I would like to ask: did anyone check if it is a good advice to throw correlated strategies after backtesting.
Let's say after backtesting we have 3 correlated strategies, A B and C. On backtesting they are similar, but in live B is the best of them. The problem is that B was created after A, therefore it was thrown away by the software, and we lost the best strategy.

I will be glad to hear your opinion.

Avraham Shenkar

1) The starting point
The starting point was simple.
Like everyone, starting in May/June 2024, I started generating EAs like hell.

But very quickly, the first real question came up:
How do you consistently produce, validate, and deploy multiple strategies at scale — without losing control?

2) Generation — creating the raw material
Everything starts with generation in EA Studio Reactor.
•    building strategies
•    defining rules and settings
•    producing a large and diverse pool of EAs

At this stage, the objective is not perfection.
It is: breadth and diversity

3) Incubation — the first real filter
The first obvious step was the installation of multiple demo accounts, which quickly turned into what we now call the incubation phase.
•    running many EAs
•    letting them accumulate trades
•    observing real behavior over time

The goal here is simple: exposure to real data and initial validation
No shortcuts.
No assumptions.
Just letting strategies run.
Over time, this scaled.

Today the environment is:

•    ~30 MT4/MT5 instances
•    ~1,000 EAs running
•    ~86,000 cumulative trades
•    ~9,000 trades per month

Every time I explain this setup, the reaction is usually the same: “You’re crazy.” Maybe.

But without this, I could not have learned what I know today.

At that scale, something important happens: you stop thinking in terms of individual strategies and start thinking in terms of systems

This is what the incubation layer looks like today:

4) The problem that emerged

As the number of strategies increased, a clear problem appeared:

•    too many EAs
•    too much data
•    discretionary top selections were not really working live
•    no consistent way to evaluate them

This is where the need became obvious: control and structure

5) Building the workflow — FxBlue + AI

To bring structure into the process, we needed two elements:

•    a reliable data source
•    a consistent way to process it

Then FxBlue became the connector.

It provides:

•    unified tracking across accounts
•    consistent trade data
•    a stable base for analysis

On top of that, we built the workflow using AI. Not to predict markets, but to:

•    process large volumes of data
•    apply deterministic rules
•    standardize classification
•    generate repeatable outputs

This allowed us to move from:

•    manual observation

to:

a structured, reproducible workflow

6) FxBlue Workflow — bringing order

With FxBlue as data source and the workflow on top (upstream), strategies are no longer just running.
They are: continuously evaluated and classified based on objective rules

Over time, each strategy moves through defined states:

•    Ongoing Incubation
•    Promotion Watchlist
•    Ready for Live
•    Pruning Box
•    Earth Birds

This transforms incubation from:

•    a collection of EAs

into:

a controlled pipeline

FxBlue governance snapshot — distribution of strategies across the pipeline

From ~1,000 running strategies, only ~40 reach the Top Band and are considered for Masters.

7) Masters — structuring what survived

When strategies reach the Top Band, they are not used directly. They are combined into Masters within Quant Analyzer.

Masters are: structured portfolios of validated EA Studio strategies

Built to balance:

•    size
•    symbols
•    assets
•    equity behavior

The objective is not to find the best combination.

It is:

to verify that validated strategies remain stable once combined into a portfolio structure

This is where individual strategies become a portfolio — return and risk combined

8) Master Governance — keeping the structure clean

Once Masters (Darwinex demo accounts) are running, a second layer (downstream) of control is applied.

This layer focuses on:

•    removing clear failures
•    monitoring degradation
•    tracking inactivity
•    maintaining structure over time

The goal is not to re-evaluate everything again.

It is:

to keep the signal pool clean and controlled

9) Moving from demo to real

Moving EAs directly from demo to real accounts does not work reliably. The fallback risk to poor performance is simply too high. So we moved to a different approach: signal copy trading.

We do not reinstall EAs on real accounts:

•    we open real accounts
•    we select the best Masters of the month
•    we copy trades from Masters into real accounts

The goal is not to overcomplicate the structure.

It is:

to preserve performance by keeping execution in the same environment where it was validated

10) Where we are today

After:
•    generation
•    incubation
•    building the workflow
•    classifying strategies
•    structuring Masters
•    validating signals

we now have: a structured and continuously filtered set of strategies

11) Next step — deployment

Tomorrow we move to the next phase → Launch of 2 new DARWINEX Zero portfolios

These will:

•    select validated Masters and copy their trades
•    understand how Darwinex risk calibration behaves on our structure
•    apply portfolio construction rules
•    target seed allocation (~2 months) and investor capital (~6 months)

Final thought

This is not about a single EA. It is about building a process that:

•    generates strategies
•    accumulates data
•    creates structure
•    filters signals
•    and only then deploys them

We might still be wrong. But one thing became clear:

the focus should not be on the single strategy, but on the system that manages and scales them

Of course, we’ll also share what happens with the new DARWINs.

P.S.: Last but not least, all of this was possible thanks to a team of five traders working together toward a shared goal. Thanks to this forum, two more members will join us next week. I can’t wait to have them on board.

Vincenzo

Here, for good readers only :-) — meaning time and patience required — are some interesting readings in the context of trading strategy selection, statistical relevance, correlations, and a few other related topics.

They might be interesting for anyone who likes to look beyond single backtests and think more deeply about how trading strategies should be evaluated.

• The Adaptive Markets Hypothesis (Andrew Lo).
https://www.researchgate.net/publicatio … erspective

Concept: Explains why trading edges are perishable. It justifies our high-turnover replacement model and the 3-month median lifespan of promoted strategies.

• ...and the Cross-Section of Expected Returns (Harvey, Liu, & Zhu).
https://www.researchgate.net/publicatio … ed_Returns

Concept: Addresses the "Data Mining" problem. It supports our 6-KPI "Filter Stack" as a necessary barrier against the thousands of "lucky" backtests generated by automated tools.

• The Deflated Sharpe Ratio (Marcos López de Prado)
https://www.researchgate.net/publicatio … -Normality

Concept: Corrects for "Selection Bias." It validates our rigorous incubation process before capital allocation.

• Does Academic Research Destroy Predictability? (McLean & Pontiff)
https://www.semanticscholar.org/paper/2 … d0a2582415

Concept: Proves that alpha decays post-discovery. This supports our "Strict Thresholds" policy—treating mature strategies with the same skepticism as new ones.

• Portfolio Selection (Harry Markowitz)
https://www.researchgate.net/publicatio … _Selection

Concept: The mathematical root of our 0.80 Correlation Cap. It proves that portfolio risk is driven more by the link between strategies than by individual performance.

• Honey, I Shrunk the Covariance Matrix (Ledoit & Wolf)
https://www.researchgate.net/publicatio … _Selection

Concept: Addresses the instability of live correlations. It is the theoretical backbone for our dynamic monitoring of "Regime Lock" in the Master accounts.

• Stepwise Multiple Testing (Romano & Wolf)
https://ideas.repec.org/p/bge/wpaper/17.html

Concept: (Stable Institutional Mirror) Supports our Hysteresis Loops (PwL/OgI) by proving that consistent performance across multiple stages is the only reliable signal of a lasting edge.

• Sequential Analysis (Abraham Wald)
https://projecteuclid.org/journals/anna … 30491.full

Concept: (Project Euclid Stable Link) Validates our rule by providing a framework for making continuous "Keep/Prune" decisions as new data arrives.

• The 7 Sins of Quantitative Investing (López de Prado)
https://portfoliooptimizationbook.com/b … -sins.html

Concept: A roadmap for avoiding common pitfalls like survivorship and look-ahead bias, supporting our "Lifetime String" data fusion.

• Crowded Trades and Tail Risk (Pedersen & Stein)
https://www.researchgate.net/publicatio … _Tail_Risk

Concept: Discusses how crowded strategies lead to liquidity holes. It justifies our 10% Global Kill-Switch as a safeguard against systemic events.

Enjoy
Vincenzo

I noticed recently that the number of strategies added to my collection seems to vary over time and I just observed it in real-time this morning.

For example, about an hour ago the collection counts across 5 Chrome instances were:
1, 5, 5, 3, 5

and now an hour later they have updated to:
1, 5, 5, 3, 4

I'm aware that in a given Chrome instance, the number of Collection strategies will be lower than the Ascended count due to the correlation filtering. This is all fine.

My concern is that I don't understand how a collection can grow and then shrink like this?

Is this a bug?
Happy to send any details you need.
Thank you

I would like to optimize for other things like the default ones in the optimizers. Is it doable to ajust the code for that?
An example would be to optimize for a specific win rate percentage ond only that. Or add other custom optimize goals.

is the Generate Best For and Optimize best for Goal accessable for a advanced user to code custom requirement?

how would one have to proceed to generate lots  of eas with WinLoss 30% or 50% hit rate. without having to look in the last pages of the collection after generated them?

I’d like to open a practical discussion about AI in the context of EA Studio.

Not the usual idea that AI will somehow generate profitable strategies automatically. I think most people here already know that this is not the real point, YET!

What interests me much more is something simpler and more practical:

Where can AI actually improve the workflow of people developing EAs with EA Studio?

From my experience, the real value is not in replacing EA Studio, but in helping us manage scale, improve decisions, reduce noise, and better understand what is really happening inside a large EA workflow.

Here are some use cases that I think are genuinely useful. Some of them we are already using in practice.

Strategy & EA understanding
    •    Clustering similar EAs*
Group strategies that are basically variations of the same idea, so we do not think we are diversified when in reality we are not.
    •    Winners vs losers analysis*
Compare profitable vs unprofitable EAs to understand what really separates them: logic type, trade frequency, exit structure, SL/TP profile, market regime fit, and so on.
    •    Feature extraction from strategies
Detect recurring patterns like trend-following, mean reversion, breakout behavior, volatility sensitivity, session dependency, and other structural characteristics.
    •    MQL code analysis*
Review and compare EA logic directly from the code. This can be very useful for debugging, understanding third-party EAs, or checking whether two bots that look different are actually doing something very similar.

Trade-level analysis
    •    Trade distribution analysis*
Study how trades are distributed across time, duration, sessions, weekdays, symbols, and setups.
    •    Winners vs losers at trade level*
Analyze what losing trades look like compared with winning trades: duration, volatility context, time of day, adverse excursion, favorable excursion, exit behavior, etc.
    •    Entry and exit behavior analysis*
Understand whether the edge is really in the entry, in the exit, or in the trade management.
    •    Floating drawdown and recovery analysis*
Look at how trades go into negative territory, how deep they go, how often they recover, and what kind of floating pressure an EA creates before closing.
    •    Trade sequence analysis*
Evaluate losing streaks, recovery sequences, and whether deterioration starts appearing first at trade level before it becomes obvious at EA level.

Incubation & live monitoring
    •    Incubator monitoring*
Detect which EAs are improving, stagnating, or deteriorating over time.
    •    Automatic labeling / classification*
Tag EAs into practical buckets like promising, watchlist, pruning, or ready for promotion based on how performance evolves.
    •    Early warning signals
Spot when an EA starts behaving differently from expectations before the damage becomes too large.
    •    Performance drift detection
Identify when live or demo behavior starts drifting away from the original profile.

Portfolio construction
    •    Diversification support*
Help build portfolios with lower correlation across symbols, logic types, and timeframes.
    •    Role classification*
Identify which EAs behave mainly as profit engines, drawdown stabilizers, or hybrids / bridge strategies.
    •    Exposure mapping
Detect hidden concentration, for example several different EAs all leaning on the same currency or market behavior.

Workflow validation
    •    Process validation at scale*
Check whether the generation + filtering workflow is actually producing better candidates over time, not just more output.
    •    Monte Carlo / WFA interpretation
Summarize robustness results across many strategies when the volume becomes too high for manual review.
    •    Success rate tracking*
Measure how many selected EAs actually survive incubation and become usable.

Operations & scaling
    •    Documentation and tagging*
Keep structure and memory across many EAs, tests, and incubators.
    •    Experiment design support*
Help organize structured tests, for example grid vs no-grid, different parameter families, or broker comparisons.
    •    Reporting and dashboards*
Produce clear summaries of what is happening across the whole workflow.
    •    Log analysis
Detect technical issues, broker execution differences, VPS instability, or unusual behavior in platform logs.

Curious to hear from others:
    •    Are you already using AI in your EA workflow?
    •    Where does it help the most?
    •    Have you found use cases that really improve results, and not only save time?

My personal view: AI is not yet the edge. The edge is still the workflow. But AI can make a good workflow significantly stronger.

* The starred use cases are things we are already actively using in our workflow.

Vincenzo

I’m trying to define a Longevity Index for EAs generated with EA Studio, based on how long strategies remain in a “top band” according to your selection criteria, filters, or KPIs once they are promoted to live trading.

The idea is not to change the selection rules, but to better understand persistence and durability after promotion.

Has anyone here worked on something similar?

A few practical questions:
    1.    Do you track how long an EA keeps meeting your “good” criteria after going live?
    2.    Do you see a real difference in longevity between stronger entries and more borderline ones?
    3.    Do you evaluate this with fixed horizons like 1 month, 3 months, 6 months, etc.?
    4.    How do you treat EAs that drop below the threshold and later recover?
    5.    If you already measure this, what is your typical expectation in terms of months?

I’d be interested in practical experience, constructive exchange, proven best practices…not only theory.

Happy Easter everyone — and thanks in advance for any insights.

Vincenzo

Is that also in FSB-Pro?

Download files:
https://drive.google.com/drive/folders/ … sp=sharing

FSB News Companion indicator for MT4 (FSBNewsCompanion_fix5.mq4)

I tried to implement something similar for FSB Pro using the same news-feed.txt data source as used by Expert Advisor Studio .

The solution is using 3 components:

1. Downloader = data source updater for MT4
2. Gate = strategy execution filter on FSB Pro and exported strategies for MT4
3. Companion = chart UI / situational awareness for MT4 (optional)

What makes this architecture especially clean is that the three components are tightly related, but loosely coupled. They share the same local source news-feed.txt and the same health globals, but they do not overlap responsibilities. The feed is downloaded and served locally for other strategies or charts for manual trading, which makes this very lightweight.

The FSB news ecosystem achieves by splitting responsibilities into three focused components: a Downloader that maintains the local data source, a Companion that visualizes upcoming events on MT4 charts, and a Gate that applies the actual trade-permission logic inside exported strategies. The result is a workflow that is both practical and lightweight: Downloader = data source updater for MT4, Companion = chart UI / situational awareness for MT4, Gate = strategy execution filter on FSB Pro and exported strategies for MT4.

1) FSB News Downloader = feed maintenance service (FSBNewsDownloaderEAv2_2_fix1.mq4)

Uses only one FSB News Downloader EA for single MT4 instance.
Default feed refresh value is 360 minutes, do not make this smaller.

The Downloader is the foundation of the entire stack. Its role is not to decide trades and not to decorate charts. Its role is simply to keep the feed current, healthy, and locally available for the rest of the system. In the FSBNewsDownloaderEAv2_2 line, the EA downloads news-feed.txt, saves it into the MT4 files area.

Need to edit MT4 settings - allow WebRequests to https://forexsb.com

FSB News Downloader EA for MT4

2) Gate = strategy execution filter for FSB Pro and MT4 (FSBNewsGate_v2_1.cs and FSBNewsGate_v2_1.mqh)

FSBNewsGate_v2_1 was designed as a directionless LIVE MT4 news permission gate. On the FSB Pro side (.cs), the indicator keeps a clean UI and parameter contract, but always outputs allow=1 during FSB Pro runtime because the indicator is forward-looking and depends on live feed behavior. On the MT4 side (.mqh), the real blocking logic happens: it reads the offline news-feed.txt, applies symbol/currency matching, applies the configured news-priority mode and pre/post windows, and decides whether long and short entries are currently allowed.

That distinction is crucial. The Gate is not a downloader and not a visual aid. It is a strategy execution filter. It is the final policy layer that tells the strategy whether to open trades around upcoming news. In v2.1, the design was improved further by mirroring the priority mode into an MT4-visible numeric parameter so operators can choose Disabled / High only / High and Medium directly from MT4 after export. That keeps FSB Pro and MT4 aligned while still preserving the execution responsibility inside the Gate itself.

FSB NewsGate indicator for FSB Pro and MT4

3) Companion = chart UI / situational awareness (FSBNewsCompanion_fix5.mq4) (Optional)

The Companion exists because having a healthy feed is not the same as seeing the market context. FSBNewsCompanion_fix5.mq4 is deliberately chart-focused: it reads the local news-feed.txt, filters events to the current symbol, optionally includes USD on all charts, limits visibility to M1-H1 by default, and draws future events as vertical dotted markers with human-readable tooltips. It does not download anything and it does not make trade decisions. It is a visibility layer a situational-awareness tool for MT4 charts.

FSB News Companion indicator for MT4 (FSBNewsCompanion_fix5.mq4)

Download files:
https://drive.google.com/drive/folders/ … sp=sharing

I made various updates and upgrades to the forexsb.com server during the last 10 days.

The upgrade includes updates to the PHP versions, the proxy servers, and the backend frameworks of the following applications:
- the Custom Indicators Repository for Forex Strategy Builder Professional.
- the Wiki software with the applications' User Guides
- the forum
- a backend control panel for managing the users' accounts.

All updates finished successfully with minimal service interruption of 20 seconds.
There are no changes in the behaviour of the updated applications.

The forum's mailing system has also been updated. I wanted to be sure our emails are meeting all security recommendations and are whitelisted by the major validation services.

I invest a lot of time and effort to provide the best service in the industry.

As a result, we have a 12-CPU-core server with 0% load. (With 15 applications running)

EA Studio loads for 300 milliseconds without cache. (And without any errors.)

The next and best strategy builder loads in the browser without internet !!! (Pure magic)

I'm happy with these results and return to the usual application development.

Have a great trading session!

Further it would be interesting to built a Portfolio out of the WF equidity lines and of course multisymbols ;) But that is easy work in excel, but the journal would help a lot.

Looking forward to hearing from.

Open Long when Macd Line crosses the zero line upward
Close Long when Macd Line crosses the zero line downward

but it has not closed, I had to close the position manually:

PLease find EA studio EA attached.

Thanks

I was curious whether swaps change frequently and whether I need to use "real swaps" in my newest and greatest backtester.

I made a simple script to collect swaps from MetaTrader 5 for each day. It stores the swaps in a local database for later use.

I checked it today and noticed the swaps had already changed twice over the last two weeks.

I'll try to find historical information for the swaps and will use it in my next program.

Trade Safe!

Hi everyone,

I've been using EA Studio and Express Generator for years now, and after a long journey of trial and error — testing different robustness approaches, losing money to overfitted strategies, and slowly figuring out what actually works — I've finally built a fully automated validation pipeline that I'm genuinely happy with.

I wanted to share it with the community because this forum and Mr. Popov's continuous work on Express Generator have been instrumental in getting me here. The speed, the scriptability, the constant updates — Express Generator is truly a lovely system, and I owe a big thank you to Mr. Popov for making it all possible. Every new version brings something useful, and the tool just keeps getting better.

So here's my complete workflow. I hope it helps some of you avoid the mistakes I made along the way.

---

The Problem: Overfitting is the Silent Killer

We all know the story. You generate a beautiful strategy with a perfect equity curve, put it on a demo or live account, and within weeks, it's underwater. The strategy was curve-fitted to historical data and had no real edge.

I spent a long time trying simple IS/OOS splits and basic Monte Carlo tests, but strategies kept failing in live trading. The turning point was realizing that one robustness test is not enough. You need multiple independent tests that challenge the strategy from completely different angles. If a strategy survives all of them, the probability that it's genuinely capturing a market pattern goes up dramatically.

---

The 8-Phase Pipeline Overview

Here's the full pipeline. Each phase tests a different dimension of robustness:

+---+---------------------------------+----------+---------+----------------------------------------------+
| # | Test                            |Data Range| Server  | Purpose                                      |
+---+---------------------------------+----------+---------+----------------------------------------------+
| 1 | Generate In-Sample              |  0%– 50% | Premium | Strategy discovery                           |
| 2 | Out-of-Sample Validation        | 50%– 80% | Premium | Overfit filter                               |
| 3 | Monte Carlo — Execution Stress  |  0%– 80% | Premium | Survives real-world execution?               |
| 4 | Monte Carlo — Param Sensitivity |  0%– 80% | Premium | Overfit to specific indicator values?        |
| 5 | Multi-Timeframe Validation      |  0%–100% | Premium | Captures a real pattern, not timeframe noise?|
| 6 | Multi-Instrument Validation     |  0%–100% | Premium | Works on related markets?                    |
| 7 | Cross-Broker Validation         |  0%– 80% | Eightcap| Real pattern or data artifact?               |
| 8 | Forward Test                    | 80%–100% | Premium | Performs on completely unseen data?          |
+---+---------------------------------+----------+---------+----------------------------------------------+

The key insight: Phases 1–4 use 80% of the data maximum, reserving 20% that the strategy has NEVER seen for the final forward test. Phases 5 and 6 use 100% data because they test on different timeframes/instruments entirely — the strategy has never been optimized on those markets.

---

Phase 1: Generate In-Sample (0%–50%)

Generation uses only the first half of the available data. This is deliberate — it gives us three more independent data segments to validate against.

I typically run this for 24 hours per instrument using `max_working_minutes = 1440`. The more strategies generated, the better the chances of finding genuinely robust ones.

---

Phase 2: Out-of-Sample Validation (50%–80%)

Feed the Phase 1 collection into Express Generator with data_start_percent = 50 and data_end_percent = 80. No generation, pure validation.

This 30% data window is completely unseen during generation. Strategies that were curve-fitted to the first 50% will typically fail here. This is your first major filter.

Settings are slightly relaxed compared to Phase 1 because the data window is smaller.

---

Phase 3: Monte Carlo — Execution Stress (0%–80%)

This is where I learned an important lesson. I used to combine all Monte Carlo tests into one phase, and it was eliminating 99.8% of strategies. Separating execution stress from parameter sensitivity into distinct phases was a game-changer — it gives you much better diagnostics about WHY a strategy failed.

Phase 3 tests execution robustness only — what happens when real-world conditions degrade?

---

Phase 4: Monte Carlo — Parameter Sensitivity (0%–80%)

Now test the opposite dimension: are the indicator parameters fragile?

Criteria are more lenient here than in Phase 3:

Why more lenient? Because parameter changes can legitimately shift a strategy's behavior more than execution noise does. A strategy using RSI(14) that still works at RSI(12) or RSI(16) is robust. It doesn't need to be equally profitable — just not broken.

Important lesson: I initially used 5% parameter variation, and it was too small to be meaningful. Use 15–20% for real stress testing. If a strategy breaks with 15% parameter variation, it's dangerously overfit.

---

Phase 5: Multi-Timeframe Validation (0%–100%)

This is where it gets interesting. A strategy generated on H1 should show some viability on adjacent timeframes (M30 and H4). If it completely falls apart, it's likely exploiting timeframe-specific noise rather than a real market pattern.

The workflow automatically maps adjacent timeframes:

- M5 → M1, M15
- M15 → M5, M30
- M30 → M15, H1
- H1 → M30, H4
- H4 → H1, D1
- D1 → H4

Criteria are very lenient — we're not expecting profitability, just "not catastrophic":

- Must pass at least 1 of 2 adjacent timeframes (50% pass rate)

The workflow automatically fetches data for adjacent timeframes before testing. This uses 100% of data since it's a completely different timeframe — there's no "seen/unseen" concern.

---

Phase 6: Multi-Instrument Validation (0%–100%)

Same idea but across related currency pairs. A EURUSD strategy should show some viability on structurally related pairs that share market dynamics.

The workflow has a hardcoded mapping of related instruments:

- EURUSD → GBPUSD, USDCHF, EURGBP, EURJPY, EURCAD
- USDJPY → EURJPY, GBPJPY, AUDJPY, CHFJPY, CADJPY
- GBPUSD → EURUSD, EURGBP, GBPJPY, GBPAUD, GBPCAD
- AUDUSD → NZDUSD, AUDJPY, AUDCAD, AUDCHF, AUDNZD

Pass requirement is percentage-based: min_pass = max(1, floor(count / 3)) — roughly 33%:
- 3 related instruments → minimum 1 must pass
- 6 related instruments → minimum 2 must pass

Tagging system: The results get tracked as tags like MI4of5 (passed 4 out of 5 instruments). These tags flow through to the final output filename, so at the end you can see at a glance:
- P8_EURUSD_H1_TF2of2_MI5of5.json = fully validated, strong
- P8_EURUSD_H1_TF1of2_MI2of6.json = passed minimums, borderline

---

Phase 7: Cross-Broker Validation (0%–80%)

Test the strategy against a different broker's data feed. I use Eightcap as the cross-broker (ECN with tight spreads, which makes it a slightly tougher test).

This answers: is the strategy capturing a real market pattern, or is it an artifact of Premium's specific price feed?

---

Phase 8: Forward Test (80%–100%)

The final boss. The last 20% of data that the strategy has never seen in any form. No generation, no optimization, no Monte Carlo — just pure validation on unseen data.

This is the closest simulation to what will happen when you go live. If a strategy survives all 7 previous phases and still performs well on completely unseen forward data, you can have much higher confidence it has a genuine edge.

Settings are moderate (the 20% data window is smaller, so absolute metrics will be lower)

---

Real Results: USDJPY H1 Through All 8 Phases

Here's an actual strategy that made it through the entire pipeline:

+---------------+-----------------------+------------------------+-----------------------+
| Metric        | Phase 3 (80% Premium) | Phase 7 (80% Eightcap) | Phase 8 (20% Forward) |
+---------------+-----------------------+------------------------+-----------------------+
| Profit        | $3,939                | $4,528                 | $3,763                |
| Profit/day    |  $0.84                |  $0.96                 |  $3.22                |
| Profit Factor |   1.43                |   1.13                 |   1.33                |
| R-squared     |  86.21                |  37.61                 |  72.22                |
| Return/DD     |   4.87                |   2.41                 |   3.66                |
| Max Drawdown  |   6.6%                |  18.5%                 |  8.31%                |
| Stagnation    |  19.3%                |  42.6%                 | 16.15%                |
| Trades        |    517                |    907                 |    346                |
+---------------+-----------------------+------------------------+-----------------------+

The strategy actually improved on forward data — profit per day jumped from $0.84 to $3.22 and drawdown stayed low at 8.3%. It also passed all adjacent timeframes (M30, H4) and all 5 related JPY instruments (EURJPY, GBPJPY, AUDJPY, CADJPY, CHFJPY).

Final output: P8_USDJPY_H1_TF2of2_MI5of5.json

---

How It All Runs: Automation

The entire pipeline is a single Windows batch script.

The script:
1. Fetches data for the main instrument
2. Runs all 8 phases sequentially
3. Automatically fetches data for adjacent timeframes before Phase 5
4. Automatically fetches data for related instruments before Phase 6
5. Prints statistics after each phase
6. Tags the final output with TF and MI results

---

Lessons Learned the Hard Way

  1. Separate your Monte Carlo tests. Combined MC (execution + parameters) was killing 99.8% of strategies. Separated, I get much better survival rates AND know exactly why a strategy failed.

  2. Don't be too strict too early. I spent months with overly tight criteria and getting zero strategies through. Better to have moderate filters across many phases than one impossibly strict filter.

  3. 5% parameter variation is useless. Use 15–20% to actually stress test. If your strategy breaks at 15% variation, it WILL break in live trading.

  4.valid_tests_percent = 80 is too strict for initial discovery. Start with 65–75% and only tighten later if needed.

  5. Reserve 20% of data for forward testing. Never touch it during any other phase. This is your reality check.

  6. Multi-instrument testing reveals a lot. A strategy that works on 5+ related pairs is much more likely to capture a real pattern than one that only works on a single instrument.

---

Final Thoughts

This pipeline isn't magic — no system can guarantee profitable strategies. But it dramatically reduces the probability of deploying an overfitted strategy. The philosophy is simple: test from every angle, and only trust what survives everything.

A huge thank you again to Mr. Popov for Express Generator. The speed, the command-line interface, the INI configuration system, and the data percentage splits — all of these features made this kind of automated pipeline possible. And the continuous updates keep making it better. This community and this tool have been invaluable.

Let's build robust strategies together!

Cheers,

Hani

I've started working on a Multi-Market functionality for Express Generator.

Why is it necessary?

I'm using Express Generator on two servers to generate and validate strategies for the Top 10 Robots App.

My current workflow is:
1) Generate strategies for each symbol and period (16 symbols and 4 periods). The Generator exports collections with capacitty of 1000 strategies.
2) Validate the strategies against the Acceptance Criteria (AC). The exported collections are up to 100 strategies.
3) Validate the AC collections with Monte Carlo (MC).
4) Cross-validation - Validate passed MC collections with data from another server sequentially. I validate M15, M30, and M30 on one additional server, and M5 on 5 servers. M5 collections also go to an application called Prop Firm Robots.
5) All new strategies are validated with already collected strategies.

My Acceptance Criteria are rather strict, and only a few strategies can progress through the workflow.

Now I decided to change the approach. Instead of throwing off the strategies at the beginning of the workflow, I want to reduce the AC and to move it into the Application itself.
Instead of strict AC, I want to use a Multi-Market validation, like:
1) Generator
2) Monte Carlo
3) Multi-Market
4) Cross-validation
5) AC in the applications

The problem is that the current implementation of Express Generator makes it easy to do sequential validations:
- generate
- test on the market A
- which pass A, test on market B
- ...

We cannot test strategies on 5 markets and take those that pass 3 markets.

I'm going to solve that problem this week.