Trades Winds: Does Sports Game AI Match Up To Real GMs?
In the sporting world, trades spring from many different motivations. Some players grow disgruntled with their team role, clash with coaches, or see the writing on the wall with their talent-starved roster and ask to be moved to a contender. The mindset of the general manager also plays a key role. The best GMs think two or three moves ahead, sensing that shedding an aging player with an expiring contract for a pick in a particularly deep draft is a wise move. Others just love to tinker. Since being named general manager of the Minnesota Wild in 2009, Chuck Fletcher has made 28 trades in hopes of bolstering the roster. Capturing the intricacies of these decision-making processes during trades is one of the biggest challenges facing sports game designers today.
Most titles don't factor in player motivations at all - only the NBA 2K series takes into account player morale, which can be affected by winning or losing and whether or not the player is being given the role and floor time that was promised. Because the other games forgo this logic, it makes it tough for them to accurately recreate blockbuster trades driven by the player, like the Seattle Seahawks trading for disgruntled Vikings receiver Percy Harvin this offseason. To the current logic system in the Madden games, it makes no sense for the Vikings to trade away its best receiving asset (especially considering the extreme drop-off at the position). Without the knowledge that the player wants out and has no intention of resigning with the club, the AI has no reason to see this trade as the best-case scenario as it was for the Vikings in the real world.
How far away are game designers from instituting a trade logic that could mirror the reasons real-life trades happen? To find out, we tried to recreate the big trades from the NHL, NBA, and NFL on both sides of the propositions. We chose not to include the MLB games, because with the trade deadline just approaching, we don't have enough of a sample size to test the most current versions.
Taking this approach, we ran into some roadblocks on certain trades. In NHL 13, some of the young prospects like Filip Forsberg aren't in the game, nullifying his trade from Washington to Nashville for Martin Erat and Michael Latta. NBA 2K13 ran into salary cap problems when performing certain trades, making it impossible to pull off the move that sent Jose Calderon and Ed Davis to the Grizzlies in exchange for Raptors players Rudy Gay and Hamed Haddadi. Here is the breakdown of the trades we were able to attempt:
Madden NFL 13
Using the most current roster update, we tried to recreate the 12 NFL trades that generated headlines during the offseason. Proposing the move from each side of the agreement, that gave us 24 total attempts. No proposals resulted in both teams agreeing to the terms, and one side of the trades agreed to the deal in only 4 of the 24 attempts. This gave Madden developer EA Tiburon a 16-percent conversion rate, the worst of the three games.
As we mentioned earlier, both sides turned down the Percy Harvin trade. Surprisingly, the same happened with the Darrelle Revis trade. Why the Buccaneers would walk away from the best cornerback in the game when they only had to give up a first- and a fourth-round pick is beyond me. Another curious trade rejection came when the 49ers shunned Anquan Boldin's services despite only having to give up a sixth-round pick.
The successful trades included the Cardinals accepting the terms for Carson Palmer, Seattle approving the Matt Flynn trade, the Chiefs greenlighting the Javier Arenas trade, and the 49ers shipping Alex Smith to the Chiefs.
Even with a player morale system and trade finder in place, Visual Concepts didn't fare much better than Madden in mirroring the player movement of its league. Only 7 trades were accepted out of 24 attempts, a 29-percent conversion rate.
However, NBA 2K13 was the only game to approve a trade from both sides of the deal. The Rockets had no problem moving Marcus Morris for a second round pick, and the Suns felt the same way. Other successes included the Magic accepting Beno Udrih, Tobias Harris, and Doron Lamb in return for sending J.J. Reddick, Gustavo Ayon, and Ish Smith to the Bucks.
The biggest trade of last year - Oklahoma City's questionable decision to trade James Harden and three others to the Rockets for Kevin Martin, Jeremy Lamb, and picks - was accepted by the Thunder but turned down by the Rockets.
As I mentioned in my review, the inflexible GM logic in NHL 13 disappointed me, so I expected the game to perform just as boorishly in recreating the real-life trades. Though its performance wasn't great, the 33-percent conversion rate was the best among the games we tested.
Like Madden, no trade was accepted on both sides of the proposal in NHL 13. The biggest moves of the trade deadline that we could attempt - Marian Gaborik being shipped off to Columbus and Jason Pominville joining the Wild - were rejected by both sides. Other shot down trades included Calgary dishing Jay Bouwmeester to St. Louis, Jordan Leopold moving to the Blues, and Rafi Torres jumping to the Sharks.
The NHL logic fared much better accepting trades with role players on the tail end of their careers. The Rangers eagerly shipped Mike Rupp to the Wild for Darroll Powe and Nick Palmieri, Dallas gladly took Erik Cole off the hands of the Canadiens for Michael Ryder and a third-round pick, and Pittsburgh willingly relinquished two second-round picks for Sharks defensemen Douglas Murray.
These feeble conversion rates demonstrate just how far game designers are from creating a logic system that accurately understands the myriad motivations driving these trade decisions. We hope the designers can harness the power of next-generation consoles to develop trade brains that fare better in simulating the motivations of disgruntled players and wheeling and dealing mindsets of general managers.
Though you could argue that being able to recreate real-life trades has no bearing on making a trade system amenable to players, I think everyone would agree that having a trade logic that sees beyond the player ratings and accurately understands team needs, brings player morale into the equation, and is capable of thinking more than one move ahead would go a long way toward improving the franchise experience of these titles.