I found another possible bug.
The site says "But the final lucky number of each round is the one after 30-second is up" on the main page, and "The final lucky number is the one after 30-second is up" on the FAQ. But I just saw a game that ended less than 17 seconds after the first player bought a ticket.
Maybe it's just a bug in the text, and the 30 second rule is no longer in effect.
But either way you should fix it, and be explicit about which txid is actually used.
the FAQ says the hash which is used is the one after 30 seconds:

Bug reports are fake, i am not going to play your little troll games.
I see you fixed one of the bugs I pointed out, even though you said it was fake:

You've changed the 30 seconds to 10 seconds, matching how the game actually plays. Good job!
Unfortunately you've made the "provably fair" system even less provable than before. Now as well as your unverifiable transaction selection you have added in a millisecond timestamp.
You can easily adjust the timestamp by a few milliseconds in either direction to get the result you want.
This is the third different failed attempt to make the site provably fair. It's good that you're trying, I guess, but you really ought to take some advice from someone who knows what they are doing. You are only making it worse.
we have now added in a "Client Seed"
We believe we are now at 100%.
I would like to ask dooglus to comment on this new added system.
Where can I set the client seed? I haven't noticed that feature.
Even with a client seed, the game isn't provably fair if *you* get to pick the *last* piece of data that affects the result.
The usual flow for provably fair is:
1) site commits to a server seed by publishing its hash
2) player picks a client seed
3) game plays using the two seeds
4) site reveals server seed
Note that the site's influence on the result happens only in step 1, and the player's influence happens after that, in step 2. That prevents the site from cheating.
On your site the flow is:
1) players buy tickets (and possibly pick client seeds?)
2) site picks a txid
3) site picks a timestamp
4) result depends on 1, 2, and 3
The site's inputs happen after the player inputs and so can account for them.
Note that in a multiplayer game it isn't good enough even if the client seeds were selected between steps 3 and 4, because we have no way of knowing whether the house is also a player.
I would refer you back to the pevpot thread if you want to see an example of how to make a lottery game provably fair such that even a player who is working with the house is unable to cheat.