amibroker
You are using an out of date browser. It may not display this or other websites correctly.
You should upgrade or use an alternative browser.
You should upgrade or use an alternative browser.
my mechanical system
- Thread starter robottrader
- Start date
amibroker
Quote from man:
this is sometimes called survivorship bias and it is a true
nightmare. if i recall correctly, including all ticker changes
the SP1500 had 1100 changes in the nineties. these
things are massive problems for backtesting long only
single stock systems. they can easily wipe out the edge
or, seeing it differently, a system can be fit to this very
phenomenon that a database does not contain changes.
i would consider trading a short component against the
long side. even if is on its own not worth doing it could
enhance the robustness big time. in fact most long short
equity players are more or less breakeven on the shorts,
which just balance out the overall market risk.
I actually thought about trading a similar short system. But the challenge is you never know if a stock is shortable. And, I've found it doesn't help lower the max DD.
The only thing I've found that I trust on EOD data is the closing price. That's the only thing you can trust (1) is accurate and (2) that you can actually hit. Having said that, since you have a great backtest and your paper trading is matching the backtested results, I would say just dive in and devote a small account to actually trying out the strategy realtime. But, it's important to compare your actual entries/exits with your parallel theoretical entries/exits to see if real life can truly match theory. If your actual results are significantly less than theoretical, then you can reevaluate your strategy or your execution. If, on the other hand, your strategy succeeds, you're profitable, and your real results don't deviate too far from theoretical, then increase your bets weekly based on how much total profits you've accumulated. If the equity curve is as smooth as the graphic you posted, then you should be making $100k bets in no time. Devoting 20% of your account to each bet sounds a little on the risky side, but the smoothness of your equity curve should dictate how much capital you should commit to each bet.Quote from lindq:
You haven't addressed the questions asked of you, but I am going to assume that your system (1) is based on EOD data, and (2) buys pullbacks.
If this is the case, then I hate to dash your hopes, but you may have a big problem with data.
In my experience with EOD data, which is pretty extensive, the daily lows are sometimes NOT what was actually printed intraday. Bad prints are logged as the daily lows, when indeed they were not. So your system would show a profitable trade, when a trade would not have occured at all.
This happens often enough that it can really skew your results over a periods of months/years.
So going forward in trading, be careful and watch for daily misprints.
I don't go above 10% of my account on any single stock, but I also will get into upwards of 30 stocks at once, so I can still easily max out my overnight margin and, consequently, my return on capital. I initially tested my own system realtime with $2k bets. Within 2 yrs I had increased my bets to $100k and still manage to match the theoretical results. I've kept that amount level for about 6 mos, but now I'm considering going up to $200k/bet for the more liquid stocks.
Good luck!