Contra-momentum strategies are usually called contrarian as they are the opposite to momentum strategies. Back tests easily reveal that contra-momentum strategies are much more effective than momentum strategies as was shown a couple of years ago at Zacks’ People & Picks. Momentum strategies are usually not long-term effective. They don’t stand the Tooth of Time with only a few exceptions to the rule like SOW’s Big Money strategy. Contra-momentum strategies are usually robust over time when you select the Bot# stocks that dipped the most over the past holding period. It is possible to construct a contrarian-momentum strategy that is robust over 30+ years of back testing and not using any other quantities than closing prices and trading volumes. However you need better data then dbcmhist to prove it. Dbcmhist doesn’t have the historical prices but uses i5 as the closing price adjusted for splits.

That makes all back tests based on i5 unreliable in RW. Zacks states that limitation on one of its sites. From a sw programming point of view it is relatively easy to program stop-losses and stop-gains into a back-test algorithm. You then could extend holding periods to, say, 13 weeks and run the back tests with such stops. The stops encompass then all daily closing prices during the holding periods. It can then easily be shown that when you would allow for multiple passes through the stop-losses, your gains would be fantastic when you don’t account for the slippages encountered when whipsawing through the stop-losses, especially when you would combine the stop-losses with stop-gains. However, when you would account for the real-life whipsawing slippages with multi-passing stop-losses, your stop-losses always appear to degrade performance significantly. As a rule of thumb, whipsawing through stop-losses on average gives 40% of the stocks in your portfolios an extra pass. If you make the stop-loss only effective during its first pass, it also appears to degrade performance in all screens I tested. So I never use them. Stop-gains usually mildly improve performance, something I would expect.