Winning Deals: using simulation analysis to assess NBA trades

In this article: 1) Philadelphia vs. Denver (the "Iverson" deal) 2) Golden State vs Indiana (the eight player "blockbuster")

What if? What if the Blazers had drafted Michael Jordan? What if players like Grant Hill, Jamal Mashburn and Kenyon Martin had been able to stay healthy through their primes? What if Karl Malone and John Stockton had not had played in the same era as Jordan’s Bulls? What if Portland had kept its 1999-2001 teams together (and out of trouble, and healthy)? What if questions like these are at the root of many sports questions, fueling intense discussion and debate. WhatifSports.com adds to the debate with in-depth, detailed, online simulations for basketball as well as football, baseball, hockey and auto racing all with full play-by-play, comprehensive boxscores and exhaustive statistics.

Possibly the most intriguing “what if” scenarios come about at this time of each season - around the trade deadline. What if Kidd goes to the Lakers? Or Ridnour to the Hawks? Or Gasol to just about any team in the NBA? Rather than speculate too much at this point though, we would like to utilize the simulation technology found at WhatIfSports.com (and available free to the public) to review two of the biggest trades from the season thus far. Then, after the upcoming trade deadline, we will provide a more condensed version of the analysis on many of the new trades. Other websites have tools for fans test potential trade feasibility under the cap. Well, consider those the precursors to this analysis.

Assuming that all franchises are pressured into a “win now at any cost” strategy (a fair, but not always reasonable assumption), we will review the Philadelphia/Denver and Golden State/Indiana trades to attempt to determine who “won” each trade. To do fully do so, we have simulated two different types of matchups for each team involved 820 times (ten full seasons). This reduces variability in the results. We simulated the roster of a team previous to trade vs. roster of the same team following the trade and then those two rosters against a control team in separate 820 game simulations. To build rosters, we used the active roster immediately previous to the trade and the active roster today. This means that Steve Blake made the post-trade Nuggets and not Earl Boykins. And, in order to test before and after with oft-used lineups and well distributed playing time, we left Jason Richardson off of both Golden State teams, giving Monta Ellis his full minutes. We also utilized a hybrid of this season’s stats and the player’s previous full-season’s stats for his non-playing time inputs as we felt that this would better represent a player’s skill, ability and role than just using 15 games of data or so from the beginning of the season. This analysis does not account for potential injuries (clearly inevitable with the some of the players involved), future seasons (downplaying the significance of draft choices and young players like Ike Diogu) or potential roster moves by these teams this season.

About the Engine:
While this analysis is not intended to go into great depth concerning the simulation engine, a few points should be mentioned. The simulation engine is designed to be able to handle teams constructed of players from various real-life squads, seasons and eras. This allows fans the opportunity to “see” how a team of Magic, Jordan, Bird, Malone and Olajuwon would do against Cousy, Oscar, Baylor, Russell and Chamberlain and so on. The significance of that fact is that the engine must account for every detail. Many simulations can simply regurgitate actual stats based on team aggregate stats and fixed individual probabilities, but this engine analyzes every individual matchup, team chemistry, shot distribution, ball-handling, individual and help defense and possession pace, often utilizing many of the more “advanced” statistics and metrics gaining recent attention in basketball and several years of our own research.

With questions or any more information concerning the simulation engine, please contact us at the email address below. These questions may lead to another article on this site that further explains every “sim” interaction and much of our research.

About the Control Team:
The control team is completely average. We built a team consisting of one player at each position who can play all 48 minutes of every game. Each of these player’s statistical inputs employed by the sim engine are exactly average for that position. This means that, while we should not see the large game-to-game variation that could normally be found in a regular season, players and teams will ultimately do the same against this team as they would over a balanced schedule including every team in the league.

As you will see, all of the teams in this analysis do poorly against the control. This has a few possible implications. First of all, these teams may not be very good no matter the result of the trade. And secondly, in the NBA it may be more beneficial not to have many weaknesses than it is to have a few strengths, implying that the difference between “good” and average is less significant than the difference between average and “weak” and that basketball is more a game about exploiting weaknesses than anything else.

The first point definitely has some truth to it. Based on records alone (not necessarily the best measure considering injury and other trouble from some of these teams), all four teams are close to or below average and none of them is within four games of its division lead (at the all-star break). Indiana and Denver would be in the playoffs as the five and seven seeds respectively and Golden State and Philadelphia would be in the lottery.

The second claim is a little bolder. And, while we cannot say we disagree, we are not prepared to fully defend it here. Instead, we will present a known success versus the control to illustrate that a team, even one with a couple obvious below average attributes, can soundly defeat the control.

We simulated the current (if all healthy) Phoenix Suns against the control 820 times to find:

The Philadelphia/Denver Trade:
The 76ers and Nuggets made headlines on December 20, 2006 when Allen Iverson and Ivan McFarlin were traded for Andre Miller, Joe Smith and 2007 First Round picks. First, let us analyze how Philadelphia, a team that was 5-19 (6-11 with Iverson in the lineup) at the time of the trade, fared in the deal.

We simulated the 76ers before the trade against the 76ers after the trade 820 times to find:

"Old vs New"	W%	Points	PPS	FG%	Oreb/gm	Dreb/gm	TO/gm	PF/gm
76ers Before	67%	100.9	1.27	46.9%	10.6	35.3	13.2	20.2
76ers After	33%	95.0	1.18	44.4%	11.4	35.7	14.2	22.3

Then, we simulated the 76ers BEFORE the trade against the control 820 times to find:

Team	Wins	W%	Points	PPS	FG%	Oreb/gm	Dreb/gm	TO/gm	PF/gm
76ers Before	26	31%	96.0	1.19	45.8%	10.6	30.4	14.1	21.3
Average Team	56	69%	101.4	1.26	47.4%	14.2	35.3	15.2	20.3

And, we simulated the 76ers AFTER the trade against the control 820 times to find:

Team	Wins	W%	Points	PPS	FG%	Oreb/gm	Dreb/gm	TO/gm	PF/gm
76ers After	16	19%	92.6	1.18	43.6%	11.5	30.1	14.9	22.5
Average Team	66	81%	101.4	1.36	47.8%	13.6	36.6	14.9	19.9

We would say that, over an 82 game season, we would expect the 76ers to be about ten wins worse (-10.2) after the trade.

The Before team probably is not twice as good as the After as the head-to-head matchup may insinuate. It is clear though, that the team became less efficient. This is very interesting considering the key part of the trade, Allen Iverson, left and he has not always been known for efficiency. However, without Iverson and Webber, the 76ers are left with a problem they have not had in ten years – more shots than players to take them (who will take them). Andre Miller may be a decent player, but he wants to set guys up and he does not really know who those guys are. The best answer now is Andre Iguodala, but he drops from an almost 50% field goal shooter, to an average shooter against the control. Still, for rebuilding, they are not that much worse than they were previously (especially taking headaches and money out of the equation).

Now, let us look at the Nuggets.

We simulated the Nuggets before the trade against the Nuggets after the trade 820 times to find:

"Old vs New"	W%	Points	PPS	FG%	Oreb/gm	Dreb/gm	TO/gm	PF/gm
Nuggets Before	41%	104.0	1.16	43.8%	13.5	32.7	17.0	23.5
Nuggets After	59%	107.3	1.21	44.9%	11.0	31.9	15.2	21.4

Then, we simulated the Nuggets BEFORE the trade against the control 820 times to find:

Team	Wins	W%	Points	PPS	FG%	Oreb/gm	Dreb/gm	TO/gm	PF/gm
Nuggets Before	34	41%	104.4	1.24	44.7%	12.8	30.3	16.4	23.3
Average Team	48	59%	107.9	1.27	46.3%	12.6	31.3	16.7	22.2

And, we simulated the Nuggets AFTER the trade against the control 820 times to find:

Team	Wins	W%	Points	PPS	FG%	Oreb/gm	Dreb/gm	TO/gm	PF/gm
Nuggets After	40	49%	107.0	1.26	46.1%	11.9	30.0	16.2	22.1
Average Team	42	51%	108.2	1.26	46.8%	12.6	31.0	17.4	22.8

We would say that, over an 82 game season then, we would expect the Nuggets to be between six and seven (+6.4) wins better after the trade.

The After team is the best team out of all four clubs – eight different rosters – in this analysis. They can push the tempo, while still being efficient and playing adequate defense (for the style). Turnovers are a little high, but that is more a product of pace. We believe this analysis illustrates how the After Nuggets are a much better fit for their style of play.

WhatIfSports Simulator Trade Summary (Team Wins)

Team	Before	After	Net
Philadelphia	25.8	15.6	-10.2
Denver	33.6	40.0	+6.4
Verdict: DENVER Wins			+16.6

We would give the Nuggets a +16.6 win margin based on the before and after differences of both teams against the control.

Clearly, the Nuggets “won” this trade as they were already the better team and improved, while the 76ers did not. That being said, is it worth it? For the buzz the team has generated by making this move, the Nuggets are essentially an “average” NBA team, completely out-classed by a team like the Suns. With the balance of power how it is in the league, being “average” in the Western Conference does not even guarantee a playoff spot.

The Golden State/Indiana Trade:
The Pacers and Warriors made some headlines as well when, on January 18, 2007, Al Harrington, Stephen Jackson, Josh Powell and Sarunas Jasikevicius were swapped for Mike Dunleavy Jr., Troy Murphy, Ike Diogu and Keith McLeod. First, let us analyze how the Warriors fared.

We simulated the Warriors before the trade against the Warriors after the trade 820 times to find:

"Old vs New"	W%	Points	PPS	FG%	Oreb/gm	Dreb/gm	TO/gm	PF/gm
Golden State Before	45%	103.5	1.19	45.6%	14.1	31.2	16.7	23.6
Golden State After	55%	106.6	1.24	47.6%	11.3	31.0	15.2	22.9

Then, we simulated the Warriors BEFORE the trade against the control 820 times to find:

Team	Wins	W%	Points	PPS	FG%	Oreb/gm	Dreb/gm	TO/gm	PF/gm
Golden State Before	22	27%	103.4	1.21	44.3%	11.6	30.1	16.5	23.1
Average Team	60	73%	111.7	1.27	47.4%	11.1	34.2	15.0	21.8

And, we simulated the Warriors AFTER the trade against the control 820 times to find:

Team	Wins	W%	Points	PPS	FG%	Oreb/gm	Dreb/gm	TO/gm	PF/gm
Golden State After	22	27%	104.6	1.23	45.7%	10.6	28.2	15.7	22.4
Average Team	60	73%	111.0	1.24	47.7%	13.2	34.0	16.8	21.5

We would say that, over an 82 game season then, we would expect the Warriors to stay the same or gain about one win (+0.4) after the trade.

This data suggests that, either way, the Golden State Warriors have been over-performing to-date this season, which is not a good sign for their fans. The new Warriors are technically better, but not significantly. Of note though, is that this is a very different team. They are not as strong at rebounding, but cause more turnovers and shoot at a higher percentage. They are now probably better suited for the tempo the team would like to run. On a game-by-game basis, this could make a big difference in outcome, but on the whole they are really just as “good” as before.

Now, let us look at the Pacers, who have played well enough thus far to warrant a five seed in the weak Eastern Conference.

We simulated the Pacers before the trade against the Pacers after the trade 820 times to find:

"Old vs New"	W%	Points	PPS	FG%	Oreb/gm	Dreb/gm	TO/gm	PF/gm
Pacers Before	66%	98.3	1.22	44.3%	12.8	32.2	14.9	22.8
Pacers After	34%	94.9	1.17	42.6%	14.9	33.8	16.9	23.9

Then, we simulated the Pacers BEFORE the trade against the control 820 times to find:

Team	Wins	W%	Points	PPS	FG%	Oreb/gm	Dreb/gm	TO/gm	PF/gm
Pacers Before	32	39%	96.3	1.20	44.3%	12.0	30.5	15.4	23.4
Average Team	50	61%	99.2	1.25	46.5%	11.8	32.7	14.7	22.5

And, we simulated the Pacers AFTER the trade against the control 820 times to find:

Team	Wins	W%	Points	PPS	FG%	Oreb/gm	Dreb/gm	TO/gm	PF/gm
Pacers After	19	23%	94.0	1.17	43.4%	14.0	31.9	16.6	23.9
Average Team	63	77%	100.3	1.27	47.5%	11.7	33.3	14.3	22.3

We would say that, over an 82 game season then, we would expect the Pacers to be 13 wins worse after the trade.

WhatIfSports Simulator Trade Summary (Team Wins)

Team	Before	After	Net
Golden State	22.0	22.4	+0.4
Indiana	32.2	19.2	-13.0
Verdict: GOLDEN STATE Wins			+13.4

We would give the Warriors a +13.4 win margin based on the before and after differences of both teams against the control.

So the Warriors “won” the trade by default. Instead of improving its team (on the court), the Pacers go from Eastern Conference contenders – a team on par with the Nuggets before their trade – to a team on par with the Golden State Warriors. With Dunleavy and Murphy, the Pacers do just about everything a little worse than before. Obviously, if Ike Diogu – an undersized power forward with serious durability concerns – becomes a star in the Elton Brand mold, this evaluation can change; but, especially given the amount on the contracts of its new players, it is difficult to understand how this trade makes any basketball sense for the Pacers. Of course that does not really denote good things for Golden State either. If Indiana was so eager to move a headache like Stephen Jackson and the inconsistent Al Harrington that it would pull off a lopsided deal like this, how long will it be before these players are shipped again?

Paul Bessire is the lead statistician and data analyst for WhatIfSports.com. Scott Eble is a senior developer and project leader for SimLeague Basketball at WhatIfSports.com. They would be happy to answer any questions concerning this analysis or anything related to the site and can be reached at pbessire@whatifsports.com