This could make a good KOTH.

If someone would be able to do that, it would be cool. I think there may be a bit to much randomness in the game for it be a true challenge though.

Can you clarify "Outputting 2 or 3 numbers will cause them to lose"? Is there any "look-ahead" strategy or is this simply "on this turn"?

@NotthatCharles Specifically on that turn, the bot will not output the numbers 9, 10, 11

So a bot can never output an 11 unless handed a 10?

Also, based on the examples, it looks like after a bot says 11, the next bot plays. Is that right?

That would be correct

One more question - are there output reqs around announcing the winner?

@NotthatCharles 1. Yes the next available bot will play 2. Didn't think of that I'll edit post I suppose it should be the same format as the examples

You said "random". What is the distribution?

Also -- the sandbox is slow.

@user202729 It was mostly conjecture. I always assumed there was to limited amount of control over the game to be able to develop strategies. As I was thinking it over today though, I believe there could be some interesting designs. So I think I'm going to try and get the KOTH to work.

@akozi Would that KOTH for n bots be averaged score over n! tournaments, or lots of smaller tournaments? (Since, for n>4 it is becomes possible for bots to lose on their first move, and I know that n=2 is a solved-game. Not sure about n>2 though?)

My first thought was to have have n! Games averaged. How dis you imagine the smaller games working?

Let's say we have 10 bots competing: n! games is 3628800 games. But, playing every order of every combination of 4 competitors is n!/(n-4!) matches (which are also each shorter!) for 5040 matches.

@Chronocidal I like that idea. I think the fewer players in each game the more control is granted to the player and allows for more strategy. I think that's a far better idea!

@Chronocidal I figure it would be information pushed to the terminal and people return the number they want to add. I think the infor passed would be the history of numbers added, players remaining, and the current count

I was thinking current count & players remaining, but full play history would let you look for patterns ("Player 3 always counts 2", etc) so is probably better. :) For single-language you can pass as an array, or for multi-language you could write to a text file and require a file or .bat that makes your bot run? (Like Let's Play Mafia! did)

@Chronocidal Awesome. I'm going to start working on this tonight.

I wrote a program to play every possible game. For two-player games, the first player wins once more. For 3-player, the first players wins significantly less than the other players. You can also print the game tree and change the bust number. bit.ly/2K1N74J

Very nice! I think I might steal some of this for the actual game as well. My code is a little... messy and this is very clean. I wasn't expecting that much of a difference. Interesting. With that in mind, how do you think scoring should work? In my mind the increases should become less between 1st and 2nd. Maybe something -> 6pts, 5pts, 3pts, 0pts. So more weight is on not dying first?

Note that this includes every game. Not just games where players play optimally. I was going to try to "solve" the game and see if there was a sequence of choices a player could make to win 100%, or what the best sequence is. I have yet to analyze the move tree to figure these things out.