Good point, to be honest. Maybe there's a better way to combine the block hashes like taking the first N bits from each block hash to produce a single number but the direction that this is going in doesn't seem ideal.

I just asked a friend about this problem and he mentioned using the hash of the proof of work hash as part of the number so you have to throw away a valid POW if it doesn't give you the hash you want. I suppose its possible to make it infinitely expensive to manipulate the number but I can't think of anything better than that for now.

I need to sleep on this for now but let me know if anyone has any better ideas.

On Fri, May 20, 2016 at 6:34 AM, Johnson Lau <jl2012@xbt.hk> wrote:

Using the hash of multiple blocks does not make it any safer. The miner of the last block always determines the results, by knowing the hashes of all previous blocks.

== Security
Pay-to-script-hash can be used to protect the details of contracts that use OP_PRANDOM from the prying eyes of miners. However, since there is also a non-zero risk that a participant in a contract may attempt to bribe a miner the inclusion of multiple block hashes as a source of randomness is a must. Every miner would effectively need to be bribed to ensure control over the results of the random numbers, which is already very unlikely. The risk approaches zero as N goes up.