New article Research of Statistical Recurrences of Candle Directions has been published:
I also had a look at this. Turns out if you bought eurusd from hour 0 to hour 1 for the last year you'd have made 1 300 pips in the last 365 days. If you then closed and sold hour 1 to 2 you'd have made another 600. There's over 30 000 pips to be had like this in a year. Certain hours of day for example hours 11, 14 and 15 have very strong directional trends. So in theory you can identify the trends for the hours and then simply sell or buy for that hour.
All we did the was deduct the close from the open to see if there were strong directional patterns at times of day. See attached.
The results were extremely interesting. There are very strong time based patterns.
If you think about it the same people get their desk at the same time and do the same thing everyday. And if that thing is selling Euro's, then they're going to sell Euro after lunch, or after arriving at work or at NY open. Data release also happen at set times.
The problem with this is that there is no way to pick up quickly enough when the trend changes. You'll lose all your profit by the time it becomes obvious that where you use to have to sell euro in hour 0 that you now have to buy it.
So the same guy that was selling after lunch might be buying after lunch now, or the guy that was selling after lunch has gone under because he should have been buying and if he was big player that pattern has now changed.
So there are very strong time trends. How to make a strategy from that still eludes me. We wrote several EA's around the principle, but there's no way to pick up a trend change in time.
This is a great trading system.
First, is it possible to know what day of the week it is with MQL? Like Sun, Mon, Tues, Wed?
Second, I'd like to design a system based on candle counting. If one looked at the week before and traded the 4hr or 1hr charts where ever they saw the strongest pattern I bet it could work. You could look for multiple signals or only ones that are on a massive streak, say (six times in a row). The best part about this is that if you add filters like a 200MA or Time of Day, I think you could pick one winner each day and pull off 55% correctness.
I'll build it and post my findings. Usually when I think I have a great idea it goes to hell though. When I think something won't work, it does... Good ol' fx
int DOW = TimeDayOfWeek(Time); /* http://forum.mql4.com/33851
// reports DayOfWeek() always returns 5 in the tester. No refresh?*/
I also did some candle analysis on the 15 eur/usd chart. Looks pretty interesting that there are patterns statistically significant differences emerging when looking at the first 3 candles. You can start to devise an entries and exits around these patterns.
How to translate :
"Mas vira maina" in english ?
"Mas vira mainas" it's not russian. But the script gives me faulse results ! What about you ?
Vous êtes français ?
"" in english ?
Go to the www.themoneyclock.com and watch the whole video. August 10, 2012