To be fair:
You shouldn't rely on the randomness of the hash if people are betting on it. Especially not after the hashing.
If you have a miner who stands to make more on the bet than they would on the block subsidy, the economics can favor them getting a hash low enough to make a block, but then withholding it, looking for a hash low enough to make the block *AND* having a final digit that will win them the bet as well. They risk losing the block subsidy if another miner finds a block first, but they can also instantly release the block they found the first instant they get a whiff of the other miner's block, and then they have a nearly-equal chance of getting the block subsidy anyhow. Meanwhile, they go on looking to win the bet.
This.
When there is no incentive to mess with the hash, it may be safe to rely on the last digit. But as was said here (and DH above previously) when there is incentive to do so, it is possible to do so.
Using the last digit of the hash of a single block as the sole random value for a sufficiently large prize is asking for problems.
Even looking at all the values showing it looks random is not sufficient if the lottery drawing (for example) happens every 1000 blocks because you only need to do it once to win a massive sum (eg in the US's Powerball drawing the subsidy is minimal in comparison). In most of the blocks there is no incentive to select a particular hash so even if someone did it successfully 10% of the time (every 10000 blocks in this example) it would be undetectable in the noise.