Quote from black diamond:
So here are the model results I said I would post last week. These results are for 1000 consecutive holding periods. Not fully realistic (no slippage, costs), just some model backtests to make the point so take it for what its worth.
The basic strategy ranks all stocks from 1-10 using signal 1 in every period and goes long the 1's and short the 10s in the next period, equal dollar weights and always in. I'm not saying what the signal or period length are, but the signal is price based and is a variation of a reversal strategy.
The vol filter takes a ratio of each stock's dollar volume in the ranking period to a historical average for that stock and only takes positions in stocks with the ratio in the lowest 10% (cross-sectionally in that period). I use the ranking instead of an absolute value to make it relative to the market, so I am always taking the "best" 10% of all the positions that signal 1 says to take.
Maybe not earth shattering and needs work before being tradeable, but it looks like the vol filter makes a difference.