A conditional base rate answers a narrow question: given a dealer state with these properties, what has happened next in the sessions we have logged? It is arithmetic over history, not a forecast, and the distinction matters more than it sounds.
A forecast says what will happen. A base rate says what did happen, in a defined set of past cases, under a condition you can state. You can check a base rate. You cannot check a forecast until afterwards.
Sample size is part of the number
A rate without its denominator is not information. Seventy percent of ten is a shrug; seventy percent of four hundred is worth reading. Any base rate quoted without the count behind it should be treated as decoration.
The same applies to the window. A percentile measured over two years and one measured over thirty are different claims, and the only honest way to quote either is with the window named.
Why our own history is the hard part
Anyone can buy an option chain. The chain is a commodity and several vendors sell the same one. What is not purchasable is what a particular dealer map said at a particular minute on a particular past day, because that is a computed state somebody had to record at the time.
That is the whole reason a conditional base rate is difficult to replicate. It is not the maths; it is the archive.
How to read one without fooling yourself
Treat it as a prior, not a signal. It tells you whether the setup in front of you is ordinary or unusual, which is a different question from what to do about it. Conditions that look similar are not identical, and a base rate cannot know what is different about today.
And watch for the setup where every case in the sample came from one stretch of market. A hundred observations that are really one week repeated is not a hundred observations.