My claim is that you can't come up with such a conjecture where it makes sense to change the probability away from 1/6. That is why you should not update.
I disagree. I'm not sure it's provable(maybe in professional poker players??), but if you've played the bet a lot of times, you could have come up with cues* about whether your friend has got the same roll(or number on the die) as the last time or not.
Another month, another rationality quotes thread. The rules are: