Honing a strategy | Page 3 | Forex Factory

Options

Go to
Permalinks
Print Thread

Similar Threads

AMAZING Strategy!!! 3,266 replies

WRB/MC strategy:Coder needed to semi-automate my strategy 11 replies

strategy builder/tester for new scalping strategy 1 reply

Is opposite of losing strategy a winning strategy? 21 replies

Daily strategy report 1 reply

Attachments: Honing a strategy
Exit Attachments

Honing a strategy

  • #41
  • Oct 2, 2015 10:45am Oct 2, 2015 10:45am
Quoting Atokys
{quote} I use backtesting as a way of very bluntly determining a strategy's technical validity. You make a fair point about those issues, if they are a concern for you then forward testing is a solution.
This is obvious (about to teach my grandma to suck eggs) but the difference between a strategy that returns a profit and one that doesn't can be tiny.

If you've got something that (after taking out spread and commission) makes x pips per month, with proper money management, you can make an absolute fortune.

And the difference between a winning strategy and a losing strategy might be a filter where you only place trades when the spread is < 1 pip.
  • #42
  • Oct 2, 2015 10:53am Oct 2, 2015 10:53am
Quoting BigOldFella
{quote} This is obvious (about to teach my grandma to suck eggs) but the difference between a strategy that returns a profit and one that doesn't can be tiny. If you've got something that (after taking out spread and commission) makes x pips per month, with proper money management, you can make an absolute fortune.
I don't think there is any strategy that can return a consistent number of pips per month.
Quote
And the difference between a winning strategy and a losing strategy might be a filter where you only place trades when the spread is < 1 pip.
If that slight variation caused a profitable strategy to turn unprofitable you are better off discarding the strategy completely.
  • #43
  • Oct 2, 2015 10:58am Oct 2, 2015 10:58am
Interesting chat guys but bed calls. Be back in three days. :-)
  • #44
  • Oct 6, 2015 4:41am Oct 6, 2015 4:41am
Quoting Atokys
{quote} I don't think there is any strategy that can return a consistent number of pips per month.
Sorry, I meant return a positive result most months, not a consistent number of pips.

So now, I've run my EA through 10 years worth of data an all 28 currencies and 12 pairs returned a loss. The 16 other returned various positive gains.

Obviously, I couldn't filter by relative strength or the status of other pairs but surely these filters can only enhance the profitability...or is that an assumption I can't make?

I guess this is where forward testing takes over from the back testing.
  • #45
  • Oct 6, 2015 5:42am Oct 6, 2015 5:42am
Quoting Copernicus
Interesting chat guys but bed calls. Be back in three days. :-)
bed...three days..if only
anyway...a lot of what I'm reading seems to apply to the position I'm currently in where by I have a strategy which 'works' and has done for around 6 months now but I'm so sceptical and terrified its going to stop. I'm not a big fan of backtests certainly over massive 10 year periods as I believe the markets change so much that no strategy will perform over single pairs over that duration. The principles behind the strategy are what need to be rock solid , that is to say when are you trying to catch the market, why and under what circumstances. The actual technicalities of the method in my eyes will always have to be altered to achieve the same principles over time. In my experience this has led to changing the groups of currencies I trade at any particular time. The filter I'm struggling to implement is one which will help recognize this on higher TF. Unfortunately I'm a numbers man and at the moment my trading still requires this certain degree of discretion which I'm really struggling with as I like things quantifiable.
I much prefer trading simulation over backtests, but don't get me wrong backtests have their place in gathering statistics to gauge TP/SL etc and in building the initial strategy.
I don't think adding extra filters equates to curve fitting unless your bending your pre set rules to say 'oh that would have been a winner if...' Its knowing that nothing is perfect and knowing when to stop adding the filters and strip it back to all you absolutely need and realise the rest is MM.
  • #46
  • Oct 6, 2015 5:42am Oct 6, 2015 5:42am
Quoting Copernicus
Interesting chat guys but bed calls. Be back in three days. :-)

If you're really going to sleep for that long, I recommend a little more caffeine, or a little less brandy, or something ...
  • #47
  • Oct 6, 2015 6:52am Oct 6, 2015 6:52am
Quoting Xela
{quote} If you're really going to sleep for that long, I recommend a little more caffeine, or a little less brandy, or something ...
I have two kids under 5, a full time job and a forex addiction...I'm lucky to get 3 hours let alone 3 days.
  • #48
  • Oct 6, 2015 7:32am Oct 6, 2015 7:32am
Quoting BigOldFella
{quote} Sorry, I meant return a positive result most months, not a consistent number of pips.
If you manage this that would also be quite a feat.
Quote
So now, I've run my EA through 10 years worth of data an all 28 currencies and 12 pairs returned a loss. The 16 other returned various positive gains.
Its common to have varying performance from different markets. I would also suggest testing it out on other commodities as, right now, your test results are concentrated in the currency markets.
Quote
Obviously, I couldn't filter by relative strength or the status of other pairs but surely these filters can only enhance the profitability...or is that an assumption I can't make?
Could you elaborate on that idea?

Anyway, only you would know the answer since you have maker's knowledge of your system. If your filter makes logical sense within the context of your system and it improves performance then it seems like a reasonable addition. This is up to your own discretion.
  • #49
  • Oct 6, 2015 8:39am Oct 6, 2015 8:39am
Quoting Atokys
Quote
Obviously, I couldn't filter by relative strength or the status of other pairs but surely these filters can only enhance the profitability...or is that an assumption I can't make?
Could you elaborate on that idea?
When I say "the relative strength" I mean how strong it is in relation to the other currencies: -
Attached Image


If, in this scenario, you wouldn't sell GBPUSD becasue GBP is currenty stronger than USD. If I filter out all the trades (from the 6 months worth of historical trades I have in a database) that go against the relative strength then it instantly removes about 55% of the losing trades and about 14% of the winning trades.

By status of other pairs, I mean if the Ash of GBPUSD is green then I want the Ash of EURUSD to be green too.

Neither of these things can be tested in the Metatrader Strategy Tester. But both of them eliminate more bad trades than good.

If I run my EA through the strategy tester on GBPUSD from 1st Jan 2010 to now I get a nett gain of just under 800 pips. I suspect if I could apply just the currency strength rule alone it would increase the gain significantly.
  • #50
  • Oct 6, 2015 8:51am Oct 6, 2015 8:51am
Quoting BigOldFella
{quote}{quote} When I say "the relative strength" I mean how strong it is in relation to the other currencies: - {image} If, in this scenario, you wouldn't sell GBPUSD becasue GBP is currenty stronger than USD. If I filter out all the trades (from the 6 months worth of historical trades I have in a database) that go against the relative strength then it instantly removes about 55% of the losing trades and about 14% of the winning trades. By status of other pairs, I mean if the Ash of GBPUSD is green then I want the Ash of EURUSD to be green too....
That seems to make sense.

Btw, what is this Ash indicator?
  • #51
  • Oct 6, 2015 9:30am Oct 6, 2015 9:30am
Quoting Atokys
{quote} That seems to make sense. Btw, what is this Ash indicator?
Thanks, it's nice to get some confirmation from a veteran when you're a newbie.

It's the AbsoluteStrengthHistogram from Udine's thread.
Attached File(s)
File Type: ex4 AbsoluteStrengthHistogram.ex4   19 KB | 250 downloads
Attached File(s)
File Type: mq4 AbsoluteStrengthHistogram.mq4   9 KB | 253 downloads
  • #52
  • Oct 6, 2015 9:47am Oct 6, 2015 9:47am
Quoting BigOldFella
{quote} Thanks, it's nice to get some confirmation from a veteran when you're a newbie.
lol

But really, dont take it as any form of confirmation. What goes into your system has to make sense to you. When markets don't go your way, your mind will do everything to make your system look broken and nonsensical. The psychological element of applying a system cannot be stressed enough.
Quote
It's the AbsoluteStrengthHistogram from Udine's thread. {file}{file}
I don't have a way of analysing it. In simple terms, what does it do or attempt to do?
  • #53
  • Edited 5:01pm Oct 6, 2015 9:59am | Edited 5:01pm
Quoting Spoon138
{quote} bed...three days..if only anyway...a lot of what I'm reading seems to apply to the position I'm currently in where by I have a strategy which 'works' and has done for around 6 months now but I'm so sceptical and terrified its going to stop. I'm not a big fan of backtests certainly over massive 10 year periods as I believe the markets change so much that no strategy will perform over single pairs over that duration. The principles behind the strategy are what need to be rock solid , that is to say when are you trying to catch...
That was a long sleep....../tic :-)

I know what you mean regarding your concern on how long that great performance will last but despite it's bad rap I really find that back-testing can help reduce this anxiety to a degree.

The main point with extensive backtests is to simply test the validity of the system over a wide range of market conditions. It is a process to identify the strong points and weak points of the strategy. While you might not capture all market conditions over a 10 year period, you would certainly be able to deduce those market conditions where your strategy flourishes versus those market conditions where your system flounders. If you find particular conditions that lead to intense drawdown periods, then this gives you valuable knowledge 'ahead of live trading' of the weak spots of your trading strategy and you may decide that there are ways to plug these holes through strategies such as 'market avoidance' (using filters or periods or trading halts) or use alternate strategies that flourish during these times and complement overall program performance. Extensive backtesting also gives you valuable information that can be used to assess whether the system is broke or whether the drawdown you are facing is simply part and parcel of a particular set of market conditions that may soon pass. Obviously there needs to be a time when you say 'enough is enough' but back-testing can help in defining when you should be calling it a day.

The tail-end of the back-testing process should always involve out of sample testing to see how the system performs with unseen data. This is a necessary element of back-testing that ensures you have not attempted to curve fit during the process. If the system performs in accordance with backtest expectations over the unseen data, it means that the system is a robust one.

Furthermore if you find that your strategy can still deliver a positive edge over this extended range of market conditions (of say 10 years) then you have found an excellent baseline strategy that is sufficiently robust to be 'turbo charged' to adapt to changing market conditions using 'rolling' walk forward testing. This form of testing follows on from the backtesting phase of strategy development and is a successive out of sample test whereby the baseline strategy is optimised over say 2 years of data history and then applied over 3-6 months or so of unseen data. Following this process the optimisation window moves ahead by a 3-6 months period and a new optimisation takes place for the next 3-6 months of forthcoming data.

If the ratio between the annualised net profit relating to the walk forward tests versus the annualised net profit results of the optimisation periods is > 100%, then the system is efficient and is performing well and it's forecasting potential remains high. Remember however that good performances of a couple of walk forward tests may be simply luck so it is best to give a length of time to ensure a number of walk forward tests are conducted before a decision is made as to whether the system is broken or not.

......anyway I hope this helps a bit. :-)
  • #54
  • Oct 6, 2015 10:26am Oct 6, 2015 10:26am
Quoting Atokys
{quote} dont take it as any form of confirmation.
I meant confirmation that my "sense" matches someone else's "sense".

The Ash shows when a smoothed moving average with a shift of 1 is above or below a smoothed moving average with the same period and a shift of 0.
  • #55
  • Oct 6, 2015 10:35am Oct 6, 2015 10:35am
Quoting BigOldFella
The Ash shows when a smoothed moving average with a shift of 1 is above or below a smoothed moving average with the same period and a shift of 0.
What does the shift refer to? What values do you use for the 2 MAs? What granularity of data are they applied to?
  • #56
  • Oct 8, 2015 5:36pm Oct 8, 2015 5:36pm
Quoting Atokys
{quote} What does the shift refer to? What values do you use for the 2 MAs? What granularity of data are they applied to?
The Ash indicator has a number of settings.

The one that's used in Udines thread (on M30) uses a smoothed MA(5) and a signal MA(3). When the signal is above the smooth it's green else it's red.

The shift is where the actual value of the moving average is shown (or shifted ahead).

I use a shifted moving average and a non shifted moving average to tell if a market is ranging or not. If it's crossed more than once in the last 4 bars then the market is ranging.
  • #57
  • Oct 8, 2015 5:43pm Oct 8, 2015 5:43pm
Quoting Xela
If it's of any interest, the only one that doesn't apply to my own trading is the trailing SL mentioned in point 8 above: for myself, I need to see really convincing statistical evidence that a trailing stop confers additional profit, before using one.
It definitely is of interest.

I've done some back tests and forward tested over the last 2 weeks and the results are in...

And you were right, running the trailing stop has actually been detrimental to my bottom line. If I'd let the trades run to either TP or SL I'd be hundreds of pips better off.
  • #58
  • Oct 8, 2015 5:59pm Oct 8, 2015 5:59pm
Quoting BigOldFella
{quote} It definitely is of interest. I've done some back tests and forward tested over the last 2 weeks and the results are in... And you were right, running the trailing stop has actually been detrimental to my bottom line. If I'd let the trades run to either TP or SL I'd be hundreds of pips better off.
Nice one BOF. The benefits of back-testing. While many Back-test nazi's here on FF state that back-tests amount to naught in the scheme of things, it is findings such as these that give validity to decisions as opposed to 'brain farts' and subjective opinion......but for even more validity IMO, extend the sample size even more to improve the confidence. Ensure your sample size is sufficiently large to cover a wide range of market conditions then you give no room for subjective opinion to move.....checkmate :-)
  • #59
  • Oct 8, 2015 6:09pm Oct 8, 2015 6:09pm
Quoting BigOldFella
I've done some back tests and forward tested over the last 2 weeks and the results are in... And you were right, running the trailing stop has actually been detrimental to my bottom line. If I'd let the trades run to either TP or SL I'd be hundreds of pips better off.

Well done - I found the same, with all but one of my systems, when I back-tested methodically over statistically significant numbers of trades. (Curiously, there was one minor system, I must admit, for which the trailing stop did actually prove beneficial, but from my own evidence I strongly suspect that for most systems, superficially attractive though they look/feel, they cost money.)
  • #60
  • Edited Oct 9, 2015 4:09am Oct 8, 2015 6:25pm | Edited Oct 9, 2015 4:09am
Quoting Xela
{quote} Well done - I found the same, with all but one of my systems, when I back-tested methodically over statistically significant numbers of trades. (Curiously, there was one minor system, I must admit, for which the trailing stop did actually prove beneficial, but from my own evidence I strongly suspect that for most systems, superficially attractive though they look/feel, they cost money.)
I have found that too Xela. It seems that there are certain dependencies that exist with stop positioning placement dependent on technique so it is a worthwhile exercise to test this assumption over large sample sizes during strategy development for each independent method. :-)

Example: If trend following using Standard Deviation Channels as your guide for entry and stop placement, a trailing stop established at say 2 SD's away from the line of best fit acts not only as a risk management measure but also an optimised trade exit condition that confirms when the directional price movement ends in accordance with your chosen definition of a trend. However if using a momentum based strategy such as a Donchian Breakout technique where you are looking for a breakout move as opposed to a sustained directional price movement, then a trailing stop is too precise to offer benefit and does not give sufficient room to breathe as volatility expands. However you then need to determine an effective trade exit condition which for momentum based strategies appear to be more successful with defined TP's as opposed to stop exits.
0 traders viewing now