason over last).

That's all fine and good, but how does that information help my fantasy team? To be honest, I'm not sure. You could try to make global adjustments to account for these changes in each category, but that might not had much affect on your overall player rankings. It's worth testing out, which is what I plan to do. More on that later...let's see what we can do on individual player projections.

Getting back to Baron (sounds like an name of the Fray's next album). Do his preseason numbers provide any information for what's going to happen in the upcoming season? Hard to say. But on average, it does look like the preseason numbers do have some value. Let's recall my recent comparison of 2005-06 stat projections. For comparison purposes, I used the 2004-05 numbers as the baseline prediction and created an index where 100 = simply using 2004-05 numbers to predict actual 2005-06 results. Below 100 meant you did better than the baseline. Surprisingly, a couple sites did worse on average than this most basic bogey.

What if we created another prediction that was based on the 2004-05 numbers PLUS the next season's preseason data? Let's call it 2004+PRE. First, I adjusted the raw preseason data because we know that shooting's going to go up, turnover's down, etc. Second, I looked at per 48 minute production, so the minutes were kept constant to the previous season. And third, I only weighted the preseason numbers by the number of games played. The 2004-05 data was given a weight of 82 games across the board. So if a guy played in 8 preseason games, his new prediction would be 82/90ths his 2004-05 production and 8/90ths his preseason production. Here are the results.

_CAT__200405_ _2004+PRE__Monster__WEGO_
Min100.0100.0 91.6 90.9
FGM100.0 98.4 93.2 93.1
FGA100.0 99.0 93.8 93.2
FG%100.0 98.7 97.0 92.9
3M100.0 96.6 94.2 87.5
3A100.0 97.7 98.2 91.5
3P%100.0100.2 103.585.8
FTM100.0 97.8104.5 97.7
FTA100.0 97.4103.0 93.7
FT%100.0 98.4 96.8 95.3
REB100.0 99.2 99.1100.7
AS100.0 97.8 92.4 84.4
ST100.0 99.1 100.0 97.5
BK100.0 97.6108.0111.0
TO100.0 98.4107.0 90.6
PTS100.0 98.7 93.7 93.5
TOTAL100.098.498.5 93.7

I showed Basketball Monster's and my predictions from last year for context. Surprisingly to me, the preseason numbers to hold some informational value (and a 1% difference can be the thin line between winning and losing your fantasy league). Of course, they are hindered by the fact that it keeps the prior year minutes constant. The minutes is where BBM and I really shine, since it's easy to embed changes in roles/teams in the minute projections. So that leads me to my next question. What if we added the preseason data on top of my original projections...would they provide even greater accuracy? Let's create a 'WEGO+PRE' using the same methodology:

_CAT__200405_ _2004+PRE__Monster__WEGO_ _WEGO+PRE_
Min100.0100.0 91.6 90.9 90.9
FGM100.0 98.4 93.2 93.1 91.4
FGA100.0 99.0 93.8 93.2 92.9
FG%100.0 98.7 97.0 92.9 92.2
3M100.0 96.6 94.2 87.5 86.1
3A100.0 97.7 98.2 91.5 90.9
3P%100.0100.2 103.585.8 87.0
FTM100.0 97.8104.5 97.7 96.1
FTA100.0 97.4103.0 93.7 91.8
FT%100.0 98.4 96.8 95.3 94.0
REB100.0 99.2 99.1100.7 99.9
AS100.0 97.8 92.4 84.4 83.4
ST100.0 99.1 100.0 97.5 96.9
BK100.0 97.6108.0111.0 109.6
TO100.0 98.4107.0 90.6 88.2
PTS100.0 98.7 93.7 93.5 92.0
TOTAL100.098.498.5 93.792.7

Not's shaves another percentage point off. Blocks are still horrible, but this is something that might've been saved with a global adjustment. I'll do that eventually, but right now I need to go back and adjust my player projections before my live drafts tonight! But for testing purposes, I'll create three sets of projections:

1. Baseline Projections as Oct 24, 2006. Same ones that I posted a few days ago. No preseason adjustments.
2. Projections w/ player preseason adjustments.
3. Projections w/ player preseason adjustments and global adjustments.

At the end of the year, we'll see which one holds up the best. And that will help me determine how to use the preseason numbers in the future. It just might help you win your fantasy league. Stay tuned.


  1. Anonymous said...:

    Interesting analysis. Got those second 2 projections done yet?

Wegoblogger #31 © 2011 | Designed by Bingo Cash, in collaboration with Modern Warfare 3, VPS Hosting and Compare Web Hosting