Trade Value

[JoshEngleman]

I've been thinking about putting a trade value calculator together for a while, so I thought this would be a good place to bring it up. For this exercise, ignore the specific talent ratings for players. Just try to look at this snippet in a vacuum.

In baseball, it's pretty easy to manage trade value, since it is difficult for any single player to make a large difference in a season. Obviously, the NBA is different. What can we look at to "grade" trades?

For example, Jae Crowder is expected to produce 17.9 wins over the next 3 seasons, while making a total of $21.9m. Amazing deal. It would produce a surplus of $34.3m over that time period. Anthony Davis, on the other hand, will produce 34.8 wins over the next 4 seasons, while making a total of $105.1m. This is a total of $6.6m in surplus value. In a vacuum, you would rather have Crowder than AD. Unfortunately, that's not how the NBA works. Every single GM in the league would trade Jae Crowder straight-up for Anthony Davis, so surplus value can't be the main measure for potential trades.

So, what is the next step? How can we best evaluate the components of a trade? Or, is everyone just wrong?

[Crow]

Some trades should have a 3-5-more year time horizon. Some are about this season. Probably should look at trades in at least 2 ways, this season and say 3 years.

A simple model would be increase net wins in either or both timeframes while being legal. So cost doesn't matter if it is within CBA constraints. Can modify and say spending over tax line or some other marker is disliked at some rate such that moves are rejected unless they produce more than a certain amount of wins or # wins above 50-55plus per million or $10 million net expenditure. If there is a hard stop, acknowledge where it is or appears to be.

Trades, especially early in off season, measured against what you actually have and that plus expected value of remaining discretionary spending BASED ON WHERE THE MARKET IS / who is left & likely cost. It is Crowder, w, x, y and the possibility of a z1 and z2 later vs. the bigger star for the same share of money or what you can / will spend legally. Overall portfolio analysis all the time, not trade in isolation. Also bed to look at minutes, actual playing time & usage projections and redundancies.

To the extent that player contract situation bring or lose cap stretching in future years via cap hold vs actual worth / future pay) their cost should be adjusted so long as you stay within the legal & desired spending zone.

Instead of having a fixed value of what you'd pay per win, have a marginal willing to pay pay curve that goes up as wins go up from 40 to 45 to 50, 55 and beyond. Respect the minimum team salad boundary. If you want, tack on the draft value implications to future year team total asset "values". Know where you are trying to peak and what division, conference and league banners are worth beyond the regular season wins, psychically and financially (operating basis and capital asset perspectives).

[jgoldstein34]

I've always thought of trade value as what can be reasonably expected in return, not necessarily what the player is individually worth.

I haven't researched it, but it seems that the better a player is, the less % of that value you can reasonably expect in return in a trade. Similarly, the better value a player's contract is, the less % of that value that can be returned. I also think real contract $'s matters as much as impact on the court. Young players obviously get a boost because of *potential*.

So while a player like Jae Crowder is a value on his contract, he still is only on a small contract and can't expect to net a huge return and that return likely won't have as good of a deal.

Trade Value is a really tough thing to model because so much of it is situational. A player can be valuable to one team and useless to another. It's much tougher to quantify than impact value.

[bondom343]

Trade value differs for each team too, it doesn't exist in a vacuum. What a player is worth depends on who is trading for him and moving him, and it can vary somuch I'm not ure this is possible to define with a calculation. An in teresting exercise for sure though, and I'd be intrigued as I'm really into this side of the NBA (just retired as moderator on RealGM's trade board).

The problem comes in in a few ways:

1. Draft picks - Picks don't have a defined value, and even assuming an average for one it depends on the GM doing the drafting for expected return on value as well as decreasing value after use. A pick before it's used is like a new car, but as soon as it's made the pick loses value to many (would you rather the number one overall pick in 2013 or Anthony Bennett?).

2. Fit - Roster construction is big here. I'd say a guy like Deandre Jordan is relatively pretty valuable. But would the Pelicans think so with Cousins and Davis? Doubtful. If they got amazing value and gave up Holiday for him you'd say "they won" but in reality probably got much worse on court.

3. Team Contention - Is the team rebuilding or win now oriented? Melo for example has totally different value to the Rockets and the Magic. Same for a guy like Fultz.

There are just so many confounding factors, but it's something I really find fascinating honestly.

[Crow]

If you had a really advanced trade model you might estimate the value of players and all the major player interactions down to four factor levels. And then it might be smart to assess needs and needs by position in the abstract and then look for the player that best fits that profile / factor mix instead of hunting for players and then considering the factor profile after falling in love with 1-2 factors or his stature. You certainly could / should rate players for your team, teams in general or specific other teams. Cases where the valuations vary most become the keepers & the guys perhaps best suited to be traded.

There isn't a clear, comprehensive, objective trade valuation model in public to my knowledge. But there could be with enough effort. I mean if I gave it a day I'd have one. A week a better one, a month, a year better & better. The fact that there isn't a complete example in public raises suspicions that most teams don't have a well documented one in use. It sounds & looks like a lot of ad hoc juggling. Most times pretty rationale but some mistakes creep in that perhaps could be caught / lessened with more systematic approach.

Teams should certainly forecast their team with and without a move using a simulator of kind, even against specific top competitors. Celtics moves might look better against Cavs & Warriors than in general (or not, I haven't run it, just guessing motivation / possible outcome of small-ball, offensive lean.)

Somewhere recently I came across someone who presented the factor level correlations who said they weren't statistically significant. Maybe not at league level. But maybe at the level of player types or specific players and in specific situations? Use that or not so much? Plenty to study, build on or move beyond, if it had a chance to be applied.

Another approach would be to build a realistic dream team then build as close to that as possible by finding the implicit values for players by role. Like maybe a top 5 wing at his peak supported by at least elite PG and big on second contracts. Supplemented by at least two above average shooters (proven quality at higher usage but also willing to accept less) and at least two above average defenders. Either young & cheap or older & relatively cheap. And then role players that fit within very tight salaries and specific cap rules.

[JoshEngleman]

I guess I should have been more specific in my original post, but I like a lot of the replies here. My original thought was more along the lines of Bill Simmons' old trade value columns. Boiling the combination of age, talent and contract down into one ranking.

In a generic trade setting, we would be setting team specific discount rates for wins. So, the Nets would value wins in 17-18 essentially at zero, while wins in 20-21 are dramatically more important. The opposite would be said for a team like Cleveland.

[DSMok1]

A couple of thoughts here:

First of all, the wins measure has to be relatively accurate, and the projection should deal with aging well. In this particular list, Porzingis is listed as the lesser player versus Crowder, and is projected to remain so. Do we agree with that? Does Porzingis have some potential upside (high variance--chance at being a star) that isn't accounted for in a "mean expectation" model?

Secondly, and I think this is significant: Superstars are worth a premium. If a team is chasing a deep playoff run, we should probably look at value over average rather than value over replacement player for their pay scale. In other words--if you're deep in the playoffs, anybody below average approaches being unplayable, a "replacement level" player in that environment. How to account for that, I'm not completely sure, but it's a very real phenomenon.

[JoshEngleman]

A couple of thoughts here:

First of all, the wins measure has to be relatively accurate, and the projection should deal with aging well. In this particular list, Porzingis is listed as the lesser player versus Crowder, and is projected to remain so. Do we agree with that? Does Porzingis have some potential upside (high variance--chance at being a star) that isn't accounted for in a "mean expectation" model?

Secondly, and I think this is significant: Superstars are worth a premium. If a team is chasing a deep playoff run, we should probably look at value over average rather than value over replacement player for their pay scale. In other words--if you're deep in the playoffs, anybody below average approaches being unplayable, a "replacement level" player in that environment. How to account for that, I'm not completely sure, but it's a very real phenomenon.

Totally agree on the second part. That's what I started thinking about. There's a transition point where talent becomes dramatically more important than salary value. I would guess each person would value it differently. If you had Jae Crowder at the minimum for the next 3 years, how much closer are we to a straight trade for Anthony Davis?

For the Porzinigs example, he's clearly someone you need to look at in longer scope than just "current contract".

[Crow]

Rich Cho was said to have developed a one number player valuation system with a couple of ex-Microsoft guys, back in his Sonics days I think. Maybe at same time Dean Oliver was there or maybe after. Dean being a former insider quoted as saying player rankings aren't something teams do a lot (maybe he didn't).


If you are a top 8-12 team, you could focus on now values at least part of the time. If you aren't top 4 you probably shouldn't overdo this. Mid pack or lower, the focus should be almost entirely on future value or asset value.

[jgoldstein34]

Just thinking out loud here, but what are all the variables you'd have to feed a model to teach it trade value?

Team record
Roster construction
Cap space, now and future
Age
Some measure of player ability
Player durability
Contract
How close a team is to winning a chip
Player potential
Draft picks
Free agent class

Is there anything major I'm forgetting?

[Crow]

I'd think you'd want to take player "ability" to team, position, player level 4 Factors.

[jgoldstein34]

Found this formula for trade value recently. Seems alright for estimating it.

[Crow]

That is from Dean Oliver and is around 15-20 years old.

You've checked it? A graph would be useful and a few examples.

[jgoldstein34]

Sure, here's a link to how it estimates trade value for 2016-17 regular season: Link

And here's a link I found for explaining how to calculate all the parts of it better: Link

The general form of it actually seems okay, but I'd take issue more with the credits/approximate value equation and the years left equation than the actual trade value formula.

[JoshEngleman]

I used the DO method with a few tweaks and wanted to share what I have so far. First, instead of units Credits/AV, I'm using my Talent metric converted to wins. I probably should use a combination of both, as someone like Andrew Wiggins will grade out worse for me, but in theory, he would be much closer to the top. Anyway, I also added a slightly different "Years Left" formula. In DO's, Lebron ends up with only 3 remaining years, which doesn't work for me. I put a quick formula together to base years on talent level + age. For the Trade Value calc itself, it will use whichever years number is larger. Finally, in the Trade Value formula itself, I switched from Wins to Wins^1.3. No reason other than I needed to bump up LeBron. If he's grading out in the 40s, it needs a tweak. This pushes him as close to 13, which is about as high as I can get him without major negative consequences. Also, the Wins are calculated using '16-17 minutes, but in theory, each one should be set individual based on expectation moving forward. If I bump Nerlens to 2000 minutes, he jumps into the top 25 of trade value. That'll likely be my next step.

I definitely want to make tweaks to the formulas that handle years left. There also needs to be a small function with regards to salary.

https://docs.google.com/spreadsheets/d/ ... 1067687166