once you add a second observation for the same pod, long-term predictions may actually get less accurate. That’s because as soon as there are two observations for one pod, the difference between those (or a divisor of it if they are multiple pod days apart) is used as the estimate for the day length (portal period), replacing the built-in estimate. The built-in estimate for the day length is pretty accurate, because otherwise predictions would run out of sync with reality too quickly. The difference between any two portals observations is likely to be less accurate than that if you didn’t measure them to a fraction of a second, and it takes many observations (or few observations but a long time apart) to average out the errors.