I Think I Found the Holy Grail

C

Corso482

Ok, the reason I'm posting this is because I'm new to system development, so I don't even know what is good or what means anything.

So I've been tinkering with system development for a while now, and I've come up with a system that returns about 200% a year on stocks, no margin, 1% of total capital risked per trade and about 5% max dd. The system is only about 6 lines of code, so it's extremely simple. It's got about 50% more winners than losers, and avg. winner is about 5X the avg. loser. I've backtested it back to 1997. The system trades a 5 min chart.

The above results seem too good to be true. Have I found something spectacular? Or do most system developers have systems like this hidden in their drawers? Do these results have any chance of working in the real world?

BTW, the reason the stats are so rough is because stockwatch pro doesn't give you many stats on the results page and I don't have excel to do all the calculations. So for max dd, I just eyed up the results manually and did what I could on Microsoft spreadsheet.
 
Quote from aphexcoil:

Commission / Slippage accountability for the system?
More...

what's the standard practice? how much money are you supposed to start out with in testing?
 
Corso482,

Try trading the system with real dollars for a day or a week or so. At the end of each day, run that day's data through the backtester to see what papertrades the system was supposed to make. Chances are, a combination of two outcomes will occur:

1) that day's papertrades will be unprofitable because your backtesting was over-optimized (i.e. you found a system that worked very well in the past, but no longer works in the present or future)

2) that day's real trades will be less profitable than the papertrades due to slippage, commissions, and system delays. (i.e., the backtester contains an unrealistic model for how real markets and trading works)

Just don't jump to conclusions about the performance until you have at least 20-30 trades (and preferably 100-200 trades) in the test. If the first day goes badly, then paper trade it for a while. But ultimately, you will want to do real trades and see if they bear any resemblance to the backtested paper trades.

Finally, even if the real trades are less profitable than the paper trades, you might still have a system that is good enough.

Best wishes for drinking from your grail-like system,
Traden4Alpha
 
Quote from Gordon Gekko:

sooooo, tell us about it. :p
More...

well, it's long only, buys at the open of the next 5 min bar after certain criteria are met. It exits based on opposite of entry criteria, 1% stop loss, or the end of the day. It's ridiculously simple. Makes 4-5 trades a day. Highest % returns come from low priced stocks. Avg profit is about .60% and avg. loss is .11%, but varies depending on what stock you're trading, what year you traded. I don't know what the hell you'd call it, it's not trend following, maybe volatility?
 
How many trades are in your backtest sample and how do you pick the stocks to trade?

If you have hundreds of trades in your backtest sample you can have more confidence in it than if you have dozens.

And you should pick what stocks you backtest on and will trade on a priori (without hindsight). Some people fool themselves by throwing out stocks where their system doesn't work very well for no reason other than that their system hasn't done very well on them. (See the "Don's Opening Only" threads for examples of this.)

I agree with Traden4Alpha...trading the system is the best way to find out if you've got a good one. Let us know if you do.

Definitely think long and hard about slippage and transaction costs. Inandlong will probably chime in with some good advice about how to account for this.
 
You must log in or register to reply here.
Back
Top