"encounter a significantly worse drawdown in actual trading, then I know that there's something wrong somewhere"
That's one useful point about testing - that it can give you a feeling for when the trading system may be failing.
"by the way, that the long-side bias in equity trading still seems to show through...easy or even easier to accumulate profitable long-side day-trading testing results as short-side ones."
I think in daytrading one has a little more freedom to go long or short regardless of the longer term trends. In longer term trades, the main trend seems to be more important.
"Even using a very long time frame such as 20 years doesn't save you from such a problem: The last 20 years have been, after all, a major bull market in the S&Ps."
I see your point. In engineering, when one extrapolates or interpolates data, one uses a subset of data corresponding to the conditions of interest, e.g., if extrapolating vapor pressure to 350 Kelvin, one would use two or three points slightly above and below that temperature - not the entire set of available data covering a large range.
The last four years have given us an end of the main bull market, a transition, and a bear market. Those are very diverse market conditions. If a model can handle the last four years, I think it should do fine for the next year or so. There is no reason to require that the trading system do well in the whole 20 year period - just my opinion. |
