There is a reason that random strategies don't make money - it's simply because they can't. When I say they can't - I mean it is against the laws of mathematics and in a way physics.
I shall explain.
There is, you will mostly all agree, a finite amount of money in the markets.
That amount is the amount of money that all players have put in over time, less the operating costs of the markets and less what they have taken out.
It is not possible for all players to take more money out of the markets than they put in. Just as it's not possible to take more water out of a bucket than is in the bucket. To take money out it needs to come from other people and your ability to extract money from the market whilst others lose it is an 'edge'.
Let's say we had a random strategy that made money. Then lets say that every investor used that same strategy. It is physically impossible for all players to take more money out as a whole than they all put in, therefore something would have to give.
If everyone implemented this 'random' money-making strategy, the markets would equalize and change in such a way that the strategy would no longer work. The edge would disappear.
If we can accept that the markets can change in such a way that a strategy stops working then we must acknowledge that the strategy itself is exploiting something within the markets at that time. The edge itself is tied to some expectation of behaviour in the market, in this case, a specific volatility profile. When that behaviour is not present - they system will not make money.
In the Van Tharp book, no-one claims to have executed this strategy as a coin toss. They discuss only testing it. The people running the tests thought they were executing a test in randomness when in reality they created a volatility specific strategy. It is understandable that they didn't see this, although a couple of the other things they mention border on cheating.
In particular, they got 80% results and added something that took the results to 100% across the same data set. This is curve fitting. One can only wonder how many things they attempted in their efforts to get from 80% to 100% for that data set. I am sure people here have done similar things - I know I have.
Note also - they didn't discuss the size of the 80% of winning runs to the 20% of losing runs, nor did they discuss overall effect of adding the rule that bumped the ratio to 100%. I presume they must have run out of ink at that point.