LA Noire

Fingerprints? We don't need no stinking fingerprints.

Despite being an all-round Rockstar fan, LA Noire and Bully are two of the more interesting Rockstar titles I never got around to playing. (I did play a bit of Red Dead Redemption, without really getting into it — I’m thinking of revisiting it once I finish LA Noire.)

This isn’t a review, and LA Noire has been out for quite some time (two years or so) so this is hardly timely, instead this is a discussion of a very interesting piece of game design that was built on top of Rockstar’s GTA Engine. As such I’m not going to discuss the content of the game (there’s a serial killer mystery which is pretty disturbing, although cleverly tucked into history) — I’m discussing LA Noire gameplay design, specifically the way mysteries are solved.

The signature feature of LA Noire is the very impressive facial capture / animation and the interrogation system built on top of it. It somewhat resembles what I considered one of the most brilliant game mechanics I’ve ever seen — unfortunately, not in a game but in a very expensive piece of multimedia computer-based training Andersen Consulting created for GE Capital back in the 90s — but I’d say the training product actually had a better thought out mechanic (more on this later).

Here’s how LA Noire works in a nutshell:

You visit a crime scene or other place of interest (e.g. the dwelling of a suspect or victim) and walk around. When you’re close to a piece of evidence, the controller vibrates. If you press “x” (often after walking around to try to get the vibration to recur) you’ll discover either one or several pieces of evidence. If one, you can pick it up, rotate it (sometimes), zoom in on it (sometimes), and it will sometimes cause a piece of evidence to appear in your notebook. There’s a subtle mechanic to tell you when you’ve found all the evidence at a crime scene (a musical swell).

Somewhat oddly, you manhandle every piece of evidence you see, despite the fact that fingerprints were used to obtain a conviction in the US as early as 1902, and you’re investigating very high profile murders (“red ball whodunnits” in Homicide terms). Perhaps the typical LAPD detective in 1947 was lax with evidence (although in LA Confidential and its sister novels, set in the 50s, the LAPD is surprisingly high-tech), but the character you’re playing is educated, ambitious, and a stickler for procedure.

This is, in essence, an even more annoying variation of the Fallout / Bioware “cursor as a white stick” mechanic for finding loot in 2d adventure games and I despise it, but it’s not so annoying that I won’t play the game.

You also interview / interrogate POIs (persons of interest). Usually you gather evidence first, but the game steers you towards interviewing a witness before — for example — gather evidence at another scene.

Once you start an interview, your character will say something (over which you have zero control) then the POI will say something, then you’ll look at your notebook and some questions will be there. You select a question and the subject responds, at which point you have to pick one of three options: Truth, Doubt, or Lie. You’re told that “Truth” means you believe them, “Doubt” means you think they’re lying, but have no specific evidence, and “Lie” means you think you’re lying and can provide evidence to demonstrate it. Before deciding you can examine their face and body-language (the facial expressions are pretty amazing, although almost everyone looks like they’re lying, and the few people with direct gazes are often lying in one or two of their responses and I for one have no clue how to tell which one).

Truth, Doubt, or Lie? Or would you like a life line?

You can examine your notes before choosing one of the three options, but once you pick an option you can’t go back. (E.g. you can’t decide you doubt them after having decided they were telling the truth, or change your mind and decide they were truthful after assuming they were lying. Even if you later find evidence contradicting a previous piece of testimony, you can’t go and put the question to the POI a second time. It’s quite infuriating. And you absolutely can’t do what detectives actually do, and that is ask open-ended questions. (If I recall correctly, in Homicide: A Year on the Killing Streets, David Simon says the most powerful question in a detective’s arsenal is “And…?”)

Sometimes a response will cause a follow-up question to appear.

When you’ve exhausted all your questions the interview is over and you’re told (usually) how many times you picked the “correct” response. You don’t know which specific responses were correct (which is especially infuriating when, for example, you’ve picked “lie” and gotten a highly unsatisfactory response despite having what looked like perfectly good evidence that you were being lied to, so were you wrong? Or was some other response wrong?)

Finally — worst of all — at the end of an investigation you don’t really get to pick your suspect (indeed, based on what I’ve read, you can apparently pretty much screw up every aspect of an investigation and the same person will be charged at the end). This may be necessary for some of the story arcs to work, but if so it’s a case of the need for story arc defeating the core gameplay.


"sadness" expression from Lie to Me

Perhaps the most egregious problem with the execution of all this comes to a single point I came across in this article on GameRant (which I found while googling for ideas on how to interrogate better, without resorting to a walkthrough):

A lot of people say that Aaron [Staton, who played Phelps] goes a little bit psycho with some of the questions you ask in the game. When we originally wrote it, the questions you asked were Coax, Force and Lie. So Force was a more aggressive answer, and that’s where we actually recorded it…But when the game came out, it was Truth, Doubt and Lie, so everyone says that Aaron on the second question goes psycho, but that’s just the way we wrote it from before.

This  is from Brendan McNamara (described as a former member of Team Bondi, the original development team of LA Noire).

Update: in the remastered version of LA Noire that came out for PS4 (and I’m assuming XBox One) Truth, Doubt, and Lie have become Good Cop, Bad Cop, and Accuse. The first two are better than Truth and Doubt (or Coax and Force) since they match the actors’ instructions, but Accuse is terrible because you’re often just challenging their testimony (see below), not necessarily accusing anyone of anything.

Note that Coax and Force are tactics, not assessments of the preceding response (and Lie would actually be better described as Challenge). Also note that the actors were quite possibly not told that they were lying, but that they were holding back. If you’re a fan of Lie to Me (or a social psychologist) you might know that the micro-expressions of a liar are very different from those of someone holding back. Good actors (and many of the actors in the game are excellent) will also know this. So when I take a glance up and to the right to indicate fabrication and pick “Doubt” I am picking the “Force” tactic instead. This is just awful, and combined with the terrible feedback system it’s near-fatal. Given that L.A. Noire places almost all its eggs in the “read the NPCs’ facial expressions” basket, this is an utter disaster.

Update: despite having played through the game before and solved every mystery, on my recent replay, despite studying the actors (at better resolution) I still do little better than chance in my interrogations.

Even so, the real problem with the mechanics is that it gives the player detailed control over stuff that often doesn’t matter (e.g. the order in which bits of evidence are discovered) but not over the actual outcome of most cases (correction: you do get to pick whom to frame for several cases where you know that you’re pursuing the wrong suspect). This is pretty much like those games where all the important stuff is in the cut scenes. “Hey player, go do a bunch of legwork so that the game engine can tell you whom to arrest.” Surely the basic task of a detective ought to be (a) gather evidence, (b) form a theory of the crime based on (a), (c) accuse your suspect. The way I would suggest it work is something like, every time you accuse a suspect with insufficient evidence your potential score drops (and possibly your ultimate score suffers if you successfully convict the wrong person).

Given the whole Dragnet/Untouchables visual feel of the game, having something like a voice description of the outcome: “Frederick Ronald Bloggs was convicted of first degree murder and sentenced to death in the gas chamber.” or “Frederick Ronald Bloggs pleaded to manslaughter in the second degree and with a good behavior bond.” depending on how well you conducted the investigation, along with a game breakdown of how well you did would be much better than telling you stuff like how many times you guessed which response to use over the course of the investigation.

I have left one thing out of my description of the game mechanics: “intuition”. This is basically a “lifeline” mechanism (you get a fixed number of intuition points — I’m not sure when your pool is restored, it’s possible that it’s simply buggy — and each point lets you “ask the audience” or “50/50” one response to one interview response: it’s a near useless bandaid on a broken mechanic).

Making Allowances

Now, I believe what happened with LA Noire is that Team Bondi worked on it for a long time and went way over-time and way over-budget. Rockstar assigned some kind of A-team to rescue the production and basically triage the content and get something out the door. (So, think So, a lot of the problems with the game are at least in part a consequence of an overly ambitious design being cut down to a (barely) functional core in sudden death overtime, but it seems to me that the overall structure of the game is actually as intended, and changing Truth, Doubt, Lie to Coax, Force, Challenge wouldn’t really help significantly because it’s a fundamentally flawed design.

The basic problem is the structure of the detective game.

(It’s pretty clear that the designers were aware of the shortcomings of the mystery content, because most of the mysteries and incidental street crimes devolve into gratuitous car chases and shootouts. So you get half-assed mysteries mixed with half-assed action. The worst examples are the linear chases with sudden death mini-games that are so horrible there’s a “skip” option for them.)

How to do it right

Here’s how the training program I mentioned earlier worked.

You (an investment banker) are asked to assess a company as an investment opportunity. You wander around gathering evidence (by walking around the company’s premises, making observations, interviewing its employees and customers, and reviewing documents) and then make a decision (“yes — loan money to the company” or “no — deny loan”). You need to cite evidence for and against whatever decision you make (compile a list of pros and cons) because your goal is to make a thorough assessment, including a  realistic risk assessment.

Is this not obviously a superior mechanic to the LA Noire game design I’ve described? You could even easily add mechanics for concealing evidence from the defense (i.e. concealing evidence that weakens your case), framing suspects, and so forth. Of course, that would be less Dragnet and more L.A. Confidential or Chinatown. I don’t want to go into the content of LA Noire — I haven’t finished it yet; so far it’s not terribly good — but it seems to me that L.A. Confidential was the obvious touchstone for this project. Unfortunately, the actual game seems instead to draw more from The Untouchables (both the movie and TV series) and Dragnet.

Update: despite all its flaws, I really ended up enjoying LA Noire because it’s pretty well written and very well acted, and I simply wanted to see how it turned out. As I write this, I’m replaying the remastered version.


There’s clearly a market for mystery-solving games: there are games based on all kinds of TV mystery shows — my wife bought me an unplayable Law & Order game for Christmas some years back; I would love a good game based on Law & Order. It’s sad to see Rockstar, which I believe has the all-round best adventure game engine on the market, try and conspicuously fail to tackle this genre, because I doubt there will be another serious foray into the genre for some time to come.