The Ghillyoar System (buying highs and selling lows)

Quoting fxtrue

Disliked

Just to clarify for those who can't understand the strategy ... 1) it trades with fixed lots 2) it's trying to pyramid i.e. add to a winning position, it's not averaging / Martingaling (adding to a losing position) 3) exit rule #2 is to get out of a stalemate situation where price has triggered trades in one direction and then reversed - this is especially relevant during flat market conditions where a growing number of buys and sells could accumulate without achieving the target profit 4) more exit rules are required to improve the strategy's ability...

Ignored

Quoting fxtrue

Disliked

As promised, I attach Paddy9's (moderately) cleaned-up code for the Ghilly Oar system that generated the provided test results. Paddy9 explained that the purpose of this thread was to encourage some "thinkers" (rather than "takers") get involved and try adding additional exit rules that might benefit the system and be adopted if useful. You will therefore find example code at lines 600-602 than can be copied and used to add additional rules. I hope that useful rules will be shared ...just as I hope that the source code of other EAs for this strategy...

Ignored

Quoting gerval

Disliked

{quote} Based on the pyramid that "FXTRUE" comments and as he says, the biggest problem I see is that the price enters the range, so we would accumulate many operations that freeze each other. Or that the price goes in one direction and turns around which would leave us with some operations in favor and more against and further away, so the price should move too much in the direction of the turn. In my opinion, we should never allow positive trades to go negative and drag them, so when opening an order, a breakeven (+ minimum pip) could be placed...

Ignored

Quoting tesiag

Disliked

{quote} thanks fxtrue.....analysis of your code shows that Buy_Bench and Sell_Bench is updated at every bar; I understood they were to be updated once the system exit from all positions in gain (rule#1)..while they re kept constant when the system exits all positions in loss (rule#2).

Ignored

Quoting gerval

Disliked

{quote} Based on the pyramid that "FXTRUE" comments and as he says, the biggest problem I see is that the price enters the range, so we would accumulate many operations that freeze each other. Or that the price goes in one direction and turns around which would leave us with some operations in favor and more against and further away, so the price should move too much in the direction of the turn.

Ignored

Quoting neosqualls

Disliked

Hi Fxtrue, Again i mention averaging / Martingaling and that's the only part of what i said that have been kept, anyway i asked Paddy if it was a hedging system. From my experience, and some others ... in real market condition, hedging strategies can give very different results from backtest. Thanks for your work, you code better than i do

Ignored

Quoting fxtrue

Disliked

{quote} Hi, Neosqualls ...as I assume Paddy9 explained, it's not averaging or Martingaling because buys are only added as price goes up, and sells are only added as price goes down (i.e. pyramiding). It's also not really a hedging system because as soon as buys = sells and the buy or sell trades have a profit of $20 then everything is closed out with a net loss. I do agree that there would be differences between live trading and what the strategy tester presents (mostly due to the fixed spread of the tester), but I think the differences would be...

Ignored

Quoting tesiag

Disliked

is it correct (from a backtesting point of view) to allow an entry soon after (i.e. within the same bar) the exit, or viceversa an exit soon after (i.e. wihin the same bar) the entry? your EA shows this can be possible. my system which (should) implement your exact rules (apart from not allowing entry and exit in the same bar) sometimes shows different results, which is quite strange for simulations on the long run (months).

Ignored

Quoting fxtrue

Disliked

{quote} This is not quite correct as it's not really possible to accumulate "many positions"... remember that the system is trading 0.1 lots and shooting for a $50 target - that's only a 50-pip move for a single entry and the target has been achieved. Also, since rule #2 closes the batch if "buys = sells" and "buy profit or sell profit > 20" then it's unlikely that there would ever be more than 2 buys locked by 2 sells.

Ignored

Quoting gerval

Disliked

{quote} ok, maybe what I mentioned was not exactly the strategy, but more generally the pyramid. I will reread the strategy to adjust to it.

Ignored

Quoting neosqualls

Disliked

{quote}i looked at the code you posted and using the normal distribution as target or filtering instead of the EMA might help

Ignored

Quoting fxtrue

Disliked

{quote} @neosqualls That's an interesting thought - how would you suggest the normal distribution be calculated (for a coder, my maths ability is oddly minimal ) Is anyone else using normal distribution? If so, please share your thoughts & experiences.

Ignored

int period = PERIOD_D1;
double x = 0;
double normDist = 0;
 
for (int i = 1; i <= Population; i++) {
     x += (MathAbs(iHigh(Symbol(),period, i) - iLow(Symbol(),period, i)) / ((iHigh(Symbol(),period, i) + iLow(Symbol(),period, i)) / 2)) * 100;
}
x *= (1 / Population);
for (int j = 1; j <= Population; j++) {
     normDist += MathPow(((MathAbs(iHigh(Symbol(),period, j) - iClose(Symbol(),period, j)) / ((iHigh(Symbol(),period, j) + iLow(Symbol(),period, j)) / 2)) * 100) - x, 2);
}
normDist= MathSqrt((1 / Population) * normDist);

Quoting neosqualls

Disliked

Also, what do you think about reverse the bench if we have for example 2 buys, 2 sells, we buy at the sell bench and sell at the buy bench ?

Ignored

Quoting fxtrue

Disliked

{quote} First up, thanks for that normal distribution example - will be interesting to see how that can be applied. As for the reversal idea, my first thought is that it turns the strategy into a martingale system as we'd be adding to a locked loss at that point. Sure, there's a 50/50 chance that price would rise, but it would need to move a long way in my view to have any meaningful impact on the original locked loss, and the bigger question would be what happens if price continues to fall now rule #2 no longer applies and there's no mechanism...

Ignored