Crano
I think you already know that with our AI — and with us in general — things are a little more interesting than simply choosing between two options.
It is not just a binary question:
Either this, or that.
Either stops, or no stops.
The search for a balanced solution never really stops.
You solve one series of problems, and eventually the combined system demonstrates a stable positive result.
Then the next important question appears.
Over the last year, we have had just over 10 situations like this — against more than 1,000 trades even in our most conservative mode.
And we deliberately HOLD these situations.
First of all, because we want to synchronize what happens in live trading with our research programs.
Quite simply: compare real market behaviour with what the tests are showing.
We can test hundreds of trades and observe certain behaviour, but the important question is:
Will it behave the same way in the real market?
So far, the tests are showing the same thing.
The current GBPAUD trade is classified as one of these extremely difficult cases.
And the current test runs are seeing the same type of outcome in their simulations.
Which is excellent.
Secondly, these difficult cases form a separate subset.
And that subset has to be physically accumulated and isolated before we can properly study it as a group of events — to determine whether they share common properties, or whether there are different categories hidden inside them.
For example, imagine we discover the following property:
If a trade, after 4,500 minutes, has fallen more than -1,200 points, its recovery becomes significantly slower, and after that point the trade should simply be closed.
At the same time, all similar trades in the historical data eventually recovered.
For example — rounding the times to the nearest minute — we have cases such as:
GBPJPY BUY — 2026-06-22 17:05 UTC
EURJPY BUY — 2026-04-28 09:40 UTC
EURUSD BUY — 2026-03-16 16:40 UTC
CADJPY SELL — 2026-03-10 22:45 UTC
And so on.
There are only 13 such cases out of approximately 1,150 trades according to this criterion.
And every single one of them eventually recovered.
Of course, GBPAUD has not yet been open for 4,500 minutes — roughly three days — but in terms of drawdown, it is already clearly approaching this group.
So we now have this rather impressive collection of difficult trades.
Trades that eventually recovered — but certainly gave us a lot of heat along the way )))
And this is now the final, and essentially the only major question we are working on:
What exactly should we do with these trades?
How can we identify them more accurately and earlier?
How should we cut them?
And, perhaps even more importantly, what should we do after cutting them?
That is what we are currently trying to answer by combining approaches, running tests, and comparing different methodologies.
And honestly, it would be quite a funny story if GBPAUD now follows exactly the same scenario and — whether in 3–4 days or simply over the coming days — eventually recovers as well )
So yes, of course you are right: stops may ultimately be one possible solution for trades like these.
But what is even more important is to look specifically and carefully at this small group of difficult trades, understand what they have in common, and learn how to cut them at the right time — one way or another )
P.S. — One important clarification: there is only ONE lot coming directly from the AI itself )))
Everything else is our own greed — well, more precisely, our confidence… or perhaps overconfidence (we’ll find out) )))
