Although this is a common criticism of utilitarianism in the Bentham traditition, it is not in fact how utilitarian ethics work as a whole. For example John Stuart Mill would say that whether or not an act in a specific case is ethical can be decided by whether or not it is in violation of a general rule, and whether or not the consequences of the general rule are positive or not. Taking clients funds and using them to go double or nothing would clearly lead to lack of trustworthiness and the consequences of this would be clearly negative in general so this is very clearly unethical according to Mill.[1]
> To maximize your expected value, you must aim for it and then march blindly forth, acting as if the fabulously lucky SBF of the future can reach into the other, parallel, universes and compensate the failson SBFs for their losses. It sounds crazy, or perhaps even selfish—but it’s not. It’s math. It follows from the principle of risk-neutrality. [0]
I think he figured in a million realities, the expected value is net very large. He just happens to live in a reality where it collapsed, but how much of his wealth is due to these games? I don't know if the probabilistic method is common among EA. For instance, if you gave me a chance to bet my entire net worth on a 51% odds game, sure the expected return is positive and if i had a million realities and netted everything out, I should take the bet. But most people would see that as insane
When the ends justify the means, anything goes.