Where SSPS Goes Wrong with AVG

I did this the other day, but the changeover to the new software slowed me down. I’ve gathered a list of hitters along with their 2004 SSPS predicted and actual averages. I present the top 25 overachievers (actual average exceeding the predicted average) and top-25 underachievers (predicted average exceeds actual average). I was trying to find some underlying cause of the “Ichiro problem” so I list some of the factors, along with their Z-scores, that I think may be associated with the mis-predictions. The factors are the AVG-to-OBP ratio and speed (I use the average of 5 speed scores presented here). I also wanted to see how the 2005 prediction changed in response to the 2004 performance, which I call “bounce back.” So what are the trends? Let’s look at the overachievers first.

Top-25 Overachievers

Rank Name Residual 2004 AVG 2004 Pred 2005 Pred Bounce Back AVG/OBP Z-Score of AVG/OBP Speed Score Z-Score of Speed Score
1 Ichiro Suzuki 0.0866 0.372 0.286 0.292 0.007 0.898 1.715 6.130 1.121
2 Adrian Beltre 0.0671 0.334 0.267 0.301 0.033 0.862 1.112 3.331 -0.575
3 Melvin Mora 0.0635 0.340 0.277 0.282 0.005 0.811 0.273 4.294 0.008
4 Ivan Rodriguez 0.0614 0.334 0.273 0.275 0.002 0.873 1.298 3.455 -0.500
5 Carlos Guillen 0.0425 0.318 0.275 0.280 0.005 0.838 0.723 6.770 1.509
6 Sean Casey 0.0407 0.324 0.283 0.301 0.018 0.851 0.933 3.724 -0.337
7 Mark Loretta 0.0391 0.335 0.296 0.310 0.013 0.858 1.044 4.342 0.038
8 Carlos Lee 0.0373 0.305 0.267 0.272 0.005 0.832 0.609 4.397 0.071
9 Javy Lopez 0.0353 0.316 0.281 0.271 -0.009 0.854 0.991 2.566 -1.039
10 Jack Wilson 0.0331 0.308 0.275 0.287 0.012 0.919 2.075 5.903 0.984
11 Juan Pierre 0.0319 0.326 0.294 0.301 0.007 0.872 1.283 7.382 1.880
12 Barry Bonds 0.0317 0.362 0.330 0.356 0.026 0.594 -3.351 4.797 0.314
13 Vladimir Guerrero 0.0307 0.337 0.306 0.301 -0.005 0.860 1.091 4.761 0.291
14 Jason Kendall 0.0290 0.319 0.290 0.294 0.005 0.799 0.075 4.053 -0.138
15 Lance Berkman 0.0290 0.316 0.287 0.302 0.015 0.703 -1.534 4.117 -0.099
16 Mike Young 0.0270 0.313 0.286 0.293 0.007 0.886 1.523 6.351 1.255
17 Scott Rolen 0.0265 0.314 0.288 0.300 0.013 0.768 -0.448 4.951 0.407
18 Miguel Tejada 0.0263 0.311 0.285 0.287 0.002 0.864 1.142 3.274 -0.609
19 Craig Monroe 0.0259 0.293 0.267 0.275 0.007 0.870 1.252 4.144 -0.083
20 Erubiel Durazo 0.0238 0.321 0.297 0.296 -0.001 0.810 0.251 3.655 -0.379
21 Carl Crawford 0.0232 0.296 0.272 0.282 0.010 0.893 1.638 10.719 3.902
22 Mark Kotsay 0.0232 0.314 0.290 0.303 0.012 0.848 0.882 4.540 0.158
23 Jim Edmonds 0.0224 0.301 0.279 0.288 0.009 0.720 -1.249 5.193 0.553
24 Raul Ibanez 0.0209 0.304 0.283 0.281 -0.002 0.860 1.079 2.924 -0.822
25 Cesar Izturis 0.0199 0.288 0.268 0.283 0.015 0.874 1.316 6.998 1.647

The most obvious similarity among these players is for AVG/OBP, which is something many people have noticed. While Bonds, Berkman, and Edmonds made the list, most of the guys are high AVG/OBP hitters. Interestingly enough, when I threw this variable into some regressions it did not have any impact on the prediction. I thought speed might matter, because speedy baserunners might be able to get on base on balls in play that would lead to outs for most players, but there does not seem to be much of a trend with this. Now, let’s look at the underachievers.

Top-25 Underachievers

Rank Name Residual 2004 AVG 2004 Pred 2005 Pred Bounce Back AVG/OBP Z-Score of AVG/OBP Speed Score Z-Score of Speed Score
1 Scott Spiezio -0.0664 0.215 0.282 0.270 -0.012 0.747 -0.799 4.546 0.161
2 Jason Phillips -0.0609 0.218 0.279 0.275 -0.004 0.733 -1.027 1.848 -1.474
3 Desi Relaford -0.0589 0.221 0.280 0.275 -0.005 0.747 -0.794 3.550 -0.443
4 Chipper Jones -0.0461 0.248 0.294 0.289 -0.005 0.686 -1.823 2.486 -1.087
5 Alex Gonzalez -0.0396 0.232 0.271 0.269 -0.002 0.858 1.047 3.609 -0.406
6 Jody Gerut -0.0386 0.252 0.290 0.288 -0.002 0.753 -0.695 5.917 0.992
7 Jay Payton -0.0366 0.260 0.296 0.293 -0.003 0.797 0.028 4.098 -0.110
8 Jose Valentin -0.0333 0.216 0.249 0.234 -0.015 0.751 -0.739 5.687 0.853
9 Scott Podsednik -0.0321 0.244 0.276 0.273 -0.003 0.779 -0.272 7.987 2.247
10 Alex Cintron -0.0319 0.262 0.294 0.283 -0.011 0.871 1.272 4.562 0.171
11 Jose Cruz -0.0317 0.242 0.274 0.269 -0.005 0.728 -1.121 6.239 1.187
12 Luis Gonzalez -0.0304 0.259 0.289 0.286 -0.003 0.694 -1.681 4.673 0.238
13 Eric Chavez -0.0282 0.276 0.304 0.306 0.002 0.695 -1.668 2.063 -1.344
14 Bret Boone -0.0279 0.251 0.279 0.264 -0.015 0.792 -0.042 3.101 -0.715
15 Toby Hall -0.0257 0.255 0.281 0.277 -0.004 0.850 0.914 0.762 -2.132
16 Joe Crede -0.0253 0.239 0.264 0.259 -0.005 0.799 0.073 2.635 -0.997
17 Bernie Williams -0.0253 0.262 0.287 0.282 -0.005 0.728 -1.119 2.355 -1.167
18 Corey Koskie -0.0251 0.251 0.276 0.272 -0.004 0.734 -1.017 5.231 0.577
19 Mike Cameron -0.0249 0.231 0.256 0.258 0.002 0.725 -1.171 5.766 0.900
20 Bobby Higginson -0.0242 0.246 0.270 0.268 -0.002 0.695 -1.673 4.021 -0.157
21 Eric Hinske -0.0222 0.246 0.268 0.269 0.002 0.786 -0.142 4.216 -0.039
22 Angel Berroa -0.0221 0.262 0.284 0.276 -0.007 0.850 0.920 6.445 1.312
23 A.J. Pierzynski -0.0221 0.272 0.294 0.303 0.010 0.852 0.953 1.232 -1.847
24 Vinny Castilla -0.0219 0.271 0.293 0.301 0.008 0.817 0.364 2.207 -1.256
25 Carlos Pena -0.0218 0.241 0.263 0.263 0.000 0.713 -1.362 5.317 0.628

There are a lot of guys here with lower AVG/OBPs, but the trend is not as strong as it was for the overachievers. Speed doesn’t seem to be much of a factor here, either. The good news for the system is that in 2005 the projections seem to be moving in the opposite directions.

Overall, I’m pretty pleased with the system. Some guys are hard to predict and may have combinations of unique skills that are difficult to quantify. I’m fine dealing with that. The next step is to move on to minor league predictions. I’ll let you know when I’m done.

2 Responses “Where SSPS Goes Wrong with AVG”

  1. josh says:

    Some of these guys, I definitely would have thought would be on these lists. Also, I wonder if Depodesta is working with something similar, picking up Phillips and Valentin. If Vinny Castilla hits .301 you deserve the nobel prize in Saberocity.

  2. JC says:

    I do think that Vinny is a bit under-rated, but I think .301 is high. In SSPS Version 1, I assume that every player plays for the the same team as the previous year, so the Coors inflation is included in that. However, I would not be surprised if he his over .290. In the future, I hope to update projections for the correct team.

    The Nobel in Saberocity? I’d like to see that! :)