Draft night trades are a regular occurrence. With so much uncertainty in the potential of different players, how can a team weigh the relative trade value of different picks?

Every year during the NBA Draft, multiple picks are moved, often for other picks. Even more are rumored to be on the move, as trade rumors and speculation are often at their zenith on draft night. Teams move up and down, trading picks for picks, and fans are often left wondering which side of the deal they would rather be on. How many second round picks would it take to fairly buy a lottery pick? Should you deal the No. 11 pick for No. 16 and No. 19?

While each draft and particularly each player prospect are different, the draft position itself has a definitive and calculable expected value, based on the average value attained by the players drafted in that position before. If one can define the expected value of draft positions objectively, one can then define which team got the better side of a deal when draft positions are traded.

Calculating the expected value of each draft position is relatively simple. There are several strong player value metrics out there, and the best can be objectively used to assign value to draft positions. For the purposes of this analysis, I used VORP (value over replacement player) derived from BPM (Box Plus-Minus). BPM, a box score based estimate of a player’s net impact per 100 possessions, is second only behind metrics of the RAPM family with regards to predictive ability amongst publicly available player value metrics.

To condense the differing VORPs of a player’s seasons into one value representative of his career, I took a two-year-max approach. This approach assigns the value of a player as the best two-year average VORP value of his career, allowing players with shorter careers to be as good as players with longer careers, while still requiring players to have at least two good years of play in a row if they are to score well. Summarizing careers in this fashion is often the standard in draft projection models, including mine.

With a set value representing how good a player is, it is quite easy to estimate the expected value of a draft pick: Simply average the value of the players drafted at each draft position over a set period of time.

Without adjustment, however, this leads to some wonky results. An oft-cited true fact is that the No. 3 overall pick has yielded better players than the No. 2 overall pick since 1984, and there are many similar examples of later picks historically outperforming earlier picks. Using BPM-derived VORP as the bellwether of value, the No. 13 pick has been better than the No. 12 pick, the No. 21 pick has been better than picks No. 14-20 and so on and so forth.

Few would argue that having the No. 3 pick is better than the No. 2 overall pick, and if they did, they would of course, be wrong. NBA teams have gotten the draft roughly right since its inception, but if teams drafted with 100% efficiency, the No. 1 pick would always be better than the No. 2 pick which would always be better than the No. 3 and etc. etc. Picking first is obviously the most desirable position.

To this end, I applied a loess (local regression) function to the average value of draft pick positions. This function serves to smooth the curve of draft pick value, and assign value as reflective of not only the pick itself, but of the other picks around it. Under this methodology, each draft pick is more valuable than the one that follows, and is thus more accurate as a measure of expected draft pick value for the purposes of evaluating trades. Both results, raw and local regression transformed, are presented below. All values can be interpreted as wins attributable to a player during one season.

For the purpose of evaluating trades, the smoothed value is likely to be superior to the raw average, save for one position, the No. 1 overall pick. The reason for this is that for all picks but the No. 1 and No. 60, the smoothed value is computed from the averages of picks both above and below the given position. The first overall pick is unfairly brought down in the smoothed curve by the results of only players picked after that position. To this end I use the raw average of the No. 1 overall pick as representative of its fair trade value.

So what can be learned from this table? In terms of expected value, our examples from above can now be quantified. It would take roughly three second round picks in the No. 30-40 range to cumulatively outweigh the value of the No. 16 pick, the last pick in the lottery. A team that trades the No. 11 pick (3.81 Expected VORP) for No. 16 and No. 19 may be giving up a chance at the best player, but the two picks would be expected to yield a higher cumulative value (4.74 Expected VORP). There are a few other lessons to be gleaned as well.

The first lesson is that draft value is hardly linear. The difference in value between the No. 1 and No. 2 overall pick is almost 3.5 times the difference in value between the No. 30 and No. 60 pick. Value falls off quick in the NBA draft.

The next lesson is a point I have tried to stress for years: The player your team drafted at 29th overall is overwhelmingly unlikely to become an All-Star. High hopes and optimism are rampant around the draft (which of course, adds to the fun), but the cold hard truth is that the majority of drafted players play no more than a few total minutes in the league, even though there are only 60 of them every year. Major sports networks have been no help in dispelling this notion, with on-air personalities comparing players just taken with the No. 16 overall pick to current NBA veterans on near-max contracts with horrifying regularity.

Another important idea to consider when comparing the trade value of draft positions is the idea of diversification. Depending entirely on the overwhelmingly unique context of each draft and each player, there is value in allocating your risk exposure to multiple picks. Would you rather have all your eggs in one basket, a basket potentially devastated by one misstep on the basketball court, or in two baskets, with two different players, each occupying different areas of space and time, each responding differently to coaches and team culture, and each with their own separate chances of developing into a star? Especially later in the draft, I quickly start to side with the latter.

It is important to remember that each draft is drastically different, both in expected value for picks and the shape at which value falls off. For the average draft, both the expected values of each pick and the curve of draft value as presented by the table should prove relatively accurate. Compare the pick values in the table as you will, and use them as a reference for when your favorite teams trade picks for picks, picks for players, or any combination of the two.

Find more NBA Draft content at our FanSided NBA Draft hub and for updated scouting reports of all NBA draft prospects, check out Upside and Motor.