My desk had 40+ traders. Some of which have been trading for 8-10 years and were very profitable traders. All gone now except for 2 guys. Kaput. The financial crisis and the subsequent market environment pretty much carried all of them out. 1 of them whom I know still daytrading is very versatile and I suspect that's the key to his survival, he could trade anything under the sun, super liquid, illiquid, tight spreads, wide spreads, everything. If it moved, he traded it, whether it was SPY or STRA and 8 out of 10 times, he did it profitably. How many traders are truly like that though? It's very rare. Longevity is a big problem in this business. The opportunity costs given up are great. In the long run, someone who has a job and career would still come out on top.
Excellent example on how extremely poor moving averages are for making predictions - yet it's mind blowing how CNBC pundits still refer to 50- an 200-day MAs - I'd bet they've lost so much money that their only income today is to sell poor advice to others.
Good evening QuantWizard. Please note that I did NOT say that moving average cross-over systems are useless in general, far from it. All I am saying is that a trader should never take ANY trading idea for granted, no matter how "brilliant" it may sound. Test it first with 15 or 20 years of historical data and THEN trade it with real money, IF and only IF it is profitable AND the maximum drawdown is reasonable (less than 20% or so).
A lot of people do forward testing, they play a strategy with small size in real-time until they are confident it works. Would backtesting be more prudent because it wouldn't risk any capital? It could be but there are a lot of issues with slippage that you can't know about until you take the strategy live, specially in a HFT era where the bots will just back out of orders. Most traders I know also do a "mental" backtest, they have seen so many situations, charts, price action reactions that everything they do have has a backtest type evidence behind it, its not a formal backtest but the effect is the same This is not to say the formal backtesting is bad, its just that a lot of the time its simply complementary or not necessary