You are right, it's SHA1, even though it is to some extent predictable, doing so is still unfeasible to do so in this case.
To read more about why SHA1 is usually bad: http://www.schneier.com/blog/archives/2005/02/sha1_broken.html
Also currently sampling the data for you.
EDIT:
2827 bets, semi-colon separated. For all bets I picked >50.50 as winning parameter and the client seed was set to 0.
First column: WIN/LOSS
Second column: The roll-outcome
Third column: Server seed used for calculation
http://pastebin.com/1kvzeGtZ
Average roll number: 50.04464803678808