Here are some initial thoughts. I haven't finished working through it, so more to come:
Perhaps you should specify if the probability of detecting another player is an overall probability or on a per undetected player basis (so when there is only one player they haven't detected yet, when they detect a player it will always be that player they haven't detected)
in the definition of L_K, why is P(K survives turn i) outside the summation? What does i refer to then? Is P(K survives turn i) a constant ? Wont it in general depend on the current state of play?
Thanks for looking at it.
Probability is on a per player basis (ie: each turn, a player has a chance p for detecting each undetected player). I'll edit this so it's more clear.
For L_K (as well as L_R and L_P), the term outside the summation is essentially (probability the player survives the whole game) (game length). It's necessary since the game is of fixed length, and the summation is adding (probability of dying on turn x) (turn x). Consider if the probability of detection is zero, and players will never die - without the term outside the summation, the expected lifetime calculation will return zero.
Update: Please use the most recent thread.
The LW Public Goods Team wants to encourage useful research projects (as well other kinds of projects) for the LW community. If you're interested in doing this kind of work, you might run into a problem that is best solved by good outside assistance. Without assistance you might get discouraged and stop working on the project or never even start it. We want to help you avoid that. Do you
Then, we want to help!
How do you request such help? For now, I think the best way is to post to the discussion section about your problem. That way other interested people can also provide help and be interested in your research. If you feel uncomfortable doing this, you may post to the public goods team mailing list (lw-public-goods-team@googlegroups.com) or if it's not too long after this was posted, post in the comments.
I personally commit to doing at least 3 hours a week of tasks like these for people doing LessWrong related projects (assuming demand for it; I'll be keeping a log) for at least the next month. Morendil has committed to doing at least an hour of this and atucker has promised to some as well.
Our goal is to find out whether this kind of help is effective and encourages people. If this kind of assistance turns out to be valuable, we'll continue to offer it.
If you would like to volunteer some time (a little or a lot), say so in the comments!