Distribution Clustering

We have probably all figured out that a bunch of distribution over a short period is worse than the same number of distribution days spread out over 25 days.
So what is a bunch and what is a short period?

I have built a hypothetical sell rule: S13 Distribution Cluster.
I used the Market School Exposure Model to test the results and offer a clue to the answers to the two questions.

In the MEM I run a model portfolio that invests in the NASDAQ at the recommended exposure produced by the model. Using the final portfolio value(after 40 years of buying and selling the NASDAQ) as a measure of goodness I find that 4-distribution events over an 8-day period provides the best performance than any other combination of look back and cluster size. My gut told me that 3 out of 5 would work well, it does but not as good as 4 out of 8.

A distribution event is a standard distribution day (close down 0.2% or more on higer volume than the day before)or a countable stall day. A distribution event may only occur on a day when we have a distribution day or stall day (prevents some double counting). Market School will count the first stall day and thereafter count a stall day only if the number of real distribution days is larger than the number of included stall days. I tested other versions of this rule before arriving at this definition.

So for intermediate term investors, the way one could use this information is to track distribution events as above and when you see 4 out of 8 it is time to lighten up the portfolio. S13 very seldom occurs in a strong uptrend but often does occur when the uptrends gets tired. Most S13 events happen in a bear market. The last uptrend showed multiple S13’s before the market went into correction and would have caused a cleaner exit.