This is not a duplicate (see below)
Before you answer “bad, the miner can influence them”, please let me specify the question.
It’s clear that things like timestamp should not flow into a random number computation because the miner mining the block can choose it quite freely.
But what about the blockhash?
The only kind of influence a miner can take here is to sacrifice the block reward and not publish the mined block (or am I missing something) - but in what situations does that give him any statistically relevant advantage?
I read somewhere else, that it would make sense to the miner to do so as soon as the possible reward is higher than the block reward, but that does not seem true, since sacrificing the block reward doesn’t automatically make him the winner, it just leaves his winning chances at what they statistically are. So it seems to only make sense when the block reward is lower than the average return expectency value of the random draw - or am I missing something?
So, how bad a source of entropy are future block hashes really, and in which scenarios are they actually fine?
And, important follow-up question: how is that assessment going to change once ethereum switches to a proof of stakes algorithm (this is not discussed in the question marked as possible duplicate)?