A conditional base rate answers a specific historical question: when the market has looked like this before, what happened next? It is the most defensible form of quantitative market analysis available, and it is routinely dressed up as something it is not.
What it is
A frequency measured over recorded cases. Of the times the configuration looked like this, the subsequent move exceeded some threshold in this fraction of cases. Nothing more.
The honest form carries the sample size, the window the cases were drawn from and the definition of the configuration. Without those it is an assertion with a number attached.
What it is not
It is not a forecast. A sixty per cent historical frequency does not mean there is a sixty per cent chance today, because today is one case and the conditions that generated the historical distribution may not hold.
It is also not an edge on its own. Knowing what usually follows tells you how to size and where to set expectations; it does not tell you which side of the distribution today lands on.
The sample size question
This is where most published base rates fall apart. A configuration specific enough to be interesting is rare enough to have few historical cases. Twelve cases produce a frequency that will move ten percentage points if one more arrives.
Always ask how many cases. A rate without a count is not a statistic.
The regime question
Cases drawn from one market regime may not describe another. A base rate computed over a period with different volatility, different participants or different market structure is describing a market that no longer exists.
The defence is a long window covering several regimes, and reporting whether the effect holds within each rather than only in aggregate.