forex software

Create and Test Forex Strategies

forex software

Skip to forum content

Forex Software

Create and Test Forex Strategies

You are not logged in. Please login or register.


Forex Software → Forex Strategies → Question: Why do we obsess about optimizing our strategies?

Pages 1

You must login or register to post a reply

RSS topic feed

Posts: 14

1 (edited by sleytus 2019-01-09 01:38:33)

Topic: Question: Why do we obsess about optimizing our strategies?

This is a rhetorical question -- since I think I already know the answer -- but I thought I would pose it anyways to see if anyone was interested or had some insight to share or even had a question.

There are many manual (forex) traders who are profitable.  In fact, manual traders may be more profitable than algorithmic traders.  I think this is because as a manual trader one has to pay attention to details and the data and have some basic understanding of the indicators being used.  An algorithmic trader simply inserts his (magically-generated) EA into the MT4 platform and then keeps his fingers crossed that it makes money.

Manual traders use indicators -- just like we do here.  However, they rarely (if ever) think about optimizing or calibrating their indicator's settings -- i.e. they simply use the default values.  It appears the default settings values for most indicators are "good enough" for most time frames and most currency pairs.

So my question -- if manual traders use indicators and if manual traders are more profitable than algorithmic traders, then why do we spend so much time worrying about optimizing / calibrating the settings used by indicators in our software-generated strategies?

2 (edited by geektrader 2019-01-09 02:28:41)

Re: Question: Why do we obsess about optimizing our strategies?

Hi Sleytus,

I think your rhetorical thinking is flawed in the first place because of this: show me the steady profitable manual traders. I do not know any! I know some that made money a few months, or maybe a year, but that´s what we can achieve too by a lucky EA. I have yet to see a manual trader being profitable > 1 year, but if they exist, show me their MyFxBook :-) And don´t believe their words if they just TELL you that they are profitable, because most people HATE to fail and will never admit it, especially not in trading.

3 (edited by geektrader 2019-01-09 03:42:03)

Re: Question: Why do we obsess about optimizing our strategies?

P.S.: From the 10 years of trading self-built EA´s live, I can tell you that highly optimized strategies (R-Squared > 98) but with FEW indicators do great going forward compared to the ones with "ugly" looking equity curves. Especially if they work on multi-symbols with at least an R-Squared > 80 too, even after optimization on the main-symbol.

Re: Question: Why do we obsess about optimizing our strategies?

geektrader wrote:

I think your rhetorical thinking is flawed in the first place because of this: show me the steady profitable manual traders.

Hello @geektrader -- thanks for your response.

What you say makes perfect sense -- i.e. "show me the evidence".  And that's part of the problem -- no one has any evidence one way or another.  Furthermore, everyone and their brother is an expert and feels compelled to give advice -- this is how the Internet works.

I don't have any evidence.  All I have is my own anecdotal experience and reading between the lines on a variety of forums and blogs.  I'm a software developer by profession and love algorithmic trading and consider Popov's applications the finest I've ever had the pleasure to work with.  And I've spent the past 6 months also involved with manual trading.  I don't particularly care for manual trading, but I've learned a lot.  And though I can't prove it I do think it is quite likely there are a quite a few manual traders who are profitable and I can understand why.

I'm not suggesting not to optimize -- all I'm sharing is an observation.  And it's fairly simple to check-out.  Take any standard MT4 indicator with its default settings and apply it to a chart -- any chart, any time frame, any symbol.  And -- surprise -- it seems to fit fine!!   So -- I'm simply asking -- why is that?

I can create wonderful strategies with Popov's software with great statistics and back testing.  And I've noticed that the settings after optimization can be quite different than default settings.  Also, about a year (or so) ago in another thread we were having a discussion about how it seemed the strategies with the better statistics seemed to trade worse and perhaps it was better to use strategies that fell more in the middle of the pack.  This wasn't just my observation, this was suggested by the Udemy instructors.

Anyways -- we all have our own path and are too set in our ways to contemplate anything different.  On the surface it may seem we disagree -- but probably not so much.  I just ask more questions than you.  And being able to read someone else's opinion is a hell of a lot better than complete silence.  I learn valuable clues from your posts -- thanks.

Good luck in your trading...

Re: Question: Why do we obsess about optimizing our strategies?

Some off-topic but interesting question:

... how it seemed the strategies with the better statistics seemed to trade worse and perhaps it was better to use strategies that fell more in the middle of the pack.

This is something I think about from several years - what are the best settings and optimizations for strategies in order to provide longer profitable run. Now, when we have Generator and tester for strategies, it appears a necessity of a next level of testing. How to say - a system that optimizes the Generator parameters in order to provide best OOS strategy performance. With other words we need a tester generator and optimizer for the Generator.

Re: Question: Why do we obsess about optimizing our strategies?

I am actually doing an experiment at the moment. I created a portfolio of strategies with 20% OOS data. Put that port on demo and then reoptimised the same portfolio with 0% OOS data and put that on a demo account.

To early to make a conclusion yet but I will posting my results in the future when i believe there is enough trades to draw a conclusion. Which is another question in itself.... When is enough trades to make a conclusion lol

Re: Question: Why do we obsess about optimizing our strategies?

Michael1 wrote:

When is enough trades to make a conclusion

This topic is not about OOS.  There are a number of other threads about OOS.

I'd really like to keep this thread focused on the question I initially posed:
Why do manual traders NOT worry about optimizing indicator settings?

Furthermore -- if you can spare 60 seconds you can test this yourself.  Open any MT4 chart and apply one of the built-in indicators (using the default settings).  You will see that most indicators -- with their default settings -- work fine with the current chart data.

In contrast -- most of us here spend a lot of time optimizing / calibrating indicator settings.  And if you pay attention you'll see the settings can differ quite drastically from the original defaults.  Why is that?

8 (edited by GD 2019-01-09 09:13:24)

Re: Question: Why do we obsess about optimizing our strategies?

Hi Sleytus

The problem of optimization
===================

It looks only small optimizations is fine. Finally I think the 10 wins rule from MT4 Sidekick is good. It depends on each EA statistics.

I hope to see MT4 sidecick used to EA Studio also.

Relative to Popov Idea on Generator
=======================

It sounds fine to me. I think it can be nice to have different Acceptance Criteria for Generator and Optimizer. It looks to be a faster step.

Also, I hope to see EA Studio having in Multitester ability to use for data from different brokers as in FSBpro.

Re: Question: Why do we obsess about optimizing our strategies?

Hi GD -- thanks for your ideas...

It looks only small optimizations is fine

Yes -- I agree 110%.

Here's the deal -- there are two approaches to optimization / calibration:
1. Don't optimize / calibrate -- i.e. stick with the default settings (or only slight variations)
2. Optimize / calibrate

As with most things in life both approaches come with their trade-offs.
1. Positive:   Simple to implement since nothing to do.
    Negative:  Performance will suffer, but may still be "good enough"

2. Positive:    There is the possibility of better performance
    Negative:  Requires some care and feeding -- recalibrating with recent data.  Also, there is the risk of over curve-fitting, which renders the strategy completely useless.

The reason why (2) needs to be recalibrated is because the optimized settings likely differ quite significantly from the default settings and were intended to be used only with the data horizon that was used to calibrate the indicators.  When the patterns in the data horizon change, then the indicators need to be recalibrated.  On the other hand, if you stick with the default settings then you may not need to recalibrate.

To Popov -- here's a suggestion -- allow users to choose one of two approaches to optimization:
Approach #1 -- the way it currently works.
Approach #2 -- only allow settings to deviate slightly from their defaults (i.e. 10% -- or some user-selectable percent)

Re: Question: Why do we obsess about optimizing our strategies?

I'm curious what "optimization" of manual trading system means. I have colleagues that trade manually and we tried to "automate" their systems, however it appeared that there were no strict systems that they can explain. I may say that experienced traders can consider very tiny changes of the market behavior, dynamics, momentum... Some may call that a Sixth Sense or trading by intuition, but I don't agree. When a person observe closely any system or a process he studies models subconsciously. After time, a person can act on nonverbal signs automatically. For example - an experienced driver can drive with hours speaking with a passenger and to even not notice when he arrives to the target. It is the same with the experienced traders. They monitor the chart, the Level 2, or simply the ticker and develop a sense about the market (Tape Reading).

About the indicators, all of them take either the price or the volume for input. All indicators of a given type correlate each other. I personally trade from time to time manually only for fun but never watch any indicators. (New traders please note: not the indicators predict the market, but the market moves the indicators) When I see a chart, I can freely imagine moving Averages, MACD, ADX, RSI... I simply don't clutter my chart with unnecessary lines (which must not be lines at all, but this is a different topic)

Trading by intuition work only for short time. Depending on the period, if the market doesn't develop in our favor for several bars, we have two choices: to close and to wait for a new entry, or to set SL and to **hope** the market will change.

Anyway, I don't see anything that can be optimized.

To Popov -- here's a suggestion -- allow users to choose one of two approaches to optimization:

Yes, this is interesting and already implemented. We have an option "Use only default numeric parameters' in the FSB Pro Generator. We have also a possibility to set the max % deviation in the optimizer.

By the way, my motives to add the "only default" parameters option was similar to your ideas. I may also add this option to EA Studio Generator.

11 (edited by sleytus 2019-01-09 11:36:57)

Re: Question: Why do we obsess about optimizing our strategies?

Thanks, Popov.  Your insights are always interesting to read and help me adjust my perspective.

I see now the option "Use only default numeric parameters".  Interestingly, I have that checked and, also, "Optimize strategy".  I never paid much attention to these since I always run the optimizer afterwards.  So -- what does it mean if both of those options are checked?

I am familiar with FSB-Pro's Optimizer and its ability to set the Minimum, Maximum and Step for each indicator setting.  However, most of us create collections and it is very tedious to adjust those for every parameter of every indicator of every strategy.  Would it be possible to have a global option to simplify this -- e.g. +/- 5%, +/- 10%, etc -- that would apply to all parameters in all indicators in all strategies that are currently open.

At this point I'm not claiming one way is better than another.  I'm still learning and experimenting.  One thing about Forex that really intrigues me is that as you pay more attention to the details then the data seems to reveal more of itself to you.

Re: Question: Why do we obsess about optimizing our strategies?

Hmm, it was exactly like that. When you select a deviation percent let's say +-15, the FSB Pro optimizer must remember that and to preset the params according to that range. I just checked it now and it doesn't work. It must be a bug introduced in some of the updates. I'll fix it in next releases.

Re: Question: Why do we obsess about optimizing our strategies?

I'm doing a trial of EA studio and I've noticed that if I pick a good strategy from a collection and get a good Monte Carlo result on it and then optimize it and rerun the Monte Carlo test again it often looks worse. So why optimize if the generation of the strategies in the first place is to find a strategy that yields a profit which by definition is optimization. So why run optimization which introduces curve fitting?

Re: Question: Why do we obsess about optimizing our strategies?

So why run optimisation which introduces curve fitting?

The Optimizer is the most used feature of EA Studio after the Generator. The question number one I receive from new users is "how to optimize my own MQL expert". It must be there because every user wants it.

The Optimizer does what it is set to do - to optimise a single stats metric while satisfying the Acceptance Criteria.   

When you set the Optimizer to search best Net Profit, it will try to find a better combination of parameters. However, there is no guarantee that the produced strategy will not be curve fitted.

To reduce the overoptimization, you can try to search the best System Quality number, R-Squared, Balance line deviation or Minimum stagnation.

Monte Carlo or testing on unseen data are means of robustness testing.

I'm curious if someone asks why there is an Optimizer in MetaTrader, NinjaTrader or the other platforms. Or why there is no Monte Carlo testing.

Posts: 14

Pages 1

You must login or register to post a reply

Forex Software → Forex Strategies → Question: Why do we obsess about optimizing our strategies?

Similar topics in this forum