As the draft looms tomorrow night, NFL teams are (hopefully) finalizing their draft boards and trying to make plans for every conceivable contingency. Whether or not they want to trade, and in which direction, is surely one of those considerations. In the past two years, we have seen two monumental trades: Atlanta’s trade for Julio Jones, and Washington’s trade for (presumably) Robert Griffin III. There are sure to be more throughout the draft as teams try and select the players they covet most. One question looms large in this picture: do teams that trade up actually benefit from the moves they make, or do they pay too high a price for the right to choose? We’ll explore the answer to this question here.
Using my previous work on the NFL draft, we can find the difference between the Career Approximate Value (CAV) of the players selected by a team that traded up to select them and the expected Career Approximate Value (eCAV) of the picks that teams gave up in exchange to select them. By comparing this net CAV of players involved in trades to the net CAV of all other players selected in the draft, we can see if players that were drafted as a result of a trade up outperformed their price relative to players who were selected normally.
Put into algebra, our equations follow. For players that were traded up for, our equation for net CAV is
Net CAV = CAVx – eCAVa – eCAVb – eCAVc …
where CAVx is the CAV of the player traded for, and eCAVa, eCAVb etc. are the expected CAV’s of the picks that were traded away. For all other players the equation is simply
Net CAV = CAVx – eCAVa
because teams only used one draft pick to select that player. This post does not include trades that involved active players or future draft picks, which are more complicated than we want to look at here. They would be valuable studies for another time.
Using data from the 1990-1999 drafts, I observed 98 eligible trades and 2584 other selections. The results are outlined in the table below:
As you can see, both the mean and median net CAV are lower for players selected as the result of a trade up, while the percentage of players selected is almost exactly the same. Using a difference of means T test, the difference between these two means is statistically significant at the 95% confidence level (T = 2.31). The 95% confidence interval for the mean net CAV of players not involved in trades was (-0.24, 1.25) while the 95% confidence interval for players involved in a trade was (-12.40, -0.53), which is more clearly seen in the graph below.
The 95% confidence interval for the difference between these means is (0.99, 12.95), meaning that players that are traded up for tend to underperform their price by 0.99 to 12.95 CAV more than players who were not traded up for.
These results suggest that teams tend to overpay for earlier draft picks – or at least they did during the 1990s. These results corroborate those found by Cade Massey and Richard Thaler in their paper on the draft. Teams that can resist the urge to trade up would likely gain a competitive advantage over teams that overpay for the right to choose, and would benefit even more by taking advantage of their competitor’s overconfidence and trading down instead. Will teams actually do so? We’ll see this weekend.
Interesting that many of the so called experts recommended trading up for RG3. The Redskins may live to regret that deal!
Your CAV model is interesting but I think overvalues lower round players hence always making it look like a better deal to trade down then up. For example, if I understand it right, trading the # 1 pick for four or five second/third rounders would be a good deal in this model. I think in real life though, that would be an insane trade. A single high CAV player is almost certainly worth multiple low CAV players regardless of whether the net CAV of the multiples players is higher.
Really enjoyed the analysis.
I agree Normalman, I think you need to consider the net replacement value of the draft pick you select. How much value are you gaining from the player you select vs. the player he is replacing on your football team. We discussed something similar where the analysis repeats a lot of what Kevin did back in November: http://bit.ly/IRGGvi
normalman makes a good point, the utility of the player and how he is able to contribute to wins is significant, but additional wins is not enough if it does not significantly raise the chances of winning a superbowl. A team that drafts well and makes decisions that allows it to be consistently finshing in the 7-9 wins category and very infrequently finishing outside of that is inferior to one that on average gets maybe 5-8 wins but some seasons they get 3 or 4 and others they get 14 and are a clearly superior team. A high varience strategy in the NFL is superior, provided the goal is winning the superbowl and anything else is considered a failure. However, with that being said, I believe the superior strategy is to enter rebuilding mode more frequently, trading every pick this year for a pick higher next year, and repeat, also trading players over the age of 28 (although some positions sooner, others later) and continuing to roll over draft picks until One has built up an arsenal of picks. I also think the NFL has yet to bridge the gap between stats and analytics and decision making like Baseball has done in the post moneyball era. NFL teams should assess probabilities and standard deviation and make decisions that comply with their goals. Personally, I think if teams regularly would trade players after the age of 28 and constantly roll over draft choices until it could “live off the interest” trading most of the picks but still getting several 1st rounders every single year, it could easily become a dynasty and become very difficult to keep up with until adjustments to the “draft value chart” are made.
I looked at this a different way and got to the same answer.
Do the picks for which teams traded up have a higher success rate than others (i.e., are the teams more certain on those so the expense might be justified)? Since the picks do not have a higher success rate, you can’t justify giving away value (net of trade) to move up. Still possibility that trading up makes sense for teams with few holes as a form of concentrating the value of their picks into a smaller set of players, but that would require more research. Check out the analysis here: http://www.sportsplusnumbers.com/2013/05/luck-vs-skill-are-picks-after-trading.html