Stacking the deck

So I was reading this website, it had an article talking about how an on-line casino is publishing the mathematical "hash" of the virtual deck of cards before the game but without revealing everyone's state of play.

An anonymous commenter wrote...

"The integrity of the deck doesn't matter worth *anything*. The hash insures (or is supposed to) that they haven't changed the cards middeck. There is NO protection against stacking the deck."

Now this got me thinking, and I think I've worked out a way to have a dealer which everyone could be sure is fair. I stress that I think I've worked out a way. If I've made an error, (or dare I say, you think it's perfect) please leave a comment and tell me.

A deck of 52 cards has 52! (or approx 8×10⁶⁷) possible shuffles. This is comfortably within the realm of 256 bits. The dealer's shuffler code will need 256 bits to perform its task.

1. Dealer selects 256 bits, his contribution to the shuffle function, k0.
2. Dealer selects another 256 bits for the secret salt value (salt).
3. Dealer calculates the dealer's hash. SHA256(k0 + salt) and transmits it to all players.
4. Each player selects their own contribution to the shuffle function, kn.
5. Each player calculates and publishes the result of SHA256(kn).
6. Once everyone has the player's hashes from step 5, each players publishes their kn value from step 4.
7. Each player and dealer may audit the other player's hash by repeating the SHA256 calculation.
   
8. Dealer calculates SHA256(k0 + k1 ... kn + salt) and shuffles the virtual deck with the result. (The shuffle function should be published in advance and be deterministic only on the seed value.)
   
9. Dealer and players play with shuffled deck until the winner is decided.
10. Once the game is over, dealer publishes k0 and salt, allowing everyone to audit the dealer's actions.
    

Note that player's kn values can be any length, within reason, depending on how difficult that player wants to make it for others to cheat. If a player trusts all other parties, that player could just use no bits at all as thier kn value.

Here's my own attempt at attacking this algorithm. Stop me if I've missed anything.

• Neither dealer nor players can "stack the deck" by choosing a favourable kn, because all parties contribute into the final hash that seeds the shuffler.
• The dealer can't choose a favourable k0 or salt after the each player's kn has been revealed, as the dealer will already have revealed the dealer's hash which the player's can check afterwards.
  
• The last player in line can't choose favourable kn material with knowledge of all the other k inputs - even if he colludes with the dealer - as the other players don't have to reveal their actual kn value until all the players have revealed their hash.
• Players can't change thier chosen kn value after step 5, as they will have already revealed the SHA256 of their kn value and the other players can check it.
  
• Players can't collude against the dealer or other players, as salt will not be revealed until the game is over.
  
• The dealer can't use different shuffled deck, as all of the dealer's actions may be repeated once the game is over and all hidden values are revelead.
  

Any comments please?

Its possible I'm talking rubbish. Its also possible I'm the millionth person to think of this.

UPDATE: Looking over it again, I don't think the salt value adds anything and I can't remember why I added it in the first place. Without the salt, the algorithm is the same for dealers and players, except the dealer's reveal of k0 comes after the game has completed.