There's only one user whose seeds I did randomize before launching for CLAM, and that's because I told him how long his current losing streak would have ended up being. He could have used that information to win huge bets, and so I reset his seeds.
You're actually able to see this? I thought it was all randomized.
It's like this:
1) the server picks a random "server seed" for your account
2) the server tells you the hash of the server seed, and you make a note of it
3) you pick a "client seed"
4) you make a bunch of bets - each roll is the result of hashing the two seeds (server and client) together with a nonce which is 1 for the first bet you make, and goes up by 1 for each subsequent bet you make
5) you click 'randomize'; the server publishes your server seed
6) go to 1
a) You then compare the hash of the server seed from (5) with the hash given in (2) to make sure that the seed you were given after you played matches the hash you were given before you played. That proves to you that the site committed to a particular server seed before you picked your client seed.
b) Then you use the server seed (5) and the client seed (3) to regenerate all the rolls you made. You can compare them with the rolls the site actually gave you.
Since the site provably picked the server seed before you played and before you told it your client seed (a), and since the site used those two seeds to generate all your rolls (b), there's no way the site could have cheated you.
So in a sense the rolls are random, since they depend on the server seed, which is random.
But in another sense the rolls are predetermined once you've picked your client seed, since the pair of seeds uniquely and completely determines all your future rolls.
Since you pick the client seed *after* we pick the server seed, we can't pick what numbers you are going to roll. The client seed has the final say on that, and that's entirely your decision.
If each roll was truly random, there would be no way for you to check whether the site was cheating. You play "less than 98", we pick "99". You lose. Did we cheat? We went to random.org and it told us 99. Honest! There's no way you can be sure we went to random.org and didn't just pick a number bigger than 98 to make you lose. So that's why we do it how we do - so that there's no way we can cheat the players.
On the other hand, we *can* cheat the investors. I can see what my next roll is going to be (it's 98.3610), and so I can be sure of winning it if I want to. That's the same with any provably fair site - you have to trust that the operator isn't going to play with knowledge of his upcoming rolls.
Here are the first 3 rolls I made with my current seed pair:
![](https://ip.bitcointalk.org/?u=https%3A%2F%2Fi.imgur.com%2FEiOeJP3.png&t=665&c=R7f-ANV0DqU9kA)
and here's my "prediction" of the 4th roll:
![](https://ip.bitcointalk.org/?u=https%3A%2F%2Fi.imgur.com%2FAjOAxHq.png&t=665&c=-X-NznBrGFI7WQ)
This is why, if I'm going to play against the investors for more than small amounts, I go to great lengths to prove that it's legit. For example:
http://www.reddit.com/r/justdice/comments/1vbo1p/justdice_owner_plays_his_own_game_and_wins/Edit: I made that 4th roll:
![](https://ip.bitcointalk.org/?u=https%3A%2F%2Fi.imgur.com%2Fu1IdZeq.png&t=665&c=grxzZ0CgBMvTJQ)