Improvement of Triple Screen -based Strategy using Expert Advisor

New comment
 

Hello,

A fast story

I use to trade using an strategy based on Triple Screen method by Elder's book Come Into my Trading Room (I describe this on this post). Manually, the operation is quite ok, but not as profitable as I want it to be, so I thought the parameters were not optimized for the market type I was operating, thus first believed that an optimization could considerably improve the profit factor of this strategy. So I implemented an Expert Advisor with the rules I used manually, and the result was very not ok; as I do not include subjective information (reasoning) and another indicators which occasionally I can use in operating time, it's reasonable to accept the bad results of this pure-logic robot. But still I believe it can modified to achieve reasonable (but profitable) results in real market (because Triple is commonly accepted as a simple and good strategy, right? Or my understanding of it may be wrong).

Strategy

With this, I operate only Buy positioning. 

In short, I use 4 EMAs and 2 MACD Histograms to find a High Trending:

  • As intermediate screen: 
    • EMA Fast : EMA (5)
    • EMA Slow: EMA (21)
    • MACD histogram from MACD (12, 26, 9)
  • As long screen (use 5 times factor for each indicator parameter):
    • EMA Fast : EMA (25)
    • EMA Slow: EMA (105)
    • MACD histogram from MACD (60, 130, 45)

And I put a Buy Stop when both for intermediate screen and for long screen occurs:

  • EMA Fast > EMA Slow 
  • MACD Histogram increases
The Buy Stop parameters are:

  • Price > currentTick.bid
  • SL = historicalRates[1].low
  • TP = bid + TP_LEVEL * (Price - SL), TP_LEVEL is optimized for 3.0
There is only a Buy (of a standard quantity) then there is no position (=there is not avalanche buy).

And when positioned, move SL and TP according to:

  • newTP = TP_LEVEL * (hRates[2].low - currentSL)
  • newSL = historicalRates[2].low
Also when positioned, Sell At Market, if any screen condition fall (EMA Fast < EMA Slow or MACD Histogram decreases, for any screen).

And finally, the new conditions are only checked when it comes a new bar. 

Operation

I put this strategy to operate BM&FBovespa, for index WDOH16, in M1 period, in a demo account. The results were not good. Not disastrous, but returned a small loss.

I realize that the majority of trades loses are due to early SL activation, so the price is not "mature" to release. I thought several ways to change this, like putting a fixed/optimized SL_LEVEL, or averaging the historicalRates[i].low, or even considered neural networks to optimize this; so I realized I needed a help from you guys more experienced than me in stocks operations.

So?

Would you please give me some advices on how to fix/improve/change/throw away this strategy?

Thanks in advance. 

New comment
Reason: