31 July 2014

Are Traded Prospects Worth Less?

Every year at the trade deadline we see top prospects traded for top veteran players. So far this year there has been one blockbuster deal made when the Cubs traded Shark and Hammel for Addison Russell, Billy McKinney and Dan Straily and it is possible that there will be more. In many of these blockbuster deals, people think that the team that traded prospects usually gave up too much to get their veteran target.

The study that I wrote discussing historical prospect value also indicates that teams are trading too much prospect value in order to get veteran targets IF projected wins have the same value at the beginning of the year as they do at the end of the year. However, it is questionable whether the above condition is true. Teams have a better understanding of their position after 100 games have been played then they do at the beginning of the year. The Red Sox and Rangers probably thought that they would be competitive before the season started but have learned since that they will not. Therefore if teams would pay $7 million per a win in free agency it makes sense that they would pay considerably more than $7 million at the trade deadline. In addition, teams that trade for veteran players usually believe that said veteran will help them in the playoffs. A win in the playoffs should be considered more valuable than a win in the regular season. If one uses the typical rule of thumb that a win in the regular season is worth $7 million in value then this will result in one undervaluing veteran players. The value of a win at the trade deadline for a team in the playoff hunt is considerably higher than the value of a win in free agency.

It seems that veterans are consistently underrated based on the reasoning above. But another relevant question is whether the traded top prospects are being overrated. All top prospects that have been traded have something in common --- their team was willing to trade them. This could indicate that they believe their prospect is overrated and therefore less valuable then an equivalently ranked prospect. What if teams that trade their top prospects do so because they believe that they simply aren’t as good as the average equivalently ranked prospect? What if traded prospects aren’t as valuable as not traded prospects?

In order to test this theory I looked at all prospects ranked in the top 100 by Baseball America from 1990-2010 and grouped them by year ranked, actual ranking, whether he is a position or pitching prospect and whether the prospect was traded.

I used two different sets of year rankings. The first one was 1990-1993, 1994-1997, 1998-2002, 2003-2006 and 2007-2010. The second one was 1990-1997, 1998-2006 and 2007-2010. The value of pitching prospects has changed over time and twenty one years is a pretty long period of time. I use different year groupings to insure that whatever results that I find remain consistent over the time period of the sample. It is important to realize that data from 2007 to 2010 is incomplete. Many players ranked from 2007-2010 are still under team control and therefore their value will change.

For each of these year ranking groups, I further split prospects into groups based on whether they were a position or pitching prospect and how high they were ranked. Position prospects are typically more valuable than equivalently ranked pitching prospects and a top 10 prospect is usually better than a top 80 prospect. I split prospects into groups based on whether they were position or pitching prospects and whether they were ranked 1-10, 11-25, 26-50 and 51-100 for a total of eight groups.

This study has a small sample size of traded prospects and therefore I only used these groupings to determine expected value. While the sample size is large enough to be meaningful for a four year period it isn’t large enough to tell me very much about whether traded pitching prospects ranked 11-25 from 2003-2006 were better or worse than average pitching prospects ranked 11-25 from 2003 to 2006. This is because there were so few traded pitching prospects ranked 11-25 from 2003-2006.

I also grouped prospects based on whether they were traded while still prospects/rookies. This is more difficult than it sounds. Suppose a player was ranked in 2004 was traded in 2008 and only subsequently made it to the majors. One could claim that he should be considered a traded prospect but this would defeat the purpose of the study. I only considered top 100 prospects traded if they were traded the year before they were ranked, the year they were ranked or the year after they were ranked.

Using salary information and Baseball Reference it is possible to determine how long a prospect was under team control. If a prospect signed an extension then I determined how long he would have been under team control if he didn’t sign an extension. For each position prospect I multiplied the WAR he was worth in each season he was under team control by the amount a free agent win was worth and subtracted by his salary. For pitching prospects I used the same methodology with the one exception being that I used RA9_WAR instead of FIP_WAR. I believe that RA9_WAR does a better job explaining what did happen even if FIP_WAR does a better job projecting the future. Then I added the results for each year for a prospect to determine the players’ total value over the years that he was under team control.

Next I determined the expected mean value for each grouping defined above. A top 10 position prospect is more valuable than a 51-100 pitching prospect and therefore the expected value for a prospect should account for that. If a prospect was ranked in multiple years then I determined his mean expected value based on his position, rankings and years ranked over the period that he was ranked. After the averages were determined for each grouping, each prospect only counts once in the rankings. Otherwise a traded prospect ranked four times has an overly inflated value compared to a prospect ranked once.

Now that I’ve explained the methodology it is possible to look at the results. This first table shows the actual value, expected value and the difference for all traded prospects in the time frames defined.



Year
Mean RA9
Mean Expected Value
Difference
1990-1993
0.589
1.709
-1.12
1994-1997
1.515
3.947
-2.432
1998-2002
3.012
4.418
-1.406
2003-2006
5.165
5.448
-0.283
2007-2010
3.128
3.733
-0.605




1990-1997
0.899
2.79
-1.891
1998-2006
3.551
4.86
-1.309
2007-2010
3.128
3.733
-0.605


These results indicate that traded prospects are worth less than the average prospect but that the difference has decreased over time. Teams were really bad at determining the difference between 1990 and 2002 but from 2003 and 2010 they’ve seemingly improved. Teams know their prospects better than other teams but the difference has seemingly narrowed.

The dataset is large enough to see how traded prospects do based on year and whether they are a position or pitching prospect. Here’s the table.



Year
P_Type
Mean RA9
Mean Expected Value
Difference
1990-1993
Position
1.085
2.446
-1.361
1994-1997
Position
2.025
4.877
-2.852
1998-2002
Position
3.951
5.004
-1.053
2003-2006
Position
7.846
6.683
1.164
2007-2010
Position
3.841
4.561
-0.720





1990-1993
Pitcher
0.008
0.845
-0.837
1994-1997
Pitcher
1.039
3.080
-2.040
1998-2002
Pitcher
1.810
3.668
-1.857
2003-2006
Pitcher
2.762
4.340
-1.578
2007-2010
Pitcher
2.415
2.852
-0.437





1990-1997
Position
1.275
3.562
-2.287
1998-2006
Position
4.972
5.626
-0.654
2007-2010
Position
3.841
4.561
-0.720





1990-1997
Pitcher
0.502
1.972
-1.470
1998-2006
Pitcher
1.950
3.997
-2.047
2007-2010
Pitcher
2.415
2.852
-0.437


This shows that traded pitching prospects were significantly worse than average pitching prospects through the whole sample although teams seem to be doing better over a period of time. Traded position prospects from 2003 to 2006 were considerably better than average although traded position prospects from 1998 to 2002 and 2007 to 2010 are not. This may be nothing more than random chance over a small sample.

It is possible to tell whether the differences in these results are statistically significant. This dataset is non-normal and expected traded prospect value is dependent on actual traded prospect value. Therefore, I’m using the Wilcoxon Signed Rank Sum Test and the Sign Test to determine statistical significance.

This table shows the significance for all years without regard to position.




Year
Wilcoxon Signed Rank Sum Test Statistic
Wilcoxon Signed Rank Sum Test p Value
Sign Test Stat
Sign Test P Value
1990-1993
-535
0.0001
-19.5
<.0001
1994-1997
-495.5
<.0001
-16
<.0001
1998-2002
-539.5
0.0024
-15.5
0.0004
2003-2006
-152
0.2056
-7.5
0.0581
2007-2010
-176.5
0.1298
-10
0.0091





1990-1997
-1593
<.0001
-31.5
<.0001
1998-2006
-1234.5
0.0006
-23.5
<.0001
2007-2010
-176.5
0.1298
-10
0.0091


The data show that the difference between the actual value of traded prospects and the expected value of traded prospects is statistically significant from 1990-2002. From 2003-2006 the differences aren’t statistically significant although they are reasonably high. When looking at the data in eight year increments the difference between actual and expected value is significant from 1990-2006 and very well may end up being the case for 2007-2010.

This table shows the significance for all years when taking position into account.



Year
P_Type
Wilcoxon Signed Rank Sum Test Statistic
Wilcoxon Signed Rank Sum Test p Value
Sign Test Stat
Sign Test P Value
1990-1993
Position
-155.5
0.0059
-9
0.0029
1994-1997
Position
-112
0.0081
-8
0.0037
1998-2002
Position
-114.5
0.1398
-4.5
0.211
2003-2006
Position
19.5
0.6298
0
1
2007-2010
Position
-42
0.3221
-4.5
0.1221






1990-1993
Pitcher
-135.5
0.0019
-10.5
0.0001
1994-1997
Pitcher
-145.5
0.0015
-8
0.0052
1998-2002
Pitcher
-171
0.0006
-11
0.0001
2003-2006
Pitcher
-120.5
0.0067
-7.5
0.0081
2007-2010
Pitcher
-52
0.2177
-5.5
0.0522






1990-1997
Position
-422.5
<.0001
-15.5
<.0001
1998-2006
Position
-175.5
0.2213
-5
0.2529
2007-2010
Position
-42
0.3221
-4.5
0.1221






1990-1997
Pitcher
-373.5
0.0001
-16
<.0001
1998-2006
Pitcher
-499
<.0001
-18.5
<.0001
2007-2010
Pitcher
-52
0.2177
-5.5
0.0522


The actual value of pitching prospects is statistically significantly different from the expected value from 1990-2006. It is likely that this will also be the case from 2007-2010. However, the statistical significance is slowly decreasing. It seems possible that teams are getting better at determining which pitching prospects are most likely to succeed and therefore the best targets.

The actual value of position prospects is statistically significantly different from the expected value from 1990-1997 but not from 1998-2006. While the 2007-2010 data suggests that the results from 2003-2006 is a fluke it also suggests that people are beginning to learn how to successfully value other teams’ offensive prospects.

Teams are learning how to successfully value other teams’ prospects but they’re doing a much better job of valuing position prospects than pitching prospects. If the Orioles are willing to trade a top pitching prospect like Gausman, Bundy or Harvey then it probably means that they don’t think he’ll be successful in the majors.

No comments: