Based on how the new M3 model has performed over the past 10 days I've adjusted the momentum algorithm slightly to reflect a speed that is about halfway between the old M3 and the beta version new M3. I'm trying to avoid the typical problems associated with curve fitting an optimized model as maximum drawdown is just as important at total return as well as short term returns. We've run a series of forward tests, backward tests, out of sample tests and latency tests to derive the M3 version that is the current benchmark.
The z-score pairs signals have also been altered to show the dynamics of the actual SPY/UPRO pair with a 10 day fixed time stop. A full report on this pair will be issued as a separate post later this weekend and maintained in the Newsletter archives for reference. The results are quite consistent and provide an otherwise hidden perspective on the value of leveraged momentum.
