About the crisis volatility filter

I tried the idea suggested here. Indeed, if my “volatility” crosses a certain border from bottom to top, then it is possible to cut off unprofitable periods in RI and stocks, including going long on February 22, 2021 (exiting the “filter” not by the volatility value, but by its dynamics from the first point). However, there were not enough periods (from the end of the day of the first date to the end of the day of the second we do not trade long, and we enter at the end of the second day if we entered the long earlier) for an unambiguous conclusion:

05/26/2006-05/29/2006 (1 trading day)
08.09.2008- 03.10.2008
09/27/2011-09/28/2011 (1 trading day)
03/03/2014 (alas, Crimea is not cut off, but I was not long on the systems before March 3, only the sold and unhedged 120th put at par (!, not GO) of the futures at 100% NAV and a partial short in Gazprom) -11.03.2014
12/16/2014-12/30/2014
01/22/2016-01/26/2016 (1 trading day)
03/10/2020-03/19/2020
02/21/2022-04/05/2022 (from 02/28 to 03/23 there were no trading, and the dynamics is calculated only on trading days)
09/21/2022-09/22/2022 (1 trading day)

Actually everything. Out of 10 cases, 4 include only 1 trading day each and obviously the results there are pure coincidence. It turns out 6 “successes” out of 6. It is clear that the probability of “success” is greater than 1/2, but it is too early to calculate it as 1.