The Ghillyoar System (buying highs and selling lows) | Page 9 | Forex Factory

Options

Go to
Permalinks
Print Thread

Similar Threads

Daily Highs and Lows to Previous Daily Highs and Lows 1 reply

Higher Highs/ Lows and Lower Highs/ Lows 14 replies

buying and selling the news, eurusd 29/4/08 1 reply

Difference between selling a Call and buying a Put? 5 replies

Buying Dips and Selling Rallies 16 replies

Attachments: The Ghillyoar System (buying highs and selling lows)
Exit Attachments

The Ghillyoar System (buying highs and selling lows)

  • #161
  • May 11, 2021 12:25pm May 11, 2021 12:25pm
Quoting tesiag
{quote} don't think so...in such case I would expect an exponential (at least a non linear) growth of the equity...like when you apply money management techniques to a succesful strategy the equity growth is completely linear...so fixed parameters..ora at least not linked to something constantly growing
Who knows
  • #162
  • May 11, 2021 12:45pm May 11, 2021 12:45pm
I have not "sort of" returned to answer any questions at all.................and I will not.

If questions are asked by an individual ...........the community can answer ..............if they wish.

If provoked ............I may from time to time ask questions.............I would rather not.



I have already asked my first question................by posting 2 charts.....I should have included words.

I did not ...........because I did not want to offend some of the kind people who have supported the thread from the start.

I am starting to annoy even myself now..............in future I will articulate the question properly ............or remain silent.




My first question was :

"" You are all saying the returns are very low.........how do you know that.?......have you really tested........ all ......or any ........of the variables that I gave you ?


The two back tests that I gave you .........politely ...........pointed out that other outcomes are available.

There was a third back test ......I am not going to post it..........I am definitely not promising to make anyone a millionaire.....
  • #163
  • May 11, 2021 2:37pm May 11, 2021 2:37pm
Quoting Paddy9
I have not "sort of" returned to answer any questions at all.................and I will not. If questions are asked by an individual ...........the community can answer ..............if they wish. If provoked ............I may from time to time ask questions.............I would rather not. I have already asked my first question................by posting 2 charts.....I should have included words. I did not ...........because I did not want to offend some of the kind people who have supported the thread from the start. I am starting to annoy even...

To all people interested in the system kindly given by paddy in this thread
We should first try to backtest the system as it was given (rule#2 included) and only modify the value of each single parameter one by one, in a defined range, while keeping all the others fixed.

We can define a table to identify all the simulation scenarios and maybe distribute the "load" of the backtests of all scenarios among us.

Let me prepare such table and propose to you for agreement and conseguente backtests execution
  • #164
  • May 13, 2021 4:29am May 13, 2021 4:29am
Quoting tesiag
{quote} To all people interested in the system kindly given by paddy in this thread We should first try to backtest the system as it was given (rule#2 included) and only modify the value of each single parameter one by one, in a defined range, while keeping all the others fixed. We can define a table to identify all the simulation scenarios and maybe distribute the "load" of the backtests of all scenarios among us. Let me prepare such table and propose to you for agreement and conseguente backtests execution
ok...sorry for the delay in here below a tentative table which identifies possible scenarios to be backtested, in order for us to collect all results and to assess the "dependence" or "sensitivity" of the EA to the change of basic input parameters.
Parameters kept fixed in all the simulated scenarios are:
1- the backtest period is 10 years (this choise will allow us to investigate the trend of the equity on long (10 years) periods as well as to correlate the equity trend to different market behaviors (ranging, trending etc.) on shorter periods (months, few years, etc.)
2- the TradeSize is 0.1...this parameter is directly related to TargetStep and ExitProfit, so there is no need to modify it when you are going to modify the other two

So find attached the table with the various proposed scenarios...all results should be collected and stored somehow to perform a final analysis to see what seems to be a "good" setting for the EA

One last comment: the simulations are to be done on one single instrument (e.g. EURUSD) on the same time period (e.g. 2000 to 2010 or 2010 to 2020 etc); the (reliable) historical bars should be shared among all people wanting to contribute to the execution of the backteststhe history ...I hope some of you have some good historical time series to be used and shared rather than the one I for example use from my demo account
Attached File(s)
File Type: docx scenario.docx   14 KB | 320 downloads
  • #165
  • May 17, 2021 12:01am May 17, 2021 12:01am
Quoting tesiag
{quote} ok...sorry for the delay in here below a tentative table which identifies possible scenarios to be backtested, in order for us to collect all results and to assess the "dependence" or "sensitivity" of the EA to the change of basic input parameters. Parameters kept fixed in all the simulated scenarios are: 1- the backtest period is 10 years (this choise will allow us to investigate the trend of the equity on long (10 years) periods as well as to correlate the equity trend to different market behaviors (ranging, trending etc.) on shorter periods...
No reply...wow.
  • #166
  • May 17, 2021 10:12am May 17, 2021 10:12am
Trying to over optimize is probably not the way to go. What you will get is the perfect set of parameters for a period of time that doesn't mean the market will be the same in the future.

Being good (or very good) for a specific range of time doesn't mean you will be good in the future. For me, parameters have to be adapted to dynamically fit the market.
  • #167
  • May 17, 2021 1:34pm May 17, 2021 1:34pm
Quoting tesiag
{quote} No reply...wow.
Hi tesiag

Thank you very much on behalf of the friends of Ghillyoar.............you did your best for us.
  • #168
  • May 17, 2021 1:49pm May 17, 2021 1:49pm
Quoting Paddy9
{quote} Hi tesiag Thank you very much on behalf of the friends of Ghillyoar.............you did your best for us.
Hello Paddy,

Could you kindly show live result of your strategy? We know with backtest how it works but can you show us 6 months or more of live results ?

Thanks
  • #169
  • May 17, 2021 1:52pm May 17, 2021 1:52pm
Quoting Paddy9
{quote} Hi tesiag Thank you very much on behalf of the friends of Ghillyoar.............you did your best for us.
We thank you for sharing your system. Following your suggestion to experiment with different settings of the EA I proposed a number of possible settings (called scenarios in the attached word file ) to try....
Don't know if some of you think it would be' worth to peform backtests with such settings or we need to redirect out efforts. In the first case we could distribute the backtest burden among people willing to contribute....
Any idea is welcome
  • #170
  • May 17, 2021 2:07pm May 17, 2021 2:07pm
Quoting neosqualls
For me, parameters have to be adapted to dynamically fit the market.
neosqualls is my harshest critic............and without him .........no one would read the thread...........including me.

I am interested to know which parameters you would be dynamically changing ........and how often.

I ask you this because I believe that you have given my system the sort of critical appraisal that I was hoping for when I started the thread.

Your opinion should be worth hearing.
  • #171
  • May 17, 2021 4:29pm May 17, 2021 4:29pm
Quoting tesiag
{quote} No reply...wow.
tesiag, neosqualls reply (post 166) nails it! Trying your different scenarios is probably going to end up optimizing the EA settings based on some past data which would not necessarily be optimal in future live trading. From my few backtests, I am convinced Paddy9's strategy has legs. I have tried the bench mark with 10, 15, 20 and 30 (the default by Paddy9) pips above/below the second candle, and 30pips for some reason produced the best result. So, Paddy9 must sure have his reason for choosing 30 pips. With the other settings, lot size (in relation to your account balance) appears to affect different pairs differently, so this would be one setting to me that needs to be considered depending on which pair you are trading, and targetstep and exitprofit can then be adjusted accordingly to lot size.
  • #172
  • May 17, 2021 5:24pm May 17, 2021 5:24pm
Guys I never meant to suggest different scenarios backtesting in order to find the magic settings who always make the system an overall winner.

Different scenarios allow, in my opinion, to anallyze the EA behavior in the same market context (same time period, same instrument) in response to different parameters...
This would help to answer questions like..."which parameters you would be dynamically changing..and.how often" (or even."when" I would add)

A backtest is not simply useful to see whether at the end of the day you will turn positive or negative in balance....but it mostly helps to assess when the system works well and in what circumstances it performs poorly (and thus other settings / tweaking appears necessary).

You can get to such assessments or by analytical thinking/investigation or by experimental backtesting ....don't see many other ways to get any useful hint about a given EA.
  • #173
  • May 18, 2021 5:48am May 18, 2021 5:48am
The question seems to be "what if anything is the point of back testing" ?

If I was the manager of Manchester United ......and I needed a centre forward I could throw in the boy who lives next door to Paddy9.
................I could forward test him........looking at his previous results ...........is after all pointless........is it not ?

On the other hand I could look at the previous performaces of Haaland........and give him a shot.........

Of course they could both fail...........nothing in life can be taken for granted.

I have seen many trading systems that perform well in a back test...........and badly in real life.
On the other hand:
I have never seen a system that peformed poorly in a back test...............and well in real life.
  • #174
  • May 18, 2021 8:48am May 18, 2021 8:48am
I am not saying backtest is useless, it give you the overall view of how a system can be good or bad.

But, Trying to find the best targetStep, the best exitProfit, the best LotSize will lead you to loose, in my opinion.

I am more looking in that way :
- if i know that 80% of the time the price move between +0.5% and -0.5% of the daily open, i should calculate to have my target inside this range.
- the bench could be calculated with this too.

I repeat, this is my opinion, i can be wrong ... or not.
  • #175
  • May 18, 2021 8:58am May 18, 2021 8:58am
Quoting Paddy9
The question seems to be "what if anything is the point of back testing" ? If I was the manager of Manchester United ......and I needed a centre forward I could throw in the boy who lives next door to Paddy9. ................I could forward test him........looking at his previous results ...........is after all pointless........is it not ? On the other hand I could look at the previous performaces of Haaland........and give him a shot......... Of course they could both fail...........nothing in life can be taken for granted. I have seen many trading...

that's why I guess backtesting is always needed...not to get the supersmart (i.e. overoptimized) set of parameters to be blindly applied in real life...but to learn when and where the background logic of the EA is robust enough to make the trader survive the market (if not even make him gain something)


opinions as well as ideas and proposals can be quite different among the followers of this thread..and that's perfectly ok and very welcome ....but the final objective should however be pretty much converging...that is to see how the Paddy9 trading system can be fruitfully exploited in the market
  • #176
  • May 18, 2021 9:05am May 18, 2021 9:05am
Quoting neosqualls
I am not saying backtest is useless, it give you the overall view of how a system can be good or bad. But, Trying to find the best targetStep, the best exitProfit, the best LotSize will lead you to loose, in my opinion. I am more looking in that way : - if i know that 80% of the time the price move between +0.5% and -0.5% of the daily open, i should calculate to have my target inside this range. - the bench could be calculated with this too. I repeat, this is my opinion, i can be wrong ... or not.

I agree....trading is some kind of balance between a forecast exercise and (maybe more importantly) a "control technique". Apart from this, I understand you're proposing not to backtest the EA as it is now but to do it after having made an adjustment of the setting of the benck level, so as to make it dynamically adapted (directly or inversely?) to the volatility of the market .

am I mistaken?
  • #177
  • May 18, 2021 11:51am May 18, 2021 11:51am
So here is an idea :

Let's assume 80% of High/Low is inside +/- 0.5% for the last 10 days ( has to be calculated by a function)

- So the Target could be +/- 0.5%.
- The bench could also be calculated using the SD

Quote
BuyBench = iClose(NULL, 0, 2) + 0.5%
SellBench = iClose(NULL, 0, 2) - 0.5%

I don't think it's very complicate to code.

What is your opinion ?
  • #178
  • May 25, 2021 12:03pm May 25, 2021 12:03pm
Quoting Paddy9
The Strategy Tester available with MetaTrader is of no value to a sensible trader. The longer the period of the test the less the reliability. So here is a test result for The Ghillyoar Trading System covering the last 21 years. The trading period is from 2000.01.01 to 2021.03.23 {image}

Hello friend,

Thank you very much! Today I found this thread. And hopes keep in touch with you guys!
  • #179
  • May 26, 2021 7:55am May 26, 2021 7:55am
Quoting neosqualls
So here is an idea : Let's assume 80% of High/Low is inside +/- 0.5% for the last 10 days ( has to be calculated by a function) - So the Target could be +/- 0.5%. - The bench could also be calculated using the SD {quote} I don't think it's very complicate to code. What is your opinion ?


If I understand him correctly Neoqualls has suggested the following two tweaks to Ghillyoar:

1. Add 0.5% to the opening price and that is your buy target..............replacing the target of 50$ profit for the basket of open trades.
I think that this means that ........he is discarding the basket concept .......and is replacing it with individual targets for individual trades.


2. At present we add 30 to the buy/sell bench. I think that Neosqualls is saying why not replace "30" with 0.5% of the opening price.


Is my understanding correct ?
  • #180
  • May 26, 2021 10:37am May 26, 2021 10:37am
Quoting Paddy9
{quote} If I understand him correctly Neoqualls has suggested the following two tweaks to Ghillyoar: 1. Add 0.5% to the opening price and that is your buy target..............replacing the target of 50$ profit for the basket of open trades. I think that this means that ........he is discarding the basket concept .......and is replacing it with individual targets for individual trades. 2. At present we add 30 to the buy/sell bench. I think that Neosqualls is saying why not replace "30" with 0.5% of the opening price. Is my understanding correct ?...
Yes the standard deviation gives you interesting data, here is an example : if i know that the price will, 80% of time, going above or below open +/- 0.5% i can use this to set my take profit and lot size.

Example : I calculate that over the last X days, the price have been above/below daily open +/- 0.5% i can use this level has target and set my lot size according to 2% or 4% (or whatever you want) of my account balance.

If anyone want to try to code this, i'm interesting to see the results, my coding skills are not good enough i'm afraid.
0 traders viewing now