Why the FX Scalping Branch Was Created
The FX Scalping A/B branch was not created to replace the earlier MACD research set.
It came from a separate research question:
Could a shorter-term M5 paper-simulation approach produce cleaner and more frequent research signals than the slower observation branches?
The earlier robots were built around broader directional logic, especially MACD-based structures with different symbols, timeframes, EMA settings, and MAX_HOLD limits.
That work remains useful, but it is naturally slower. Trades may take longer to open and close, which means that collecting enough weekly data can also take longer.
For a research project, this creates a practical limitation:
fewer trades mean slower feedback.
The new branch was therefore designed to generate more observations without simply increasing noise.
The objective was not to create a high-frequency trading system or an aggressive live trading robot. The objective was to identify small, clearly defined paper-simulation setups that could be tested, filtered, and compared in R.
From Higher Frequency to Better Filtering
The first step was to examine short-term M5 structures with enough historical activity.
Several raw ideas were reviewed, including:
- EMA pullback logic;
- Donchian breakout logic;
- different symbol and session combinations;
- different candle and trend conditions.
Most raw signal groups were not strong enough by themselves.
This was an important result because it showed that higher frequency alone does not create an edge.
A frequent signal can still be a poor signal.
The useful direction appeared only after the raw results were divided into narrower subgroups and tested under more specific conditions.
Two components survived the early research process.
Logic A — EURUSD EMA Pullback SELL
The first component became Logic A.
It uses:
- EURUSD;
- M5 timeframe;
- EMA pullback structure;
- SELL direction;
- EMA gap between 3 and 6 pips.
The underlying hypothesis is that EURUSD may sometimes continue moving downward after a controlled pullback toward the moving-average area.
However, the setup is accepted only when the EMA gap is neither too small nor excessively extended.
A gap that is too small may indicate weak directional separation.
A gap that is too large may indicate that the move is already overextended.
The 3–6 pip range therefore acts as a narrow structural filter rather than a general trend signal.
Logic B — GBPUSD Donchian Breakout BUY
The second component became Logic B.
It uses:
- GBPUSD;
- M5 timeframe;
- Donchian breakout structure;
- BUY direction;
- late New York session;
- a specific M5 candle-body range.
This logic is based on a different market behaviour.
Instead of entering after an EMA pullback, it looks for a controlled breakout condition.
Logic B also uses a smaller TP relative to SL. This means that its potential value depends heavily on:
- trade frequency;
- win rate;
- execution consistency;
- control of losing clusters.
Logic B is therefore not intended to behave like Logic A.
It represents a separate short-term trading hypothesis.
How the Two Logics Were Selected
Logic A and Logic B were not selected because they looked attractive on a chart.
They were selected because they survived several stages of analysis.
The research process included:
- raw signal screening;
- subgroup mining;
- reverse-signal audit;
- train and validation separation;
- combined-candidate analysis;
- external validation with Dukascopy bid/ask data.
The reverse-signal audit was especially important.
Weak signals were reversed to determine whether the original problem was simply incorrect direction.
The reversed signals did not produce a strong opposite strategy.
This suggested that the main problem was not direction alone. Better filtering and more controlled position management were still necessary.
After these checks, Logic A and Logic B were combined into a single paper-simulation candidate.
This became:
FX_SCALPING_A_B_V1
Why Logic A and Logic B Were Combined
The two logics were combined because they are different enough to function as a small portfolio-style research branch.
Logic A trades:
EURUSD SELL using an EMA pullback structure.
Logic B trades:
GBPUSD BUY using a Donchian breakout structure.
They are not the same setup repeated on two symbols.
They represent:
- different currency pairs;
- different entry structures;
- different market behaviours;
- different directional hypotheses.
The purpose of combining them was to place two independent short-term ideas inside one controlled paper-simulation framework.
FX_SCALPING_A_B_V1 — Conservative Baseline
The first version was built conservatively.
FX_SCALPING_A_B_V1 uses a global maximum of one open paper position.
This means that when one paper trade is already open, all additional signals are blocked until the existing position closes.
The baseline rule is:
Global maximum: one open paper position.
This structure avoids:
- stacked positions;
- overlapping exposure;
- simultaneous risk across multiple signals;
- unclear interaction between open trades.
V1 therefore serves as the conservative control version.
It answers the question:
How does the scalping branch behave under strict global position control?
Why a Second Version Was Needed
External validation using Dukascopy bid/ask data showed an important difference between the broad signal set and the strict one-position version.
The broader signal set appeared stronger because it included more opportunities.
However, unlimited parallel trading was not considered acceptable.
Allowing every overlapping signal would increase exposure and make the research structure less controlled.
The practical question was therefore not:
Can more trades be opened?
The more useful question was:
Can selected additional trades be allowed only when there is a defensible reason?
This led to the creation of a second version.
FX_SCALPING_A_B_V1_1 — Controlled Re-entry
FX_SCALPING_A_B_V1_1 was created to test controlled re-entry.
It was not designed as unrestricted pyramiding.
The key condition is that a second same-symbol paper position may only be opened when the existing same-symbol position has already reached at least:
+0.5R floating MFE
The robot is therefore not allowed to add a second position simply because another signal appears.
The second position is permitted only after the first paper trade has already moved favourably enough to demonstrate some evidence of momentum.
The V1_1 position rules are:
- maximum two total open paper positions;
- maximum two open paper positions on the same symbol;
- same-symbol re-entry only after the first position reaches at least +0.5R floating MFE.
This version answers the question:
Can part of the blocked signal value be recovered without allowing uncontrolled exposure?
What Is Being Compared
The entry logic is identical in both versions.
Both use:
- Logic A — EURUSD EMA pullback SELL;
- Logic B — GBPUSD Donchian breakout BUY.
The difference is only in position management.
FX_SCALPING_A_B_V1
- global maximum of one open paper position;
- all new signals blocked while a paper trade is open.
FX_SCALPING_A_B_V1_1
- maximum two total open paper positions;
- maximum two same-symbol positions;
- controlled same-symbol re-entry after +0.5R floating MFE.
This separation is important.
If V1_1 performs better, the improvement should come from controlled position management rather than from changes to the entry rules.
If V1 remains more stable, that would suggest that the stricter one-position structure remains the safer research path.
Separation from the MACD Research Set
The FX Scalping A/B branch should be evaluated separately from the earlier MACD observation set.
The MACD robots are:
- slower;
- broader;
- more trend and momentum oriented;
- more dependent on longer holding periods.
The scalping robots are:
- short-term;
- M5-based;
- built around smaller TP and SL structures;
- designed to generate faster research feedback.
These are different research hypotheses.
They should not be ranked directly against each other as winners or losers.
Each branch should be assessed according to its own logic, risk structure, trade frequency, and development objective.
Execution Safety Status
All work in the FX Scalping A/B branch remains:
PAPER_SIM_ONLY
The execution status remains:
orders_sent = NO
No live trading is enabled.
No order_send execution is used.
The purpose of the branch is controlled research only.
It is not currently a live trading product, demo robot, or completed trading system.
Immediate Research Goal
The immediate goal is not to claim that the branch is ready.
The objective is to observe whether the two-version test produces useful evidence.
The main comparison is between:
- the conservative baseline V1;
- the controlled re-entry version V1_1.
The next reviews should examine:
- whether V1_1 generates useful additional trades;
- how often the +0.5R floating MFE condition is reached;
- the results of second same-symbol positions;
- whether Net R improves;
- whether drawdown increases;
- whether trade clustering becomes excessive;
- whether the added exposure remains controlled.
An increase in trade count alone will not be treated as success.
The additional positions must demonstrate useful contribution without creating disproportionate risk.
Conclusion
The FX Scalping A/B branch is a new active research line inside FX Trading Robot Lab.
It is a compact M5 paper-simulation experiment built from two validated short-term ideas:
- EURUSD EMA pullback SELL;
- GBPUSD Donchian breakout BUY.
These ideas are being tested through two position-management versions:
- FX_SCALPING_A_B_V1 as the conservative baseline;
- FX_SCALPING_A_B_V1_1 as the controlled re-entry experiment.
The branch remains under strict safety limits with no live execution.
Its purpose is to determine whether carefully controlled additional exposure can improve the use of short-term signals without weakening overall risk control.