Monday Tip-Off: The Mixed Bag of CPU Teammates in 2K Pro-Am

We’re at midcourt, and the ball is about to go up…it’s Monday Tip-Off! Start your week here at the NLSC with a feature that’s dedicated to opinions, commentary, and other fun stuff related to NBA Live, NBA 2K, and other basketball video games. This week, I’m tipping things off with a few thoughts on the quality of CPU-controlled teammates in NBA 2K17’s 2K Pro-Am.

As the artificial intelligence in basketball video games has become more sophisticated, CPU-controlled teammates have thankfully become much more reliable. That’s not to say that there aren’t any frustrating moments where they seemingly forget how to play basketball, but compared to early hoops titles, there’s less of a need to frantically switch control of players and take charge of everything yourself. Needless to say, if you’re locked to controlling a single player – as is the basis of modes such as MyCAREER and 2K Pro-Am – it’s even more important that CPU-controlled teammates are competent.

Barring a connection problem, you’ll never start a game of 2K Pro-Am with more than two AI players. If you consistently run with a full squad of five players, they’ll seldom be an issue at all unless someone fouls out, or is booted due to a low teammate grade. However, if you frequently jump online to play 2K Pro-Am, chances are you’ll deal with a CPU-controlled teammate at some point. It’s at that point you’ll discover that although AI in basketball games has come a long way, the quality of your CPU teammates is still very much a mixed bag.

At first glance, this shouldn’t be possible. After NBA 2K16 masked the Overall Ratings of CPU teammates in 2K Pro-Am while giving them randomised names, NBA 2K17 went in the opposite direction, calling them all “AI Player” and revealing their Overall Rating to be 55. With no variation in the Overall Rating, you wouldn’t expect the individual ratings to fluctuate very much, and thus the quality of CPU teammates should remain fairly consistent. At 55 Overall – equivalent to a newly created MyPLAYER in NBA 2K17 – you’d also expect them to be consistently useless. Fortunately, that isn’t the case…well, not always, at any rate.

However, there does seem to be a noticeable difference in the quality of CPU teammates, and like Forrest Gump’s infamous box of chocolates, you never know what you’re going to get. Although many of NLSC THRILLHO’s games have featured a full complement of five users, there have been many games of 2K Pro-Am where we’ve had to run with one or two CPU-controlled teammates. Sometimes, those teammates have been an asset: they’ve helped keep the ball moving, made some extra buckets, done the dirty work, and basically filled their role fairly competently. Other times, they’ve felt like a complete liability.

The more I’ve thought about it, the more I think that Archetypes are coming into play here. It’s the only way that ratings for AI Players could feasibly differ so much from game to game, and the Archetypes of CPU teammates do seem to be chosen rather randomly. While a 55 Overall player obviously isn’t going to have a lot of strengths, they’ll at least be better in certain areas compared to others, according to their Archetype. Those skills aren’t necessarily going to mesh well with the users’ MyPLAYERs, or fill a need. It therefore seems logical that the random Archetypes are responsible for the varying usefulness of CPU teammates in 2K Pro-Am.

Using NLSC THRILLHO as an example, no one on our squad has created a seven foot centre that can dominate the paint at both ends. Despite the implementation of Archetypes, there are still cheesy builds, which is an issue that I’ve discussed before. Even putting aside any cheesiness, our lack of size hurts us when we come up against a big frontcourt, especially if there’s a Glass Cleaner on the other team. There are times when our CPU teammates seem to be able to at least help out a little bit in those situations, or show uncharacteristic aggressiveness in the paint when on offense. I’d suggest that those are the times when their Archetype covers for our deficiencies.

Naturally, there are also plenty of times when they don’t help balance out our lineup. Of course, at 55 Overall, even players who do suit our needs are limited in their effectiveness, so a CPU player with an unsuitable skill set essentially brings nothing to the table. CPU teammates are seemingly prone to over-dribbling, are often slow to make the pass (which leads to mashing of the pass button, which leads to queuing up unwanted passes), and have a frustrating habit of either not taking open shots or immediately taking contested ones. Some seem to be a bit smarter though – again, according to Archetype – and all AI Players tend to be good for some cheap steals.

Perhaps even more interesting than the varying quality and effectiveness of CPU teammates is how the AI handles MyPLAYERs when their user is disconnected, fouls out, or is booted from the game after falling below a D-minus teammate grade. At times, it seems as though the AI tries to imitate the MyPLAYER’s tendencies, but suffice to say, it doesn’t really take advantage of having a far more competent player at its disposal, or make the smart decisions a user likely would. If a user gets kicked and the CPU takes over, chances are you’ll end up treating them like one of the anonymous AI Players, keeping the ball out of their hands as much as possible.

In all fairness, CPU-controlled teammates have improved a little in NBA 2K17, at least in 2K Pro-Am. They’ll make the same mistakes as CPU-controlled players in other modes, especially when it comes to rotations and defending the pick and roll, but they’re a little more reliable than they were in NBA 2K16. In NBA 2K16, playing with a couple of AI Players was a lot tougher, whereas this year we’ve been able to pick up a fair amount of wins with only three users. It takes some strong performances all around however, especially when you’re matched up with a squad that has four or five users. Such games are winnable, but you’re definitely at a disadvantage.

To that end, I’d advocate increasing the ratings of the AI Players so that they’re at least on par with an average NBA journeyman. On one hand, you could argue that this defeats the purpose of 2K Pro-Am, which is more fun when you’ve got five users a side. You also don’t want to make the mode too easy, with the AI Players being a cheat. At the same time, there will be occasions when a squad will play with less than five users; since that’s an option, it shouldn’t be an unfair disadvantage. A team of five users who’ve maxed out their ratings will still have an advantage, but more competent CPU teammates will at least give undermanned squads a fighting chance.

Future NBA 2K games could also choose more suitable Archetypes for CPU teammates, or perhaps even give squads their choice of players. In fact, if Visual Concepts really wanted to put the “Pro” in 2K Pro-Am, randomly assigned or selectable NBA players to fill out squads with less than five users would be a fun idea. That might sound like it would be stacking the deck, but no more so than having five MyPLAYERs who are all rated 95 Overall or better. In the meantime, however, I guess it will remain a game of chance, with CPU teammates continuing to get frozen out for most of the game.

Support The NLSC on Patreon!