The Best and Worst Trades, Graded Twice
Every NBA trade since 1999, judged once on what was knowable the day it was made and once on how it turned out. They are not the same list, and the gap between them is the whole story.
Poker players have a word for judging a decision by how it turned out: resulting. It is the most common mistake in sports analysis, and the most fun. The team that traded for the guy who got hurt made a bad trade. The team that drafted the guy who became a star made a great pick. Case closed, next caller.
The trades page grades every NBA trade since 1999-00 two ways, and this post is about the difference. Process is what the Trade Lab would have said on the day, using only what was known then: the ratings, ages, contracts and picks as they stood. Result is what actually happened to the players and picks that moved. Both are measured as the change in each team’s roster value, in today’s money.1
Across 2,763 team-sides the two grades are correlated at 0.24. A good process buys you a better result on average, and not much more than that. Remember the number. It comes back.
The best process
The single best trade the model has ever graded on the day it was made is one nobody thinks of as a trade at all: the 2010 sign-and-trade that sent LeBron James to Miami. The Heat’s side grades at +$1708M; Cleveland’s is the worst process on record, at −$2191M. In fairness to Cleveland, its alternative was getting nothing at all, and this is not the kind of decision the Lab can see.2
| Trade | Side | Process | Result |
|---|---|---|---|
| 2010 LeBron James | MIA | +$1708M | +$871M |
| 2007 Kevin Garnett, Ryan Gomes, Gerald Green +3 | BOS | +$758M | +$378M |
| 2005 Eddy Curry, Antonio Davis, Jermaine Jackson +2 | CHI | +$392M | +$457M |
| 2017 Jae Crowder, Kyrie Irving, Isaiah Thomas +1 | CLE | +$383M | +$68M |
| 2017 Jimmy Butler III, Kris Dunn, Zach LaVine +2 | MIN | +$326M | +$80M |
| 2004 Caron Butler, Brian Grant, Lamar Odom +1 | MIA | +$324M | −$294M |
The best result
And here is the headline. The best result of any trade on the board is Oklahoma City’s side of the 2019 deal that sent Paul George to the Clippers and Shai Gilgeous-Alexander the other way, along with a stack of picks: +$1910M. On the day, the model graded Oklahoma City’s side at −$23M, which is to say roughly even. The Clippers’ side graded at −$33M and has returned −$1910M.
That is not the model being wrong. It was a fair trade the day it was made. A second-year guard with a good rookie season was worth about what the model said he was worth, and nobody, including the Thunder, knew he would become an MVP. The Thunder deserve credit for the process: the picks, the patience, the bet. They do not deserve credit for knowing. Nobody knew. That’s what makes it a great story instead of a great decision.
| Trade | Side | Process | Result |
|---|---|---|---|
| 2019 Danilo Gallinari, Paul George, Shai Gilgeous-Alexander | OKC | −$23M | +$1910M |
| 2012 Cole Aldrich, Daequan Cook, James Harden +3 | HOU | +$26M | +$1109M |
| 2011 Al-Farouq Aminu, Eric Gordon, Chris Kaman +1 | LAC | +$280M | +$1037M |
| 2012 Kyle Lowry, Gary Forbes | TOR | −$58M | +$1030M |
| 2013 Rudy Gobert, Erick Green | UTA | −$5.8M | +$930M |
| 2010 LeBron James | MIA | +$1708M | +$871M |
Good process, bad luck
The sides the model liked on the day that went on to lose the most. In most cases the reason is on the other side of the trade: somebody they gave away became much better than anyone could have expected.
| Trade | Side | Process | Result |
|---|---|---|---|
| 2012 Kyle Lowry, Gary Forbes | HOU | +$9.1M | −$1032M |
| 2013 Rudy Gobert, Erick Green | DEN | +$21M | −$930M |
| 2000 Chucky Atkins, Grant Hill, Ben Wallace | ORL | +$235M | −$794M |
| 2004 Luol Deng, Jackson Vroman | PHO | +$123M | −$750M |
| 2021 Alperen Sengun | OKC | +$169M | −$695M |
| 2002 Marcus Camby, Nene, Mark Jackson +2 | NYK | +$8.4M | −$660M |
Bad process, good luck
And the mirror image: sides the model graded as losers on the day that came out furthest ahead. Some of them were smart in ways the model couldn’t see. Most of them were lucky. The model can’t tell those apart, and, if you are being honest, neither can you.
| Trade | Side | Process | Result |
|---|---|---|---|
| 2019 Danilo Gallinari, Paul George, Shai Gilgeous-Alexander | OKC | −$23M | +$1910M |
| 2012 Kyle Lowry, Gary Forbes | TOR | −$58M | +$1030M |
| 2013 Rudy Gobert, Erick Green | UTA | −$5.8M | +$930M |
| 2000 Chucky Atkins, Grant Hill, Ben Wallace | DET | −$338M | +$794M |
| 2004 Luol Deng, Jackson Vroman | CHI | −$154M | +$710M |
| 2021 Alperen Sengun | HOU | −$212M | +$695M |
What to take from it
A correlation of 0.24 means a front office that gets every process call right will still look foolish on a good share of its trades, and one that gets them wrong will look like geniuses on some. Judge the process. It’s the only part anyone controls. Then check back in five years anyway, because it is extremely fun.
Figures: the process and result grades on the trades page, 1999-00 to 2025-26, as of this build. Wins are on this model’s own scale and dollars at the open-market price of a win; the method explains both.
- Each side’s value as a share of that season’s cap, times today’s. Result lists include only trades made at least four seasons ago, so most of the contracts involved have run their course. How the Lab works has the rest. ↩
- The Lab compares a roster with the trade against the same roster without it. For a player who was leaving anyway, “without it” is not really on the menu. ↩
