Misleading results in back tests
It's a broad question but more than likely your system takes advantage of a trending market or a ranging market, and once the behavior of the market changes to one or the other, the EA falls on it's ear. Find out which it is taking advantage of, up-down-sideways and try to figure out how to filter the rest of the stuff. Also if you are not using a stop loss that could be a huge issue.
Here is what I do know about the nature of the data.. prior to 19-SEP-06, the mean daily ticks/volume are higher, and the volatility (standard deviation) of ticks/volume is low. After that date, the mean ticks/volume drops significantly and suddenly, while the volatility goes up dramatically on an hourly basis. It's like there was a fundamentally change in the way data was recorded/sent. Early data also has a larger difference between high and low on any given bar (1M), and since my EA is basically an oscillating scalper, it makes great use of this. What puzzles me is the overnight change in the nature of the data, which is permanent from that point forward. By "oscillating" I mean, it uses SELL signals to close a BUY, and BUY signals to close a SELL, and of course opens the appropriate trade immediately following the close. I do use a stoploss, but just for emergencies. Bad trades are closed by timing out, thus the parameter "TimeOutMinutes" -- it's like a stoploss, but simply closes the trade if it expires. I ran several optimizations on this one parameter, and 31 minutes was best. The principle for the timeout is simple, scalping programs do best with high # of trades, so it doesn't want to waste time waiting for bad trades to turn around.
JDC
btw, here is the .ex4 in case anyone wants to try this back test. Try the dates in the report above and let me know if you get the same result (use EURUSD/H1).
btw, here is the .ex4 in case anyone wants to try this back test. Try the dates in the report above and let me know if you get the same result (use EURUSD/H1).
Is your Ea with Averaging Strategy, If true maybe your Margin not enough, or maybe you can use microLot.
Is your Ea with Averaging Strategy, If true maybe your Margin not enough, or maybe you can use microLot.
I don't use any averages... I don't use the chart either for that matter. I create a custom "tickstream" by filtering out insignificant ticks (i.e. +/-1pip repitition) then look at the tick by tick differential. It trades that fast, which is why there are 68k+ plus trades in the back test illustrated above.
I'm in no way saying this is a good strategy (actually, I have said the opposite!). The question I am trying to get to with this post is whether anyone is aware of a policy or software (client or server) change in Sept-06, or perhaps a change in the source of historical data.
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
You agree to website policy and terms of use
ok, so I wrote this EA that does "great" on back tests (see above) up to September 19, 2006, and then it starts losing money (EURUSD) - literally overnight on that day. I've posted the above picture to show what ridiculous extremes can be achieved, and fully admit that on a live account this EA will lose, but I'm trying to understand WHY. I've been at this for less than a year, so someone please enlighten me. Thanks!
JDC
p.s. take a close look at "total net profit" in the summary (MT4 doesn't show enough digits on the vertical axis of the chart)