Is there anyone with experience (and a robust trading system) that can tell me if the following example of a backtest - optimize- walk forward would be acceptable to use. This would be for a daily close trading strategy. Optimize system using prices (01/1998 - 12/2000). Backtest using optimized variables (01/2001 - present). In order to keep a system's parameters current for actual trading, is it best to optimize for the last 6 months of data and once every month?
Your settings certainly are acceptable. It also depends on the rationale of your system and the market it trades. I think instead of adjust parameter every 6 months or so, it makes more sense to consider seasonality. i.e., what parameters performed well in 1Q last year, what adjustments are needed for the summer dead zone last year etc. good luck.
I have quite a bit of experience with backtesting and system trading. I think your backtesting strategy is fine, assuming you have a sufficient number of trades in each timeframe to make the results statistically significant. For your three year backtest time frame, you will likely have several hundred or more trades. However, if you only have a dozen or so trades over this period then I would be very cautious about trading such as 'backtested' system in the future. Best of luck, -Eric