• Registered: 2024-02-23
  • Posts: 118

Topic: Survey: Incubation, Yes or Not?

Hey everyone,

I’m curious to see how different traders handle the demo incubation phase before going live.
The goal isn’t to decide what’s right or wrong — just to share, understand, and improve how each of us approaches this step.

If you’d like to take part, simply copy (or “quote”) the questions below and check (✅) the options that match your workflow.
Even short replies are useful to get a clearer view of what’s common (and what’s not) in our community.

⸻

1. Do you use a demo incubation phase in your workflow?
☐ Yes☐ No☐ Sometimes / depends on the strategy

1.1. If not, why?
(Free text answer)

2. Average duration of the demo incubation phase:
☐ <1 month☐ 1–3 months☐ 4–6 months☐ >6 months

3. Average number of strategies incubated at the same time:
☐ <20☐ 21–50☐ 51–100☐ >100

4. Broker or environment used for incubation:
☐ Same broker as live trading
☐ Different broker but similar conditions
☐ Any broker (just for technical testing)

5. Main criteria to promote a strategy to live:
☐ Profit factor
☐ Win rate
☐ Drawdown
☐ Consistency with backtest
☐ Other (please specify in comments)

6. Do you use incubation only to test logic, or also to check live robustness (spread, execution, slippage)?
☐ Only logic☐ Both☐ Depends on the EA type

7. Do you update strategies during incubation (e.g. rebalancing, optimization, filters)?
☐ Never☐ Occasionally☐ After each new data cycle

8. Do you have fixed thresholds to decide when to promote to live or discard?
☐ Yes, which one: ………………
☐ No
☐ Variable depending on the group

9. How do you monitor performance during incubation?
☐ Manually (Excel, charts, logs)
☐ Dashboard / automated tool
☐ Python (custom scripts)
☐ Artificial intelligence / automated analysis
☐ Just by general observation

10. In one sentence: what have you learned or improved thanks to demo incubation?
(Open answer)

⸻

Goal of this thread: gather experiences and ideas to improve how we test and qualify strategies before going live.
All input is welcome — even short answers help map different workflows in the community.

If at least 50 people participate, I’ll compile the results into a community summary post.

⸻

  • Registered: 2023-11-26
  • Posts: 125

Re: Survey: Incubation, Yes or Not?

So, Vicenzo, I don't use any kind of demo incubation phase before running my live setup. Precisely because what I realized is this: if you already have a validated method for creating strategies using only historical backtesting, and every time you use that exact same method and deploy the strategy to a real, live account, it delivers results... then all you need to do is keep using that same method to generate strategies that no longer need incubation. Got it? That's the reason why.

I'll try to give you an example. I spent a year developing my method and testing it on a demo account. After a certain point, I noticed that every single time I used my method to generate a strategy and put it on a demo account, it yielded positive results.

So, what did I do? I stopped using demo accounts. Nowadays, I have an effective method for creating strategies using EA Studio, and I add the risk management to them afterward—assuming the strategy is worthy and has passed the robustness tests. I add it later. So, what I do is basically very simple: I go in, generate the strategies, create my portfolio, add the risk management, and deploy it directly to a real account.

I don't need the incubation testing now because I already used the incubation phase in the past to validate my creation method. Make sense?

I believe I managed to express myself well in this text, even though I am translating from Portuguese to English. The bottom line is that if you have a well-defined and previously validated method, any strategy you create using that method is going to bring you real results, you know?

  • Registered: 2024-02-23
  • Posts: 118

Re: Survey: Incubation, Yes or Not?

Hi Gabdecster,
thanks, I fully understand your message (no worries about lenguages an translation) and perfectly your logic.

It makes sense that after validating your method for a long time, you trust the process and do not want to repeat a full demo phase every time. Nothing to object.

I’m still learning a lot, and the extensive incubation I built up (30 account ~ 1000 running EAs) is for me an unbelievable sorce of learning.

However, my the flow is just a bit different:

Generating method/s -> incubation phase -> performance evaluation -> copy signals to real account

And not directly:

… > incubation phase -> move to real account

The incubation step helps me observe whether the strategy behaves in line with expectations before giving it direct live capital, to benchmark different methods, and to build and prepare portfolios made up of 10–15 strategies for live capital.

Copying the signals, instead of moving the EA directly from demo to a live account, helps reduce the risk of losing continuity in performance when a strategy moves from incubation to direct live capital.

So I see incubation as a bridge between the method and the real account, and signal copying as a smoother way to make that transition.

Thank you again for sharing methods/thougths, this I a great place for sharing experiences supported by evidences.

All the best
Vincenzo

  • Registered: 2024-02-23
  • Posts: 118

Re: Survey: Incubation, Yes or Not?

Hi Gabdecsters,

Hope you are doing well.

I fully understood your explanation. I was just trying to translate it into numbers, and if I understood correctly, this would mean that your EA success rate is 100%:

Number of successful (profitable) strategies / Total number of strategies created = 100%

Is that correct?

My second question is about the phrase “...add the risk management...”. What exactly do you mean by that?
Are you referring to position sizing and portfolio balancing, or to something else?

Thank you in advance.

Vincenzo


gabdecsters wrote:

So, Vicenzo, I don't use any kind of demo incubation phase before running my live setup. Precisely because what I realized is this: if you already have a validated method for creating strategies using only historical backtesting, and every time you use that exact same method and deploy the strategy to a real, live account, it delivers results... then all you need to do is keep using that same method to generate strategies that no longer need incubation. Got it? That's the reason why.

I'll try to give you an example. I spent a year developing my method and testing it on a demo account. After a certain point, I noticed that every single time I used my method to generate a strategy and put it on a demo account, it yielded positive results.

So, what did I do? I stopped using demo accounts. Nowadays, I have an effective method for creating strategies using EA Studio, and I add the risk management to them afterward—assuming the strategy is worthy and has passed the robustness tests. I add it later. So, what I do is basically very simple: I go in, generate the strategies, create my portfolio, add the risk management, and deploy it directly to a real account.

I don't need the incubation testing now because I already used the incubation phase in the past to validate my creation method. Make sense?

I believe I managed to express myself well in this text, even though I am translating from Portuguese to English. The bottom line is that if you have a well-defined and previously validated method, any strategy you create using that method is going to bring you real results, you know?

  • Registered: 2024-02-23
  • Posts: 118

Re: Survey: Incubation, Yes or Not?

Longevity Index — From Concept to Portfolio Reality

Hi everyone,

following up on the Longevity Index idea, I wanted to move beyond the concept and test it in a more practical, real-world setting.

The starting point was a simple question:
What actually happens after a strategy reaches Top Band and is treated as “ready for live”?

Framing
This is not about how to build or validate strategies. It assumes that step is already done. The focus here is different: what happens to strategies after they reach Top Band and are deployed at portfolio level. What do we need to do to grant >70% of high performing EA?

The context
This analysis comes from a live environment where:

* ~900–1,000 EAs are running
* across ~30 MT4 instances
* continuously generated, validated, and tracked
* running since 06/2024

Our python End of Month check point labels each EA in:

- EB=Earth Birds, newly incubated
- OgI=On going Incubation, after x trades just for monitoring
- PwL=Promoting Watching List, after y trades very good performance
- RfL=Ready for Live, the best in class

Down into the appendix all KPIs listed.

This is not a curated or optimized portfolio.

On purpose, we are running strategies across:

* different assets
* different timeframes
* different logics
* different parameter sets

The goal is not to find the “perfect strategy”.
The goal is to observe what happens at scale, under real conditions.

What the data shows
Only ~8–10% of EAs reach Top Band (RfL + PwL)

Top Band is not a soft label — it is defined by strict KPI thresholds
(PF, Win%, SQN, sample size, max consecutive losses, recovery factor, etc.)

→ Only statistically high-performance strategies are promoted.

System nature
We are effectively operating a: high-volume, low-conversion system that produces a limited number of statistically validated high-performance EAs.

The test
* 19 EAs (all first-time Top Band entries during the observation period)
* first entry into Top Band (Sep–Oct 2025)
* tracked ~5–6 months

These are not optimized results, but observations from a deliberately broad and unfiltered environment, designed to reflect real operating conditions rather than ideal scenarios.

In practice, this comes down to one question: what would have happened if, 6 months ago, we had built a portfolio using the 19 EAs promoted by the incubation pipeline?

Baseline results after (6 months)
* Top Band: ~21% still high performing
* OgI: ~26% neutral, not good not bad and not damaging the portfolio
* PB: ~53% heavy degradation, most probably will die soon

More than half degrade into failure state, but can recover.

Behavior along the time
* PwL duration ≈ 1 month (median)
* RfL duration ≈ 3 months (median)

So, the Top Band is not a stable state — it is transient. This creates a structural need for continuous replacement.

Constraint
* conversion to Top Band ≈ 9%
* EA generation inflow ≈ 4–14/month
Supply is limited.

Scenarios tested (what happens if we manage or not the decay)
1. No action → portfolio collapses (soon or later)
2. PB only replacement → ~63% Top Band
3. Replacing everything immediately (PB + OgI monthly)
            → works in theory
            → breaks in practice (not enough inflow)
4. PB immediate + OgI with ~1 month tolerance (only viable configuration)
           → ~74% Top Band
             → 0% PB
             → stable portfolio size

Operational reality
* only ~3–10 replacements/month
* not every month
   
    → the system behaves in an event-driven way, not a linear or continuous process

The Re-entry
What really surprised me is that ~30–50% of Top Band inflow comes from re-entry, typically after 1–3 months.

    → the system is cyclical, high performing strategies cyclically move up and down within different performing areas.

Conclusion
The limiting factor is not strategy selection. It is the system’s ability to replace decaying strategies fast enough to sustain portfolio quality.

Even with strict statistical filters, strategies do not remain stable indefinitely.
At portfolio level, performance becomes a function of decay rate vs replacement capacity.

The Real asset
The real asset is not the individual strategy. It is the pipeline. I’ll never regret the incubation implementation as this is per sé the most important move we did to enable learning journey.

Open question
Curious how others approach this:

    * Do you treat Top Band strategies as something to hold?
    * Or something to rotate?

And:

* How do you measure performances and system effectiveness?

⸻

Appendix — Definitions & KPIs

RfL (Ready for Live)
* PF > 1.5
* and Win% > 60%
* and SQN ≥ 2
* and Net Profit > 0
* and Recovery Factor > 0.5
* and Max consecutive losses ≤ 5
* and Trades ≥ 50
⸻
PwL (Promotion Watchlist)
* PF: 1.3–1.5
* and Win%: 55–60%
* and SQN: 1.6–2
⸻
Top Band (TB)
* RfL + PwL
⸻
HB (High Band)
* RfL only
⸻
OgI (On going Incubation)
* neutral state
⸻
PB (Pruning Box)
* PF < 1.1 OR Win% < 45% OR Recovery < 0.5
* OR consecutive losses > 5
* OR SQN < 0.5
⸻
EB (Earth Birds)
* Trades < 10
⸻

Final note
All this work is possible because of the Team Effort. Thanks to the invitation in this forum and other groups, we are now 5 people (Hez, Fabio, Eliseo, Alessandro and myself) managing the whole system.

And not only: ~30 MT4 instances running 24/7 on high performing & reliable VPSs to properly get at market 1000 EA need to be monitored, managed, and paid.

Thank you in advance
Vincenzo