This is one of the more challenging aspects of modern risk assessment. One of the keys is to select the future date for which the PoF is being estimated. This is often ‘next year’ but can really be 5 years, 10 years, 20 years, etc into the future.
Another key is to ensure that both likely and rare-but-possible scenarios are captured in the conversion. A recommendation is to set up your model so that, for instance, a 10% chance of a 50mpy degradation rate can be modeled to be equivalent to a 5 mpy degradation rate. This is readily done, but you must pay attention to the downstream processes to ensure that this equivalency is legitimate.
There are several important nuances to this issue. Please consult other references on this site or the book for details.