So I finally got around to playing A Study in Emerald and confirmed what the comments told me, that my previous post mentioning the game was wildly off the mark. That'll teach me to listen to rumours and heresay (Side thought: Is it an accident that heresay and herasy are so linguistically similar or do they share a root?). I did have a good time with it though, I'm not sure how replayable it is, but there's enough going on that being ruled over by great old ones was enjoyable as a first experience.
Our game had three players and very early on the other two effectively revealed their identities through their actions and subsequent discussion. As they were on opposite teams this provided me with a substantial advantage, they knew that they could hinder each other safely but neither could be sure that ruining my day wasn't hurting their own chance of winning. I found myself really invested in continuing to keep my identity a secret to preserve this advantage.
Contrast this to the traitor in Shadows over Camelot, in which an early reveal can serve the traitor more effectively than attempting a subtle game, to the point that the rulebook actively encourages traitors in small games to stay hidden in order to give the knights a chance. When a designer implements a secret identity mechanic they have an idea for how they hope players will use it, but ultimately the mechanics that underly the games reward functions determine what actually happens in play.
Any ideas concerning the best mechanics to get players to interact with secret identities in desirable ways need to be rooted in what a desirable interaction is. Ultimately secret identities introduce depth to a game without increasing complexity as they allow a great deal of other rules to be brought into a game from the outside world. Unless a player has lead an exceptionally sheltered life they will already know what it is to be lied to and what it is to lie to someone, they'll have hundreds of tiny lessons that they've picked up about both of these situations and so will every other player anymore. Furthermore these differerences will be asymmetrical, with different players having different ideas about the concept, drawn from their different experiences. So a desirable interaction is one in which players are motivated to utilise these skills to produce interesting effects.
Ideally, in a secret identities game, all players desire to know the other players identities and keep their own a secret. This is enchanced if there is a tension between the actions that a player needs to take in order to win and the actions that would best hide their identity. I think that seeking this enchancement is where some designs can fall apart, to keep the tension interesting some actions have to provide big bonuses while making it hard to maintain a deception - but if the bonus is bigger than the penalty for being caught there's no reason not to simply reveal ones identity. This needs to be considered across all bonuses, rather than on an individual level - for the tension to be constant the game needs to constantly offer "subtlety or power" choices and someone who gives up on subtlety can get the best out of all of those choices.
I think that Battlestar Galactica has enjoyed such success because it avoids this issue of cumulative benefits for being revealed. A choice often comes down to "do I risk revealing myself to sabotage this check?" and the consequence for being revealed is a dramatically reduced capacity to sabotage checks. The cumulative problem is avoided by having the large benefit for revealing apply only once at the point of reveal, rather than allowing a traitorous player to continue to accumulate benefits for revealing their treachary.
An alternative solution is seen in the likes of Discworld, in which a hidden identity has no impact at all until the game is over. The choice there is between 'proceed objectives' vs 'remain hidden', but on no occasion will anyone have proof positive of your identity before the end of the game. There's no explicit consequence to being identified, but it gets harder to win as players begin deliberately thwarting your objectives. Whether it's desirable to disengage from the hidden identity mechanic boils down to how effectively other players will be able to neutralise your ability to complete your objective.
Ultimately if a game designer wants to use hidden identities they need to examine their payoff mechanism. The trick is in balancing the tension between these things:
* If a player takes an action that risks revealing their identity then it should often lead to more progress than not taking that action (otherwise there is no reason for a person to risk exposing themselves).
* If a players identity is revealed it should set back their capacity to win by some amount (otherwise there is no reason for a person to try to hide themselves).
* The penalty for being revealed in the early or middle game needs to exceed the total bonuses for taking all "risks revealing" actions available (otherwise there's no reason to engage with the mechanic over openly revealing on turn one).
It's a tricky balance and some games manage it more successfully than others, as the reward functions that a game is imbued with also need to drive a host of other behaviours, but the sort of experiences that come out of seeing this mechanic executed well are irreplaceable.