Don't rely on strategies optimized on historical data. Quantum Funds of George Soros are being talked about as upcoming cause of further market decline after sub-prime. These quantum funds (huge) have invested only on basis of computer modelling of strategies based on historical data. And these models are failing now. (May be then, the Wolf wave on SENSEX posted in this forum will meet its target)
Our markets are highly manipulated ( see GMRINFRA on Tuesday). Try FOREX. Past few days have been money making days in FOREX on all four majors, even a simple trend following system is making money. I've found most of the technical indicators work on four majors in FOREX market.
hmm...where to start
Are you suggesting to the OP that backtesting doesnt work.Optimization is a completely different topic than backtesting.Even there, someone who optimizes on In-Sample Data then runs it on Out of Sample data and again LIVE for a few days is in a much better situation than someone who uses RSI(14) because its the 'DEFAULT'.How does one know if the Input 14 is not curve-fitted in the first place?
Backtesting just answers the question whether an approach has or hasn't worked in the past. It makes no guarantee that the strategy will work in the future if market conditions change.Having said that, about the current 'conditions', its nothing new, the patterns have existed before and will continue to do so.
Also about funds losing money, if you analyze the COT data for US markets, the big guys were short long before the current clustering effect took place.Its a Zero Sum game remember, someone always benefits.