Off topic discussion from http://www.elitetrader.com/vb/showthread.php?t=283227 continued... There are two lively debates happening here. 1) Does HFT really affect daytraders PnL? 2) Is backtesting helpful for a new (unprofitable) daytrader? There's a lot to discuss on both points, which I will add my thoughts a bit later. In the meantime post your thoughts...
I will rephrase your second question. 2) Risking your hard-earned money on untested, unproven day-trading systems: pure insanity or a brilliant idea? By the way, how is testing a trading idea with historical data harmful in any way, like the title suggests?
My 2 cents: 1) of course it does by increasing slippage or the friction of taking or removing larger size positions. 2) backtesting for 100 years says nothing about the next week. It is nice to feel good about the past, un fortunately all my trades happen in the future. What's more every methodology generates 3 systems - system, opposite system and neither.
IMO, yes, it provides a methodological framework for evaluation which can lead to insight. I will also say that reliance upon a 100% systematic approach which yields successful results implies finding a systematic edge upon which only you can execute on - this is where I think most fail. This isn't attainable for the vast majority, furthermore if you have that, you probably don't need a backtest to validate that for you. Backtesting provides a frame of reference but can't provide an edge. Edge isn't found at inception and is only obvious in hindsight (for the most part), and even more rarely is it booked at inception.
A backtest reveals that system A has been returning 10% a year on average for the last 100 years, while system B lost 10% a year on average for the last 100 years. So you want us to believe that you would have no problem trading system B because the 100 year long backtest says nothing about the next week, is that it? There is NO past or future in backtesting! If you forward test from January 1, 2013 to January 31, 2013, the year 2013 simply becomes a backtest year in 2014. Forward testing is an illusion, the "fresh" data you are using are also an illusion because in 2014, data from 2013 are now part of the backtest and belong to the "past". Get it?
HFT does not affect a day trader. If you understand how order flow works and trading desks execute orders then you will always see past the BS and make money. Back testing does not work well because it does not take into account how bars are formed, the mood of the markets and the general mechanics of the markets. It does and can work but there is information missing. The best thing to do is let a strategy run on a demo account to see the results as they happen in different market conditions.
That's the other way around, a backtest simply reveals if your trading idea has a statistical edge, period. For example you believe that buying a stock when the RSI is under 30 ("oversold") or shorting a stock when the RSI is above 70 will make you money in the long run, so you backtest this idea with let's say 15 years of historical data. You run the test and in 20 seconds you discover that this is a losing system, so you dump the idea. Now get this: some "smart" traders here on ET say the exact opposite, they will try to convince you that backtesting is for losers and that we need to trade this idea immediately with real cash! Yep, risk your money NOW, and worry about the results LATER, that's their motto. And they dare to call themselves traders!
---HFT, and is backtesting helpful or harmful? --- i wouldn't call HFT. it should be named-is current market micro structure helpful in back testing or not? the answer will depend on what exactly you are backtesting. i have a system that i've been trading for years(on halt now). it is on halt for a following reason(s)-there is no way to predict what you are going to get(fill ratio). between backtesting(using tick data and small,fixed size per position) and real results actual difference is about 500$ for each day. yep. when it's +500 on paper-i got zero.when it's a zero on paper-i got -500$ on actual account. and i have filters in place,that only counts(as a fill) if there is at least 3 trades at 100 shares each-then i count that i got 100 shares. still-huge disappearances between real and paper results. all due a current microstructure in US stock market. take it from not some sort of academic\scholar,but from a market practitioner,who use to trade thousands shares a day. now sitting on paper
Entire books have been devoted to that subject (backtesting). And a serious backtest always takes into account commissions and slippage due to poor fills. But seriously guys, why do you keep scalping thinly traded markets with low liquidity/volume and then complain about poor fills? Stick to high volume/liquidity financial instruments like the Forex and you will immediately notice the difference.