AI ORASignal - First test results (we are always in profit) | Page 4 | Forex Factory

Options

Go to
Permalinks
Print Thread

Similar Threads

Always profit exit strategy 70 replies

Profit From Profit-Taking —Scalping the Session Close 21 replies

Why EA test is forward test better than back test?? 1 reply

Would you take profit with 500pips profit or? 26 replies

Created more than 10 EA's and always end with 1 profit ratio 4 replies

Attachments: AI ORASignal - First test results (we are always in profit)
Exit Attachments

AI ORASignal - First test results (we are always in profit)

  • Linked: FreeSingal_440_f30 | Brokerage Account XXXXXX45
  • Expand
  • Last Sync: 6 min ago
6 min ago
Last Sync: 6 min ago
6 min ago
Last Sync: 6 min ago
6 min ago
Last Sync: 6 min ago
6 min ago
Last Sync: 6 min ago
to
No trading activity in this period.
  • Linked: Manual trades only (copying ORA) | Brokerage Account XXXXXX61
  • Expand
  • Last Sync: 6 min ago
6 min ago
Last Sync: 6 min ago
6 min ago
Last Sync: 6 min ago
6 min ago
Last Sync: 6 min ago
6 min ago
Last Sync: 6 min ago
to
No trading activity in this period.
  • #61
  • Apr 18, 2026 4:17am Apr 18, 2026 4:17am
Additional note on the histogram.

Trades from the last 5 days are highlighted separately.

Their holding-time distribution reproduces the previously observed structure: a pronounced concentration in the initial intervals (particularly within ~15 minutes), followed by a monotonic decay in frequency.

For us, this is a meaningful observation.
It suggests short-term reproducibility of the statistical properties of the distribution within the current sample.

At the same time, we are fully aware that this observation is limited to a specific time window and sample, and does not in itself imply long-term stationarity or persistence of these characteristics.
Nevertheless, the recurrence of this distributional shape in a stochastic environment is of particular interest, especially in the context of analyzing tail trades that fall outside the main body of the distribution.
Attached Image (click to enlarge)
Click to Enlarge Name: f9.png Size: 13 KB
The future is determined solely by the present. More: ORASignal.com
  • #62
  • Apr 18, 2026 7:05am Apr 18, 2026 7:05am
Dear traders,
a quick clarification about the purpose of our posts.

We are sharing our work here mainly as part of an ongoing research process — observing and documenting how the model behaves on real market data as it develops.

We often get questions about practical use, so just to be clear:
We are not providing trading signals, selling EAs, or offering any automated trading systems.

Everything shared here is purely for research and analytical purposes at this stage of development.

Thanks for following along and being part of the discussion
The future is determined solely by the present. More: ORASignal.com
  • #63
  • Apr 18, 2026 2:14pm Apr 18, 2026 2:14pm
Quoting OraSignal
{quote} Thank you for the discussion — I’ll try to add some perspective based on our current observations. In one of the screenshots (AI mode F9), we presented a histogram showing the distribution of trades by holding time. Looking at it overall, the distribution appears to be significantly left-skewed: the majority of trades are closed within a relatively short time frame — roughly in the 5–60 minute range. At the same time, there is a clearly decreasing pattern, with a strong concentration in the earliest interval (for example, trades under 15...
....At the same time, there is a smaller group of trades that extend beyond this range and remain open for significantly longer.... yes, this is the one that interests to me to know more. It’s the time bomb for most of us and most of the system. The killing part was floating for unknown ending time. It can kill all the small pips ptofit that we have developed.
  • #64
  • Apr 18, 2026 4:51pm Apr 18, 2026 4:51pm
Quoting ruparaw
{quote} ....At the same time, there is a smaller group of trades that extend beyond this range and remain open for significantly longer.... yes, this is the one that interests to me to know more. It’s the time bomb for most of us and most of the system. The killing part was floating for unknown ending time. It can kill all the small pips ptofit that we have developed.
Yes, exactly — and I’d like to clarify one important point.

Situations like these (long-lasting open trades, in any system) are actually even more complex than what you described. Calling them a “time bomb” is understandable — but in reality, they are even more uncertain than that. A bomb has a defined mechanism. An open position, however, can evolve in multiple ways — depending on market context, volatility, and external factors.

In other words, an open trade is a probabilistic object with a wide range of possible outcomes.

Understanding the nature of such trades within a system is a key challenge. The real question is not just what it can do, but what it is more likely to do, and under which conditions less probable scenarios may unfold (for example, sudden news events — like the Iran-related news on Friday that sharply moved the market).

What you described is indeed one of the core problems in trading:
You open a position, you have expectations — but then price moves against you.

How do you determine whether the trade is still aligned with the original hypothesis, or whether the underlying assumption has already failed?
And of course, this matters because the downside risk is asymmetric — in extreme cases, the account can be damaged very quickly.

From our current perspective, we are focusing first on achieving stable win rate behavior.

At the same time, we are researching another important dimension — extending target ranges (TP levels). This effectively gives more time and more informational clarity to evaluate the position and make a better decision: exit or hold.

Personally, I consider win rate to be the primary parameter.
It answers the fundamental question:
are we just guessing — or do we actually have a statistically grounded edge that allows us to be right more often than wrong?
If that condition holds, it unlocks an entirely different set of strategic possibilities.
Because starting from randomness (a ~50/50 win rate) makes the entire problem significantly more difficult.
The future is determined solely by the present. More: ORASignal.com
  • #65
  • Apr 18, 2026 8:28pm Apr 18, 2026 8:28pm
Hope you are not giving up answering my Doubt. May I know those equity fall down due to cut loss or withdrawal?
Attached Image (click to enlarge)
Click to Enlarge Name: IMG_7924.jpeg Size: 56 KB
  • #66
  • Apr 19, 2026 12:10am Apr 19, 2026 12:10am
Quoting ruparaw
Hope you are not giving up answering my Doubt. May I know those equity fall down due to cut loss or withdrawal? {image}
Good question.

Equity drawdowns can be caused by two different factors:
• trading results (losing trades / risk outcomes)
• or external actions such as withdrawals
These should always be considered separately.
In our case, we focus specifically on the trading component — that is, how positions behave and how the system manages outcomes.
Withdrawals (if any) are external to the system and are not treated as part of performance.

That said, it’s important to clarify the main point of what we are doing.
The goal of this experiment is not to present a “final product,” but to demonstrate the evolution of the system in real time, based on observable and verifiable data.
For example, the results of one of the modes (e.g. F30) at the early stage and now are not the same — the system evolves continuously.
If you look at recent periods (for example, the last couple of months), you can see that performance characteristics have improved compared to earlier stages of the experiment.

At the same time, an interesting observation:
the win rate remains relatively stable across different periods.
It does not “jump around” — it stays consistently within its range.
For us, this is an important and objective characteristic of the system.

As for absolute profitability — it simply does not exist in any real system.
There is no such thing as a perfectly smooth, always-growing equity curve.
Profit factor is, by definition, a balance between gains and losses — and improving this balance is one of the key directions of the current research.

The core question we are exploring is quite simple:
Does the evolution of the system improve the quality of this balance over time?
We don’t assume — we observe.
And the only way to answer that is through continued experimentation.

There is also a practical aspect:
If positions are held for hours or days,
it would take years to reach statistically meaningful conclusions.
With shorter-duration trades (as in our current setup),
we can reach statistically relevant observations much faster.

In short:
There is no system with only positive outcomes.
There is no 100% certainty in markets.
But if a system demonstrates a statistically significant edge (for example, consistently above a random baseline), then the real question becomes:
Is that edge sufficient to produce stable, repeatable outcomes over time?


Does having “knowledge” (in a probabilistic sense)
actually translate into a sustainable ability to extract profit?
That is exactly what we are trying to understand

Even simpler (the hypothesis is as follows):

If I “knew” — in a statistical, probabilistic sense (since knowing the future with certainty is impossible) — with a probability realistically achievable in nature (for example >90%), would that be sufficient to generate stable profit over time — or not?
That is exactly what we are testing.
In other words:
Many traders start with some form of “knowing” (a hypothesis or assumption), and then act based on that.
The question is:
can this probabilistic knowledge actually be converted into profit?
The future is determined solely by the present. More: ORASignal.com
  • #67
  • Apr 19, 2026 1:25am Apr 19, 2026 1:25am
Quoting OraSignal
{quote} Good question. Equity drawdowns can be caused by two different factors: • trading results (losing trades / risk outcomes) • or external actions such as withdrawals These should always be considered separately. In our case, we focus specifically on the trading component — that is, how positions behave and how the system manages outcomes. Withdrawals (if any) are external to the system and are not treated as part of performance. That said, it’s important to clarify the main point of what we are doing. The goal of this experiment is not to...
thank you, sir for the very detailed explanation. If I’m understand correctly, did you mean that your system do have a cut loss but it’s not mentioned in your published telegram right?
  • #68
  • Apr 19, 2026 2:33am Apr 19, 2026 2:33am
Quoting ruparaw
{quote} thank you, sir for the very detailed explanation. If I’m understand correctly, did you mean that your system do have a cut loss but it’s not mentioned in your published telegram right?
Not exactly.

Profit factor reflects the balance between total profit and total loss.
To understand this balance properly, we initially avoid imposing strict external constraints on the system.
The signals we publish on Telegram are primarily for demonstration purposes — to show that events actually occurred, in a transparent and verifiable way.
At the current stage, the system itself does not explicitly calculate or enforce stop-loss levels.

This aspect is still under research.

In practice, decisions such as whether to close a position or extend it are currently made manually, based on context.
This approach is important because it allows us to observe the system’s behavior in a relatively “unrestricted” form and obtain cleaner statistical characteristics (for example, as seen in the distribution of trades).

Going forward, we are of course working toward finding a more structured balance.
For example:
• high profit factor with low trade count → unstable
• high win rate with poor expectancy → not sustainable
• positive expectancy with drawdowns → realistic scenario

What we are trying to understand is:
what does “normal reality” look like for this system?
Where is the balance point between these factors?
We believe this is not only relevant to us, but also interesting from a broader research perspective.

At its core, this is a serious statistical question — and it deserves careful observation and open discussion.
The future is determined solely by the present. More: ORASignal.com
  • #69
  • Apr 19, 2026 2:44am Apr 19, 2026 2:44am
Quoting ruparaw
{quote} thank you, sir for the very detailed explanation. If I’m understand correctly, did you mean that your system do have a cut loss but it’s not mentioned in your published telegram right?
Let me add one more point.

At the core, we are trying to understand something very simple:
is it actually possible to make money in the market in a consistent way?

It may sound a bit bold, but if we follow a strict, statistically grounded approach, we do make a working assumption — a hypothesis (or, if you prefer, a form of expectation) — that the market may allow for stable profitability over time.

To put it more precisely:
is it possible to push the probability of failure far enough away from sequences of successful trades, so that losses no longer erode the overall result?

Is that achievable?

This is exactly what we are testing — observing, studying, moving forward, and showing the results as transparently as possible.
The future is determined solely by the present. More: ORASignal.com
  • #70
  • Apr 20, 2026 12:21pm Apr 20, 2026 12:21pm
Earlier we mentioned one — so far the only — “difficult” AI trade on GBPAUD, along with a few additional positions we opened manually in the same direction (longs).

We’d like to share some further thoughts on it.

A bit of context first.
Let’s repeat something important: the physics of the universe does not allow 100% certainty.
Nothing is ever 100%. This is not just a nice phrase — it’s a scientific fact.
So even with a 90–95% win rate, the remaining percentage naturally consists of these “difficult” cases — roughly 1–3 out of every 20 trades.
And these are exactly the trades that interest us the most.

We focus on them because they are the most informative.
Moreover, each “difficult” trade behaves differently — they are not identical cases.
In essence, these are outliers within an otherwise strong statistical flow.

And here we have one such trade that has been “stuck” for more than 3 days.
Which makes it a very interesting case.
A system with a clearly observable win rate (not boasting — just look at the account stats) produces a trade like this.

So the key question is:
Can even these “difficult” trades be fundamentally wrong? Completely misplaced?

Our working hypothesis is: no — they are not purely random failures.

Even when new external factors emerge (strong news flow, unexpected events), they do not fully invalidate the detected structure.
At most, they introduce conflict — competing forces — but not a complete reversal of the underlying signal identified by the AI.

So what do we observe in this specific case?

First — the pair is not behaving like in previous patterns after the entry.
It is hovering around a kind of equilibrium level. Even now, price is still around that zone.

Second — part of the manually opened positions (after the AI trade) has already been closed.

Third — we continue trading in the direction identified by the AI (bullish bias from several days ago).

And overall, we are holding steady, managing the situation, and continuing to generate profits.

There was one downside — the original AI trade hit a stop (not adjusted in time), but those losses have already been offset by profitable trades.

So what is the main point here?
With a system showing a statistically strong win rate (>90%, see account data),
each trade — even a delayed or “difficult” one — can still be viewed as an opportunity, not necessarily a failure.
At least for now, it is very hard to classify even these rare “difficult” trades as outright mistakes.

How this specific case will end — we don’t know yet.
But we will definitely share the outcome.

For now — the experience remains consistently positive
Attached Image (click to enlarge)
Click to Enlarge Name: aud2.png Size: 208 KB
The future is determined solely by the present. More: ORASignal.com
  • #71
  • Apr 21, 2026 2:10am Apr 21, 2026 2:10am
Markets are an incredibly vast and complex domain, giving rise to many different approaches.
Ours is strictly quantitative, based on statistical analysis.

Today, as yesterday, our AI is active (https://nabstat.orasignal.com/), but it is not identifying sufficiently strong conditions to open new positions.
Based on the broader news backdrop, market activity appears relatively low, with participants waiting for geopolitical developments (there are also various interpretations of why conditions look this way).

From our perspective, the system’s behavior in this context appears consistent with its internal logic.

Now a bit more on the GBPAUD trade mentioned earlier.

Price is currently fluctuating around a relatively stable level.

We generally avoid overly abstract interpretations, but if we allow a simplified view, this situation may suggest the following:
Recent news may have shifted short-term boundaries, but may not have fully altered the underlying factors that initially contributed to the long setup identified a few days ago.

At the same time, it is important to recognize that market conditions can evolve in ways that are not always immediately observable.
In this context, it is also useful to consider alternative scenarios.

For example, even previously successful trades, when exposed to different conditions, could have developed differently. Considering such cases helps us better understand system behavior under less favorable conditions.

Looking at the current situation on #EURJPY, one could assume that following the direction identified by the AI may, in some scenarios, allow for recovery after adverse movement.
However, this remains a hypothesis rather than a conclusion.
More broadly, any reasoning about the future is inherently hypothetical and probabilistic.
We treat such interpretations with caution.

What matters to us is that the system demonstrates internally consistent behavior, and that its results can be observed and evaluated over time.

This leads to a key question we are exploring:
Can a system with stable statistical characteristics (see account data) behave unpredictably under certain conditions, or does it tend to remain within a consistent behavioral framework?

At this stage, based on current observations, we have not encountered evidence of completely irrational system behavior.
These are exactly the types of questions we aim to study.

For anyone considering more serious or large-scale trading, understanding how a system behaves under non-ideal conditions becomes especially important.
So far, our observations remain generally constructive, including in more complex situations.

We continue to monitor and analyze
Attached Image(s) (click to enlarge)
Click to Enlarge Name: AUD5.png Size: 180 KB
Click to Enlarge Name: JPY.png Size: 152 KB
The future is determined solely by the present. More: ORASignal.com
  • #72
  • Edited 7:50am Apr 21, 2026 7:39am | Edited 7:50am
The market today is not the easiest to read.

But still — what are these interesting zones the AI keeps finding?

The structures before the trade are clearly one type of pattern.
The structure after the AI places a trade — looks completely different.

The AI’s task is simple in definition: capture a certain number of pips within a given TP range.
And overall, it manages to do that (+manual positions, of course).

What’s also notable — it simply stops trading when, from its perspective, “something isn’t right.”

And honestly, that’s a very good sign.

Looking at the structure after the trade — it really does feel like “something changed.”

More precisely, the market structure and pattern shift quite noticeably

We do plan to scale the AI to larger timeframes, and we expect to see similar behavior there as well (markets are fractal by nature).
Of course, on higher scales the pip ranges will be larger — up to the point where, again, “something stops making sense.”

In other words — these are not random or “naive” trades.

They’re actually quite interesting.

At the very least, they consistently align with certain structural boundaries.
I’ll even make a tentative call — we may see a trend shift here, possibly a strong one (similar to what happened with EURJPY in the earlier examples).

PS
The AI is not a human.
And yet, the level of logic and apparent “intelligence” is genuinely striking.
This isn’t some “talking cats” kind of AI
Attached Image (click to enlarge)
Click to Enlarge Name: CHF112.png Size: 137 KB
The future is determined solely by the present. More: ORASignal.com
  • #73
  • Apr 21, 2026 9:28am Apr 21, 2026 9:28am
Today’s trading.

The market woke up — and so did the AI.

Another great day, with some very clean and well-placed trades

That said, it’s currently trading around historical highs, and unlike the AI, we actually do have something to lose — so we’ll pause it manually until 00:00.

At the same time, everything we expected from the AI was delivered — all target TPs were reached.

There is one losing trade, but that one’s on us… a bit of a “misclick” on the ever “wonderful and user-friendly” MT5 terminal (based on the data, it would have closed in profit as well).

Also worth mentioning — on a separate manual-only account, where we place trades based on signals from Telegram (not all of them of course — no one has the capacity to follow everything, but a good portion), the picture is very similar.

And, there’s a snapshot showing the current situation with the only “difficult” AI trade so far on GBPAUD (more details above).
This is that same “unsuccessful” AI trade, along with our manual additions.
The account is back in overall profit.
We continue to open long positions, following the logic of the initial AI trade.
Day three now


In short, there’s a reasonable hope that a kind of “set it and forget it” system might be achievable — though that’s still just a working hypothesis.

Research comes first above all.

Tomorrow we’ll take another look — specifically whether there’s been further expansion around recent highs on GBPJPY and EURJPY.
There’s a cautious sense that the AI didn’t enter these higher-risk trades without reason
Attached Image(s) (click to enlarge)
Click to Enlarge Name: EUR_D.png Size: 137 KB
Click to Enlarge Name: GBP_d1.png Size: 139 KB
Click to Enlarge Name: GBPAUD_d4.png Size: 250 KB
The future is determined solely by the present. More: ORASignal.com
  • #74
  • Apr 21, 2026 9:57am Apr 21, 2026 9:57am
I really like the structural aspect of what the AI is doing — especially how it aligns with market patterns and adapts to changing conditions.
Just one observation from a practical execution perspective:
Some of the projected target ranges(reffered to your latest telegram input figure) appear quite tight relative to real market spread, especially on pairs like GBPJPY.
For example:
Entry: 215.309
Target: 215.330
Distance: ~2.1 pips
Given that spread alone can often take up a large portion of that range, the effective net movement becomes very limited in real execution.
So while the model seems to consistently “reach targets” in a structural sense, the actual tradability of those targets may vary depending on execution conditions (spread, slippage, timing, etc.).
It might be interesting to explore how the model behaves with slightly wider target ranges or spread-adjusted thresholds — especially if the goal is to bridge the gap between structural accuracy and practical execution.
  • #75
  • Apr 21, 2026 10:54am Apr 21, 2026 10:54am
Quoting ruparaw
I really like the structural aspect of what the AI is doing — especially how it aligns with market patterns and adapts to changing conditions. Just one observation from a practical execution perspective: Some of the projected target ranges(reffered to your latest telegram input figure) appear quite tight relative to real market spread, especially on pairs like GBPJPY. For example: Entry: 215.309 Target: 215.330 Distance: ~2.1 pips Given that spread alone can often take up a large portion of that range, the effective net movement becomes very limited...

Great question — and spoiler: yes, we are already working on exactly this, although it does require additional resources.

First of all, just to clarify — GBPJPY is widely considered one of the most “aggressive” and volatile currency pairs (I assume you already know this, but it’s still worth stating).

As I mentioned earlier, for us right now the priority is speed — in order to build a large, statistically meaningful dataset as quickly as possible.
If we look across all accounts combined (and in total this is already around 10,000 lots traded), the win rate remains consistently above 95%.
This means the system has been tested for more than six months, and there are strong grounds to assume it is genuinely able to “understand” market structure — at least to the extent reflected by its success rate.

The second key point is diversity of instruments.

Understanding one single pair is a narrow task.
So if we allow ourselves to say that the AI “understands the market,” then we must demonstrate it under the hardest possible conditions — across multiple currency pairs, multiple directions, and essentially all types of market environments.

All of this reduces the probability of randomness to an extremely low level. If we frame it in binomial terms, the probability of producing such a sequence today is on the order of 10^50 — meaning that even the lifetime of the universe would not be enough to “randomly guess” this. So it is clearly not randomness.

Where I’m going with this is simple: we cannot selectively choose pairs too carefully, except for obviously irrational or rare edge cases in the overall dataset.
That said, in real-world application, some pairs — like GBPJPY, especially at historical extremes — might indeed be candidates for exclusion or special handling.

Now regarding expansion — absolutely.

That is the entire point.
The baseline is already established.
In fact, moving toward higher time horizons (hours-based analysis) is one of the core research directions. Only on longer horizons can you realistically and physically capture meaningful pip ranges, because on short intervals such movement simply does not exist except in rare spikes.

So the idea is to move toward a predictive layer that gives more time for analysis — allowing better management of exits, better understanding of when to cut losses, or when to hold positions more deliberately within the expected trend — while still maintaining a high win rate.

In simple terms, these rare “difficult” trades will still exist, but there will be more time to evaluate them properly — including whether the target zone is statistically deviating and by how much.

And honestly — if we manage to get this right, it will be something quite significant

Spreads.
At this scale, we use ECN accounts.
It’s also interesting how the trade copying works in practice.
It effectively allows us to slightly extend the execution window by waiting for a more favorable entry.
And if we miss it (in manual execution), we simply skip the trade
The future is determined solely by the present. More: ORASignal.com
  • #76
  • Apr 21, 2026 11:45am Apr 21, 2026 11:45am
Earlier (tooday) we wrote about the first trade of the day on CHFJPY, highlighting the interesting zones identified by the AI.

At that time, I also made a cautious prediction (based on our experience with the system) that the trend would likely reverse quite strongly and return toward the level identified by the AI.

Not only did this actually happen (which, in hindsight, would have helped protect the deposit), but what is even more interesting is that the price is again circling around that same level — the very zone the AI initially marked as interesting.

What exactly are these areas? That’s a very intriguing question in itself.

But the fact remains — even if we look at it retrospectively — that the trades were successful, and the market did not break away into a completely different direction in a way that would label the AI’s initial trade as irrational or meaningless.

It definitely doesn’t look irrational.
A few hours later, the price is still moving around the same level on this pair, repeatedly attempting to continue into a bearish trend, but without a decisive breakout so far.
Attached Image (click to enlarge)
Click to Enlarge Name: CHF2.png Size: 301 KB
The future is determined solely by the present. More: ORASignal.com
  • #77
  • Apr 21, 2026 4:27pm Apr 21, 2026 4:27pm
Quoting OraSignal
{quote} Great question — and spoiler: yes, we are already working on exactly this, although it does require additional resources. First of all, just to clarify — GBPJPY is widely considered one of the most “aggressive” and volatile currency pairs (I assume you already know this, but it’s still worth stating). As I mentioned earlier, for us right now the priority is speed — in order to build a large, statistically meaningful dataset as quickly as possible. If we look across all accounts combined (and in total this is already around 10,000 lots traded),...
Thanks for the detailed explanation — really interesting insights.
Just to clarify one point regarding coverage vs what we actually see:
Is the system currently operating across multiple pairs simultaneously, but only a subset of those trades is being published (for example, today we mainly see GBPJPY)?
In other words:
- Are all instruments actively being traded/processed in the background,
but only selected examples are shared on Telegram?
- Or is the current output already representative of the full set of trades being taken?
I’m asking just to better understand whether what we see publicly reflects the full dataset, or just a filtered view for demonstration purposes.
  • #78
  • Apr 22, 2026 12:11am Apr 22, 2026 12:11am
Quoting ruparaw
{quote} Thanks for the detailed explanation — really interesting insights. Just to clarify one point regarding coverage vs what we actually see: Is the system currently operating across multiple pairs simultaneously, but only a subset of those trades is being published (for example, today we mainly see GBPJPY)? In other words: - Are all instruments actively being traded/processed in the background, but only selected examples are shared on Telegram? - Or is the current output already representative of the full set of trades being taken? I’m asking...

I won’t be able to go into the full details of how the system works, but I can outline the key idea.
All trades that appear in Telegram are обязательно reflected on the account.
In other words, the account represents a specific, isolated trading process.

So it’s basically:
a signal in Telegram → the exact same trade on the account → and then the final aggregated statistics of all those actions.

At the same time, we also add manual trades on the accounts. These do not appear in Telegram — Telegram shows only AI-generated trades.

Now, there’s an interesting point here.
We also have pure AI statistics from publicly available modes (and I won’t even touch non-public ones — that would just be “words,” so we stick to what can be verified).
Initially, we assumed that pure AI trades, at the current stage of evolution, could maintain a win rate above 94–95%, but that the Profit Factor might weaken without manual support.
However, over time, it turns out that even pure AI trades alone are currently showing a Profit Factor in the range of 3–4.
So, in a sense, a fully automated mode has already emerged.
That said, I treat the idea of a completely human-free AI system with caution.
Although, with a stable win rate, it is definitely reasonable to consider.
And this is exactly where the question of “difficult” trades becomes important.
Are these trades completely irrational — flying off in the opposite direction?
If that were the case, then yes, the balance would gradually shift toward losses.
But the key question is: how many such truly negative trades exist?
Enough to outweigh, say, a 15–20 to 1 ratio within a sequence of 20 trades?
The important detail is that all “difficult” trades are different.
So far, what we observe is that these rare “difficult” trades are not severely loss-making, and the even rarer truly bad ones are not strong enough to destroy the accumulated profit.

Of course, this still needs to be validated further — by continuing to trade and by increasing the sample size as much as possible.
That’s exactly why we are currently operating on shorter timeframes — to quickly build a large dataset and observe this balance statistically across hundreds of trades.
The future is determined solely by the present. More: ORASignal.com
  • #79
  • Apr 22, 2026 12:34am Apr 22, 2026 12:34am
And a few additional thoughts in the context of what was said above.
For example, on one of the accounts we have the following data (this is the largest public series and therefore the most representative — and here we are only discussing what anyone can verify):
Closed Trades: 7,987
Open Trades: 0
Long Trades: 4,212
Short Trades: 3,775
Winning Trades: 7,727
Losing Trades: 260
Win Rate: 96.7%
(these figures are copied directly from Forex Factory)

The lot flow is substantial — literally a continuous stream.

At this point, one can cautiously begin to separate two layers: what the AI is doing internally, and how it can be configured and used without needing to fully understand every detail of its decision-making.

In other words — and keeping conclusions very careful — at this level of win rate, combined with such a large volume of trades, mathematics itself starts to gain predictive power. And once that happens, it becomes something we can actually work with.
Meaning: we can begin to study a balance that may have forward-looking (not purely retrospective) value.
How does this translate into real life?
A simple analogy: if a car is loaded in a certain way, engineers can calculate its fuel consumption — and over time, that expected consumption does indeed tend to hold.
In other words, these become estimates you can rely on with a certain margin of error.
Within this analogy, sensitive news events can be seen as bumps or uneven terrain on the road — they may cause fluctuations, but they don’t fundamentally break the model.
What could break it is something more like a deep hole — something extreme and unexpected. But such events are not something you would expect around every corner (although, in the current geopolitical context, for example around Iran, the number of such “holes” may increase — or perhaps they are not holes at all, but just larger bumps; and that itself is also worth studying).

Everything needs to be researched and verified, and of course this requires resources.

But the key point remains: there is clearly something here that can be worked with.
The future is determined solely by the present. More: ORASignal.com
  • #80
  • Apr 22, 2026 2:56am Apr 22, 2026 2:56am
Here is the latest series from the AI + our additional manual trades.

We didn’t even wait for the original AI trade to play out (closed it at a loss), although it would clearly have ended up in profit.

But that comes down to individual approach.
I admit we may be overly cautious — that’s just our attitude toward the process.
On the other hand, if the overall series of trades is already showing a solid positive result (PF > 2.0), then what’s the point of waiting for more?

P.S. — while I was writing this post, the edge limit increased even further
But we’ll stay conservative and avoid unnecessary emotions
Attached Image (click to enlarge)
Click to Enlarge Name: CHF.png Size: 163 KB
The future is determined solely by the present. More: ORASignal.com
1 trader viewing now