Thanks Leon,
I'm encouraged by your action.

Hi
Why Optimizes What with Demo Data?
Ex Ea is optiemiert on Pepperstone when I switch on the Demo Data trades are twice as much the capital curve looks much different.

Out of Sampel 35%
So only with realtime data Optimize the rest is rubbish.
lg Pit

EA 7.1

With some added opening and closing conditions.

Leon,

7.0 looks good for today, I just started testing it in my demo and now it has 370 pips floating profit.  Good work!  Is this EA from Generator or your own?

Still good job Leon even if the EA is aided by FSB's tools.  That's the reason why we are here, to use FSB's capabilities to increase our chances of finding profitable EA.

Without FSB, neither could I have the chance to develop profitable EA.  I can't do without it.

One last comment, why most people will use per unit lot allocation instead of % because, in experimenting with things, we need to reduce or remove the influence of independent variable on your dependent variable.  In other ways, you want to use a consistent unit of measurement to understand and measure how good is your EA.

Using % lot allocation (likewise, Martingale) confuse your results in terms of your understanding whether the profit is the outcome of a good consistent/predictable strategy or the results of having higher % lot allocation to compensate those earlier losses, etc.

As for the use of "winner" feature, in my opinion, I'm using "Winner" features to narrow my search for better predictive EA that has the characteristic of entering during Breakout or able to identify longer/stronger trending entry.  Because if the EA entered in weaker or shorter trend, the "winner" option will be a disadvantage and ended up with multiple losses.  Hence, my rationale is, if I use "winner" feature, I eliminate those EA that doesn't enter during stronger/longer trending period.  And hence, increase my chances of finding better EA suited for Breakout or stronger trend.

Thereafter, when you found and tested out an excellent EA, you can go ahead and use your % lot allocation method in your demo or live trading. Development an EA is like doing an experiment.  Keep your control on each variable so that you understand their influence on one another.

Hi Leon,

First of all, thanks for your words of encouragement.  I'm glad I can be of help.  What seems "common sense" to me due to my training, may not be common sense to others.  Likewise, what seems common sense to Footon, Popov etc may not be common sense to me because I'm not mathematically inclined as them.  That's why diversity brings in different "flavour" into this forum.  I believe we all can contribute one way or another from our own perspectives/training/experiences.

Today, 7.0 (now floating 309 pips), 7.1 (176 pips) and 7.2 (176 pips) entered well.  There again, I'm not too sure of it's closing.

If you don't have time to track their price movement, do consider using MyFxBook to help you track.  When you enter your account details, you will be able to see how well your EA enter and exit (accuracy).

In the past, without the help of MyFxBook, I discard EA that aren't profitable.  However, now before I discard, I try to understand where is the "fault", bad entry or bad exit, or really totally bad theory.

If your EA enters well, be of good cheer, half your battle is won!  You then have to work on finding a good close while trying find ways to further enhance your opening accuracy.

Regards
Hannah

Leon,

7.2 has made 775 pips yesterday.

Ps: I've changed all the 7 series EA to lot size instead of % for ease of tracking their results.  And changed SL to 500 and BE to 70-150 range.

Hi Leon,

Each set of my EA are organised according to their "concept" of trading plan, eg. some EA, such as scalper mode, I use 5 to 15 min time frame's indicator to determine their entry conditions.  And there are some I use 30min and higher time frame indicators to as the determine factor for entry.  As such these EA are numbered and saved like the way I saved your 7.0 series (I use 7 because I want to differentiate yours from mine for easy identification at a glance, other prefix numbers are taken up for my EA, except for 7, 8 and 9 and usually leave these numbers for other people's EA when I help them fine tune it or when I do some "crazy" experiment such as revere trading some time ago).  Eg 60min entry EA, I use the prefix 6 (60.1 H1 entry with Trailing 300, this will be an example of my EA file name so that the moment I export it to MT4, I already have the magic number 601 to use and track it's progress). 

Each time I add some opening conditions I may save the file differently to see how the additional condition influence the EA's behaviour or performance.  Eg, 7.0 and if I change the BE from 150 to 100, instead of overriding the file, I save it as 7.01 etc so that I can compare which is a better option simultaneously.

Once I have a systematic way of numbering my EA, next to try to have all the same series or similar one in each MT4, eg MT4 terminal 1 will be for scalper EA, T2 will be for 1hr EA (all these are still using 1min time chart but different concept for entry).

Each day, I try to take a screen shot (like the one below) of my EAs open positions to help me keep track of how much floating they made and to later compare whether they closed near to their peak or too slow in responding to market changes.  Thereafter, I'm clear which EA need adjustment and which EA's concept is better in opening and closing accuracy. By looking at the screen shot, I can compare all my EA at a glance and see which is best and which has "weaker" entry point.

I'll have a note book to record those good EA, eg, each page is for each MT4 and then I'll have different column for different EA series in the page.  Each time I noticed an EA that performed well, I write it down in the column of the series it belong.  I'll put a side note to the EA whether it close well or not (EO/EC means excellent opening/closing and SC means self closing and SL means it close due to SL feature, TP1000 mean this EA closed at TP1000, ssm means Super Scalper Mode, LR means this EA is able to capture Long Run/Trend etc, I favoured EA that all these characteristics SSM, LR, SC and EO/EC, in order words, this type of EA is able to function as a scalper and a breakout EA according to market situation and have good closing conditions, hence self close with excellent entry/exit accuracy) in this manner, after a while, at a glance, I can differentiate which EA to pick/shortlisted for further testing or live trading.

Lastly, to track and analyse each EA performance, I either manually glance thru' the trade history to see which EA performs well and if I need further analysis, I go to MyFxBook and use the magic number analysis function to examine further the EA's entry and exit accuracy and other statistical performance.

I hope I have painted to you a clear picture.  I won't say that my method is totally systematic, in the past I used excel spread sheet to tabulate their results...too overwhelming to do it for so many EA.

All these EA are based on the same theoretical framework, that's why it's easy for me to reproduce so many of them but with different approach in using various time frame as the entry/exit determining factor.  I can't possible have so many different trading theoretical framework.  It's the same framework that guide me through all these EA development.  That's why this framework is very sound and strong yet robust for me to adapt and try various approach in term of using different time frame determining factor and yet still can produce profitable EA. Now the quest is to find which time frame is more "reactive" and responsive to market changes yet stable to bring in consistent profit.  Do spend more of your time to fine tune and search for a trading concept/framework you can be confident and then try to develop EA based on that concept.  Using the Generator is one approach to narrow your search.  From the way your EA behaviour, I got another idea/approach (using my own indicators) to enhance my scalper mode EA, I was thinking about it last night, and I couldn't sleep, I was so excite to try this new approach but yet feel exhausted already by now from the thousands I've made.  I think I badly need a break. 

Regards
Hannah

Poor thing Popov, you have my utmost admiration.

Actually, develop an EA is easy, but it's the repetitive of manually optimizing the base, signal n smoothing method that is tedious.  I find it a utter "waste" of time if only I've my mini optimizer...I sincere hope you can consider this aspect (one day).  I guess not many people realised that by fine tune these aspects, some of their losing EA can be winner and those that did well can be made better.  And creating an EA would be a breeze for me.

You want to perfect your FSB while I want to perfect my EA.

I wonder what drives us, perfectionism? workaholic? or passion?  Probably a cocktail of these 3.

Thanks Dave, it's such an encouragement when someone finally grasp what I am doing because I was beginning to wonder how can anyone not understand that "optimizing" the base, signal and smoothing period is so essential to fine tune their EA.  I can't imagine that most people don't even bother to tweak these setting at all and hence, a "mini optimizer" isn't important to them. Sometimes, I feel like I'm "ahead" of time or way beyond the crowd that most didn't understand what I'm doing and think I'm making lots of fuss.  It's like some artist whose artwork was not appreciate during their era till they pass away.  Of cos, I don't wish that to happened to me...ha ha ha.

How can anyone ever find any profitable EA to start with if they never even venture into this simple step of fine tuning their EA, unless most people don't develop their own EA but simply rely on FSB Generator or optimizer to "think" and work for them.  As good as it sound, these people will never gain good, solid trading knowledge.  Unless they learnt from the profitable EA created by FSB and use it as their starting point for learning and understanding.

In my opinion and I'm rather confident in this direction, that if we have the "mini optimizer" function, coupled with Generator, we can nail our profitable much easier.  Because Generator will help trader find the appropriate period/parameters and if Generator uses the built in mini optimizer in it's calculation, then that will greatly "sharpen" Generator's ability to find profitable EA much more efficient than the usual FSB Optimizer or Generator by own self, by then the current Optimizer would be redundant (no offence).  Which is better? to Optimize by random selection of parameters/period (current Optimizer) that has is not guided/filtered by (Artificial Intelligence, AI) based on sound trading rules/theory or to optimize (mini optimizer) our own EA with our own parameter/period based on our own consistent trading rules?  Personally, I would never venture into the current optimizer, why bother to reinvent the wheels (by searching for new EAs, that may be theoretically unsound) when I already have a very sound theoretic framework to rely on, now I'm just limited by the lack of "mini optimizer" tools to let the machine does the work/calculation in splits of seconds and I ended up having to manually do it myself, which taken me 15 -30 min for an individual EA, with the mini optimizer, I believe I can get in done less than 5 min.  That's the whole lot of difference for people when they have lots of EA to work with.  Given the mini optimizer, I'm certain I will have nailed and developed my "perfect" EA by now.  Because I would by then be able to put in every opening and closing conditions one by one, effortlessly instead of taking the easy way out of using close and reverse as my closing conditions.  I would also be able to compare, analyse better the effects of interaction between different time period and or time frame while the setting of the base, signal and smoothing period is well taken care of by the mini optimizer.  My energy would be focus on more productive observation and analysis and not wasted in manually optimizing these settings that left me too mentally drain to "play around" with different parameters/LTF settings to discover and find profitable EA.  I wish Popov and his team can "see" what I'm driving at and appreciate the value behind it.  Nevertheless, I also understand, there is a lot on his plate now.  The challenge is to prioritize and weight which efforts will yield the best results that would enhance FSB's performance and ability to help users develop profitable EA.

Just like the 1 min time chart, not many understand or appreciate the value behind using it...now more and more people are beginning to catch it.  Maybe one day, people will understand about the PBV issue that I'm so adamant about.  I suggested to Popov, instead of using PBV, why not use PTV (Previous Time Value).  Instead of using Bar (eg H1 or H4) as the unit of measurement, use time setting which users can determine how long, eg a cross over need to "happened" to confirm the condition instead of waiting for the whole bar (which is also make up of fixed time of the chart).  I suspect the reason why PBV is necessary because most people use the conventional period setting that are not sensitive to market changes and hence, you ended up having two curves intersecting each other several times throughout the 1hr or 4hr Bar time period and without PBV, the EA will ended up with multiple "unwanted" or "false" signal.  But this issue can be overcome (of cos in the first place, by using the "correct" parameters) using shift and repeat features to eliminate such "false" and unwanted signal.  I think it's all a matter of understanding and overcoming the odds instead of using a blanket rule that restrict other traders (becos of PBV) who understand how to use other features to prevent those issues.  Nevertheless, this PBV is no issue at all to those who use 1 min time chart (in fact, very essential to have in 1 min time chart) and hence, no big deal to me, if nothing changed.  That's why I rest my case, no point "fighting" for those who doesn't see the value of what I'm fighting for.

Footon, I didn't request for shift and repeat features.  I'm saying with these features we can use it as an alternative in the absence of PBV.  Furthermore, PBV issue is not about backtest accuracy, it's about speed of execution.  Anyway, I've already rest my case.  Let's us agree to disagree, cos we definitely are missing each other's perspective and can't see eye to eye.