>>>Artificial Intelligence ~ EA Available ~ FREE Daily Signal<<<

Sorry, but I think I don't really understand what you mean

But now that you mention it: Most mondays will be skipped in the future, as the algorithm is now including the day of week in its predictions

Prediction for next AUS200 Session: LONG @ 100% Signal Strength

Result for last trade:-26 pips (-1.10%)

Kudos for posting the live calls. It's a rare thing here!

Some observations and queries;

Your signal strength to win rate seems to have very little correlation.
35% strength yields a great win, 100% strength yields a loss.

Without knowing what your signal is based on, I can only assume it's some kind of indicator/strength index, which would be why you're not using fixed distances. Win 60 pips one day, lose 26 another.
Probably won't answer this (and that's ok) but does the EA take into account any relative wave function (previous extremes) to determine a cut off, in either loss or profit? Ort is it based on whatever you are using as your trigger?
Also, do you determine your risk settings based on signal strength? I notice the 35% strength 60 pips gave you 0.85%, where as today losing 26 pips cost you 1.1%. It appears you risk according to strength, am I right?

Looking at your Myfxbook data, you're running approx 56% win rate so far, with a little over 1:1 average R:R ratio.
Not bad, certainly workable but of course with a straight single entry win or loss system you will inevitably take longer to recover from fixed DD.
Will you take any steps in terms of money management if/when this happens?


One final observation on your theory regarding the ASX200...
You're kind of right, Aus is active when the major indices are closed.
However...in broad terms the AUD tends to follow the Wall St lead. Good day on the Dow often leads to a good day on the ASX.
However, having an economy heavily reliant on China and the commodity price, you still have to be aware of those movements.
Today was a good example.

There was a fairly weak lead from Wall St the night before, so it could be logical to assume a follow on.
However, with the cooling of the iron ore prices after the recent surge, the fairly sharp drop in the Chinese market over the past day or two, and the fact that the ASX is at almost 2 month Highs, I'm not sure what you are using to label the strength at 100%.

Still, well done for putting yourself out there and good luck.

Well, I promised, so I do it At the moment it does not run perfect, but it works perfectly normal haha



The signal-strength and win-rate do correlate over a meaningfull sample of trades (so, in the long run), 1 month of data is just not enough for this.



The EA is not using any kind of indicators, wave functions, or anything else. It is pure price action that is used as inputs to the internal models. Indicators are a human way to look at things, they are used to boil down complexity, but that process always loses data. So, for machine learning - not only in the forex domain, but more general - it is mostly better to give the algorithms enough freedom to find their own "solutions" :-)
And yeah, the results of simulations did confirm that for the concrete application as well.

This might sound counter-intuitive, but the algorithm is not using any kind of fixed targets, and this worked out best in the simulations. Here is an article I found that might explain what I mean http://mathematicalforex.net/?p=283


I am not totally sure if I understand what you mean with the money-management and "single entry win or loss" part. Can you explain?
And regarding the win-rate, it could have been higher if I would have implemented the day-of-week-awareness earlier. As you can see on myfxbook the algorithm is trading worst on monday (because then the signals from the world-economy - the other indices - are 1-2 days old already, and a lot can happen in that time!)



And you are absolutely right with the observations about ASX200. And believe me, its itching me to optimise it with hindsight to the last days.
However, what I have learned in the past is: Stick with a system that works, dont try to optimise it after every loss, that will only lead to something that never works on a broad scale.

So, while you are right, and it would most likely be good to have these observations in the model, it will introduce short-comings in other market conditions for sure. Would it be better overall? Perhaps. I dont know.

In the end, all you can do is stick with what you know is working (or at least should, nothing is sure in trading of course), and be a man when its about losing periods (not my words haha, but words - true words - that were spoken to me ).

Anyway, regarding simulations, even a month-long drawdown would be expected. But drawdowns are normal, so I will just continue to update this thread until the system is back in profit and beyond ;-)

Prediction for next AUS200 Session: SHORT @ 100% Signal Strength

Live account results (with 1 fixed lot) as of March 10th:

a prior point of mine, and the point still is...
if the OP says he found that variable risk is the way to go, should respect that and not mess in fixed lots stat, or for that matter anything else that irrelevant on a very short term statistic ( ie. like selected trading day or such).
if anytime, maybe after a month or more data such things can show us something, in the meantime i think the focus should be still on the system AS IS?

Sure, what I mean by single entry win/loss is that you are taking one signal per day, one direction. No reversing, so secondary entries. So each day is either win or lose. This in itself has ramifications to the equity curve and features of DD.

Money management wise is a little more complex. Without understanding your approach, some examples would be:

Using modelling to determine the approximate max consecutive win/loss rate, and say after "x" consecutive losing days, increase the risk under the assumption that the series is unlikely to continue.
conversely:
After "x" consecutive winning days, decrease the risk under the assumption that the streak is likely to end.

Without understanding your methods there's little point listing more, because likely many will not be applicable, but you get my drift.
If the modelling shows a fairly constant set of certain parameters, there can be opportunities to adjust the risk accordingly, in hope of optimizing the equity curve.

With regard to the more fundamental aspects of the ASX, I'm pretty sure these would be hard to code in, as they require a certain amount of discretion. As you said, at least initially it's best to allow the EA to run some decent test sample size and look at what you have stats wise. Sometimes something will jump out at you.
It may also be that, for example, when there is a major event in the leading markets the night before, you notice a certain repeated pattern with the ASX.
It's possible you can adjust, or even switch off the EA to take into account this feature.


Anyway, I've said enough.
Good luck

Live account results (with 1 fixed lot) as of March 11th:

(The results - with a fixed 1 lot - seem to be about the same as with signal strength adjusted lots. See comparison statistics below.)

Result for last trade:-23 pips (-0.98%)

Why there has been no trade on Monday AM...? (Maybe because of broker time change to GMT+3...?)

Prediction for next AUS200 Session: SHORT @ 45% Signal Strength

@MasterGuru: the AI is now time-sensitive and hardly trades on mondays anymore. And that is a good thing, as it has an 83% lossrate on mondays. Why? Because AUS200 opens before all the other markets, so the signals the AI uses are already 1-2 days old, and a lot could have happened in that time :-)

Result for last trade:+60 pips (+1.29%)

Live account results (with 1 fixed lot) as of March 15th:

Prediction for next AUS200 Session: NEUTRAL - do not trade!

Prediction for next AUS200 Session: NEUTRAL - do not trade!

Wow..neutral for 2 days ! ( i don't mind after Fed )

I know it did not trade for a while - but even if I would prefer more trades that I can show, there is just no way around :-)
The algos want what the algos want.. or so