Yes. That's not an argument against my point: they could achieve the same biased result using recordings of real bets. Be happy they (maybe) haven't yet, because that conartistry is much harder to show.
Don't trust anything anyone says about probabilities of real-world future events. We still can't even predict the weather a week in advance, despite huge computational resources being thrown at it.
