I’ve just done 5 year GBP/JPY backtest on this really weird strategy of Kathy Lien’s called the “7 day price extension fade”. Its very weird though because it gives signals VERY rarely. Also positions are entered in two parts, so actually the total trades is half what the stat sheet shows, giving a total trade frequency of about 2 trades per year!
Time:
Days processed=1795
Months processed=58.98
Trades:
Total trades=18
Profit trades=14
Loss trades=4
Profit trades cons.=8
Loss trades cons.=2
Trades / day=0.01
Trades / month=0
Profit trades / month=0
Loss trades / month=0
Max profit trade=22.62
Max loss trade=9.69
Income:
Net profit=115.35
Gross profit=154.12
Gross loss=38.77
Profit / month=1.96
Average profit=11.01
Average loss=9.69
Max drawdown=27.14
Profit factor=3.97
Return, %=11.53
Other statistics:
Max lot used=0.03
Restoration factor=4.25
Reliability factor=0.07
Is this just a pointless exercise I wonder? I had an idea I could backtest another bunch of pairs using the same five year time frame, and if it works perhaps I’d be getting 10-12 trades per year, which I guess might make it a useful secondary strategy?
I feel like I’d be dead before I actually proved it in demo testing though…
I cannot vouch for your test results obviously. But the sheer frequency of trades is probably quite correct i.e. it’s not often that an instrument goes straight up or straight down for seven days solid in a row.
Go ahead and test it on other pairs though and you may be pleasantly surprised. Some pairs for sure trend better than others so you’re more likely to get more of these seven days in a row type movements.
Overall it made a profit though so let’s not forget that. Assume then that you get the same or similar results on say another ten pairs for the same period??? Then all of a sudden those results are not that bad anymore.
There is another possibility (I cannot remember if it’s mentioned in her book by her but this is not something new this particular type of trading system) and that is to trade the same type of thing on the hourly timeframe. That would give more trades obviously but the profits will be less and the chances of getting stopped out become higher of course.
One very important thing with her systems though: they work MAINLY because of the way the trades are managed NOT because they’re the best systems in town. Most if not all will have you TP at a certain predetermined level and then move your stop to breakeven and hope for the best thereafter. Not sure if this is factored into your testing but bear it in mind.
Sometimes the gain from a project like this is not so you can adopt the whole strategy but so you can extract helpful elements. Right now two of my manual exit rules is to close a profitable long position as soon as it has made 3 consecutive higher daily closes or 5 consecutive higher daily highs, whichever comes first. Which I think is parallel to Kathy Lien’s observation on price sequences - they fail. Though I wouldn’t use that knowledge to select entries.
Points well made thanks again guys. Also credit where it’s due, I actually found that system from a link to an e book @dpaterso posted.
I might try out that exit system too @tommor if I end up backtesting a system that could use it.
Backtesting is surprisingly great btw. I just couldn’t do it effectively without forex tester, and pine script in trading view was awful. Backtesting is all I’ve been doing for about two weeks. Getting to see price play out quickly and make decisions really helps accelerate the adoption and discarding of ideas. Also it reminds me of computer games!
Those two manual exit triggers make sense in a trend-following set-up. So as soon as I’ve closed the trade I’m watching that chart for a place to set a new entry order and get right back in, next day if possible, but I allow the price some slack so it has to prove to me again its going to continue the trend.
And presumably at some point you have reasoned out that these manual exit triggers for you give a superior gain over the long run than say a 1 or 2x ATR trailing stop? Or does a trailing stop figure in there somehow also?
I’ve noticed during backtesting over the past few days that if my trailing stops are wide enough I can end up doing something which I imagine is similar to pyramiding, where the first trade is still trailing, but I go in again because a new entry trigger has been received.
Yes, trial and error. Simple reasoning that the initial SL represents the maximum individual loss, and if the initial SL gets hit too often, I will incur the maximum aggregated loss. So the idea is to set the SL where I can afford for it to be hit but where I can almost guarantee it won’t be.
I don’t like trailing stops, I’ve played around with them but find its too crude - half the time they’re triggered where I would have over-ridden the stop and stayed in, and half the time I have already got out before they can be triggered.