Artificial Intelligence

Hi traders,

Are there any wizzkids among us who’re already using AI in their forex trading and who can give us some insight in this? I think it may be interesting for i.e. analyzing currency strength or economic data, or generally in fundamental analysis, but not really familiar yet at this point with AI and it’s possibilities.

Cheers, Peter.

Yes, I’m already convinced AI can make a lot of logical deductions about FA, economics, interest rates etc. but I’d like to understand what AI does with TA.

AI, better known as algos attempt to be a step ahead of humans

Thing is they use what humans have used from day one in the market - guys folding little cardboard price sheets became pivot points - a weird math from over a thousand years back became Fibs, and a Fib that’s not a Fib became part of that because it’s easy to eyeball (50%)

Then the intelligence part kicks in 21st century - again be ahead

MrTopStep tweeted recently (he’s an old timer still in the game)

I wanted to buy a lower but the banks pushed the ES up to higher on the day… what rookie move did I do? Sold the ES and NQ into the pop but kept stopping myself out.

That’s AI - smarter but not really intelligent :slight_smile:

Edit: or is the other way round?

I have no doubt that AI will be developed and refined to increase a winning trade percentage than most traders would not reach.

What I believe is that once any change in market sentiment is immediately assessed by an AI, that will be the day when retail FX trading will be left behind - as it is now beginning.

@peterma I’m more looking for ways in which AI can make our life easier while doing your analysis, especially on fundamentals/sentiment. So I’m not necessarily looking for algorithms or automatic trading, but more in ways to quickly analyze economic data, gauge market sentiment, analyze currency strength, etc… Now we need to evaluate data ourselves, but I’m sure AI can do this job for us, or will be able to do soon.

@steve369 Exactly, as well in analyzing your current performance and getting advise from AI to optimize your trading, i.e.optimization of your stoploss, TP’s, etc… Heard anything about development in this sphere? Think it’s a missed chance not to utilize this or at least be aware it exists.

I created an AI bot that trades the DAX on technical indicators. It had decent results in back and forward tests. I wanted to let it loose with a real account, but it was loads of work to migrate it to MT5, and I don’t have time.

As for what it does, I don’t know. You pass it as many inputs and indicators as you like and it decides what to do with them, then keeps what you tell it were good decisions


A secret strategy - secret even from the owner. Sounds like a literal magic black box.

1 Like

That’s the way AI works. You can put rules in place, but it does the learning bit itself and you just give it feedback on whether what it did was good or bad. In this case, telling it that profit is good and bigger profits are better. Losses are bad.

Things like Chat GPT are obviously much more sophisticated. I’d love to do more with AI and analysis of charts with better models, but it’s too time consuming with young kids and demanding job. Maybe one for retirement…


Surprisingly, I haven’t yet come across a trend trader here who focuses on the T/P as being the most critical part of an order, yet S/L is always attended to first of all. It seems like RRR is good enough, but it’s pure guesswork and should be in a garbage bin.

An AI would be most useful in analysing where to place a T/P with a positive probability of being in the right side of a winning trade. As it happens I do this manually for every trade. .

1 Like

Hi Steve,

If I may ask, how do you manually optimize your TP? I’m currently back to using a 1:1.5 RR as standard for all trades as I experienced that moving around the TP is giving me inconsistent results and induces greed in some occasions. Nevertheless I do see that often I could have gotten more profit out of my trades.

I personally do prefer smaller profits over a higher win rate as I hate losing :sweat_smile:.
I’ve also experimented with 2 TP’s, one at a 1:1 and the other one by moving SL, but I found it more complicated and still rather hard to get to a 1:1.5 average by using 2 TP’s, easier it seemed to just set 1 target at 1:1.5.

I’m really interested to find ways how to optimize both SL/TP’s in an efficient way.

Yeah, that’s pure guesswork and really not at all helpful in finding the right T/P place.
My trend based strategy uses PSAR 0.09 - 0.50 on every chart

These little balls are very useful in showing a price movement on the chart, and where previous support and resistance zones, however minor, were reached by price action. This is key.

I also find that downward trends are more likely to be consistent than those going up. At least, I can match the candle’s lowest price point previously, and place my T/P one ball above where the lowest price fizzled out. On the assumption that the price movement would test the lowest point again, my T/P has a positive probability of being successful.

Then I would add a S/L at a reasonable distance where the balls last hit the zones on the price movement back up. Usually I make sure that the difference between profit and loss is more or less equal. IMO, the only need of a S/L is to protect my account if a sudden retracement and spike occurs.

As it happened yesterday, my CAD/CHF selling trade worked exactly as I envisaged. A $30 profit.

Hope that helps.

Hi Steve,

Interesting, so in general you’re aiming near the last swing low/high as your TP? I’ve been doing that as well but saw my winning rate suffers with that. I’m still collecting data and registering the TP/SL on all my live trades and review after the trade if they could be optimized in any way.


Sorry, I failed to explain that a whole TA takes place prior to zoning in on a probability trend trade. And it doesn’t end there, as I look for confirmation that I would be on the right side of a trade. Several Apps, including Heat Map on this site, Trading View platform, Currency strength, Buyer & Seller numbers, and pivot point tables included.

I use:
Ichimoku Cloud that is followed EXACTLY at a 8-22-44 setting which caters for a five day trading session, not six as was first used by the Japanese trading week. Buying trades above the Cloud and selling below. The Daily, 4hr and 1 hr must be in accord with the Daily trend, otherwise end and move on to another chart.

MACD at 3-10-16 setting. The histogram bars monitors the strength of the price movement for each candle. Good for entry and likely exit points.

RSI 10 with a 50 line only. Above is buy, below is sell.

ATR 10 for each T/F. So as not to exceed Time frame movement when placing T/P.

As you could imagine I am fortunate to find a daily trade that meets all my criteria, and even then, I could get spiked out. I use 0.2 lot size order, which translates to around $30 movement either way once hitting the right PSAR balls. I would rather be conservative than seeking out a max profit.

I can leave this overnight and not have to worry about emotional challenges along the way.
55% winning trades over the last year.

Hope that helps.

It seems to me that AI nowadays isn’t perfect yet. Well, it may predict some data and calculate something, but from the point of TA it’s completely useless.
Well, we have expert advisors which are considered to be AI as well. Whether they help traders to bring profits? According to my observations from time to time. I believe the process of teaching AI to be a good fellow partner for trader may requires some more years.

Í am developing trading bots using machine learning as a hobby, mostly based on technical indicators (just EMA and RSI set to a bunch of different time-frames), some historical candlesticks and a few other simple bits and bobs (general market direction, ‘price roundness’, that sort of thing).

It performs reasonably well in forward testing (close to 100% profit in 300 days at about 65% of winning trades out of about 250).
It seems to work out in the real world as well the last month, trading actual money. Actually significantly better than the forward testing results, but the markets it has been trading in mostly went up anyway, so it’s difficult to tell for sure.
It definitely outperforms me as a beginner trader, but that may say more about my trading prowess than anything :upside_down_face:.

It doesn’t try to make absolute price predictions per se (tried that before but that was far too unreliable), but it’s just trained on predicting entry/exit points. The bot just makes trading decisions based on probabilities that an entry/exit will be profitable or not in the near future (no matter how much profit it may be).
As I see it, my main goal is not to make such a bot the best trader in the world, it just has to be reliable enough and consider many more markets at once than I have time for.

I utilise AI regularly to make a judgment call on my Metal trades (Gold, Silver, Platinum and Palladium).
I ask research questions to ChatGPT, Bard and Claude to gain an insight to areas I could not summarise in the short time it takes AI.