Secret from combined transaction ids

[dipsy]

We've been working on some game ideas where different people play against each other.

I'd just like to check that a secret for a provably fair game can be made out of 2 transaction ids; where the secret is used to create a signature with those tx ids (i.e. satoshi dice)

Let's say that there were 2 transactions in a 2 player game:

player_1 = "d200c3d88a578e32c6d024aacdac11b8f790c34679e60a114b4af9bcc07b39ad"
player_2 = "6bfafd42a665e14b3aa2f107e9c09164697bfe6f5dfb48ba163198a749fca3cb"

Apply a 512 Digest to the combined hash of the 2 players

secret = Digest::SHA512.hexdigest("d200c3d88a578e32c6d024aacdac11b8f790c34679e60a114b4af9bcc07b39ad6bfafd42a665e14 b3aa2f107e9c09164697bfe6f5dfb48ba163198a749fca3cb"))

secret == "0775c9758a8d63f66098e461bb185f7adc1c080f7d3f0656e3d968cdc45de4512abfff756eea6b3 379a8eb7eae110d3f95363a1c570fcbea6a2a48eaca5e0282"

Things to bear in mind:
- Players play against each other - 2 players could gang up and try to match their tx ids together in order to screw other players, but they don't know the tx ids of those other players.
- Because people are playing against each other, can any one (or many) users exploit the game?

Thanks for your help 

[1538089369]

1538089369

[piotr_n]

But what will be the winning criteria?
If you make it like "whichever hash is lower, then it's quite exploitable, no matter a method".
Bit if you make it smart, like e.g. satoshidice, then it can be pretty secured and fully auditable.
Its usually enough to just xor two hashes together to get a unique enough value - unless you want to keep the combining algo secret.
But people who gamble don't like secrets, not that I'd known any

[kjj]

In a A-vs-B game where the criteria is a hash of the combined transactions, it should be perfectly fair.  Just don't reveal either txid until you have both in hand (reveal a [randomly] salted hash if you want to prove that one party or the other has made their move).

[piotr_n]

hash of the combined transactions - indeed. and good solution, btw.
but I bet, there are so many ways to build a story around it, how to interpret the winning hash...
and some of the stories might be exploitable, if you don't think them through enough.
or I should say: if they don't think..

[dipsy]

Thanks for the help guys.

The idea is to create a HMAC SHA 512 signature for each player from the secret (combined transaction hash) and the original tx id. We'll use the last few hex of the signature to generate a number.

I'll probably post again when we have the app live, it's important for everyone that it's provably fair. Hopefully somebody won't be able to game the story though!