[indrax]

You have to weight the validity of the reasoning across all turtle observers through history. If you reject probabilistic reasoning because it sometimes tell you things that are wrong, then you have no way to reason about uncertainty.

[addingnumbers]

If your reasoning about uncertainty tells you there's a 99% chance something is false while it's true, what good is it? All the math in the world won't save you from the specious premises in this no-seafarers argument.

[epivosism]

"If your reasoning suggests you shouldn't play the lottery, but playing the lottery actually results in you winning, what use is your math?"

Well, what matters is expectation. The fact that some people win the lottery doesn't prove that playing the lotto is profitable. In fact it isn't money-profitable. The fact that people sometimes win doesn't change the nature of reality, that even for them, the win is unexpected and playing was negative expectation for them.

You seem to be confusing expected value with results.

Do you revise your decision process every time you get unlucky? On a meta level, sure, paying attention to unexpected failures is important to look for systematic errors. But randomly getting unlucky on 99% certainties does not automatically disprove your decision method.