Robot branch: EURGBP V4
Version reviewed: EURGBP V4.1.28 M5 Body Filter
Observation period: Week 28
Week 28 result: -3.1288R
Research decision: Replace the broad V4.1.28 structure
Next research version: EURGBP V4.2.29 Selective SELL Groups
Current priority: Tested Logic
Mode: PAPER_SIM_ONLY
orders_sent: NO
Live orders: No live orders were sent
Executive Summary
EURGBP V4.1.28 completed Week 28 with 35 closed paper trades and a final result of -3.1288R.
The version continued to generate profitable trades, producing +12.1972R in Gross +R. However, the negative side reached -15.3260R, leaving the overall weekly result below zero.
The Week 28 audit showed that the M5 Body Filter was not sufficient to make the broad V4 structure selective enough. Instead of adding another small blacklist filter, the EURGBP V4 history was reviewed across all available deduplicated V4 and V4.1 trades.
That review identified two selective SELL groups with a combined historical result of:
- 18 trades
- 16 winning trades
- 2 losing trades
- 88.89% win rate
- +10.4094R Net R
Based on this result, the broad EURGBP V4.1.28 branch was replaced by a new paper-only research version: EURGBP V4.2.29 Selective SELL Groups.
V4.2.29 is not a Demo Candidate. It remains Tested Logic and requires fresh forward paper observation during Week 29.
Purpose of This Report
This report documents the Week 28 result of EURGBP V4.1.28 and explains why the EURGBP V4 research branch moved from broad signal filtering to selective signal admission.
The report covers two separate stages:
- the live paper result produced by EURGBP V4.1.28 during Week 28; and
- the subsequent full-depth review used to design EURGBP V4.2.29.
This distinction is important. The historical group analysis was used to create the next research hypothesis. It was not a forward result produced by V4.2.29.
EURGBP V4.1.28 Research Logic
EURGBP V4.1.28 was created after the V4.1.27 audit.
Its main change was an M5 Body Filter designed to block candidate entries when the current M5 candle body was moving against the intended trade direction.
The purpose of the filter was to reduce low-quality entries without changing the underlying broad V4 corridor structure.
Week 28 was used to determine whether this controlled change could improve the balance between positive and negative trades.
Week 28 Results
| Metric | EURGBP V4.1.28 |
|---|---|
| Closed trades | 35 |
| Winning trades | 16 |
| Losing trades | 18 |
| Zero-result trades | 1 |
| Win rate | 45.71% |
| Gross +R | +12.1972R |
| Gross -R | -15.3260R |
| Net R | -3.1288R |
| Average R per trade | -0.0894R |
| TP exits | 9 |
| SL exits | 12 |
| MAX_HOLD exits | 14 |
Internal weekly target: +5R
Target status: Not reached
Interpretation of the Week 28 Result
EURGBP V4.1.28 was not inactive and did not fail because it produced no useful trades.
The version generated +12.1972R of positive trade material. The problem was that this was accompanied by -15.3260R of negative trade material.
The resulting -3.1288R showed that the broad V4.1.28 structure still admitted too many losing or low-quality entries alongside its profitable signals.
The exit distribution supported the same conclusion:
- 9 trades reached TP;
- 12 trades reached SL;
- 14 trades closed through MAX_HOLD.
The large number of SL and MAX_HOLD exits showed that the M5 Body Filter did not sufficiently isolate the stronger parts of the V4 signal population.
Why the M5 Body Filter Was Not Enough
The M5 Body Filter was a controlled and logically justified modification. It was based on the observation that some negative trades entered while the current candle body was moving against the proposed direction.
However, Week 28 showed that candle-body direction was only one part of the problem.
The broad V4 model continued to admit signals from many different sessions, position areas, slot types and body conditions. Some of these combinations produced useful trades, while others repeatedly reduced the final result.
Adding one more blacklist condition would have continued the same development pattern:
- allow a broad group of signals;
- identify one negative feature;
- block that feature;
- repeat the process after the next weekly audit.
The Week 28 decision was to stop extending that pattern and test a different research structure.
Full-Depth V4 Review
After the Week 28 result, the available EURGBP V4 and V4.1 paper history was reviewed as one deduplicated research dataset.
The objective was not to optimise only Week 28. The review searched the available multiweek history for:
- signal groups that repeatedly generated large negative contributions;
- signal groups that retained positive results across more than one week;
- conditions that could form a simple, auditable allowlist;
- and groups that could be monitored independently during fresh forward paper observation.
After deduplication, the broad V4-line baseline contained 218 trades.
Full Deduplicated V4 Baseline
| Metric | Full V4 Baseline |
|---|---|
| Trades | 218 |
| Winning trades | 117 |
| Losing trades | 100 |
| Zero-result trades | 1 |
| Win rate | 53.67% |
| Gross +R | +88.0792R |
| Gross -R | -85.1035R |
| Net R | +2.9757R |
| Average R per trade | +0.0137R |
| TP exits | 69 |
| SL exits | 69 |
| MAX_HOLD exits | 80 |
The broad baseline finished slightly above zero, but the average result was only +0.0137R per trade.
Gross +R and Gross -R were also almost equal. This confirmed that the broad V4 structure contained useful positive material but almost no meaningful aggregate edge after all admitted trades were included.
Research Direction: Allowlist Instead of Blacklist
The next test reversed the previous filtering approach.
Instead of asking which additional trades should be blocked, the review asked:
Which signal groups have earned permission to remain in the active research model?
Two SELL groups were selected for further testing.
Selected Group A
SELL + NY_LATE_16_20 + FIRST_SLOT
Group A contained first-slot SELL trades opened during the NY_LATE_16_20 session.
The group appeared during Weeks 27 and 28. It was small, but no losing trades were present in the available sample.
| Metric | Group A |
|---|---|
| Trades | 6 |
| Winning trades | 6 |
| Losing trades | 0 |
| Win rate | 100.00% |
| Gross +R | +4.5171R |
| Gross -R | 0.0000R |
| Net R | +4.5171R |
| TP exits | 3 |
| SL exits | 0 |
| MAX_HOLD exits | 3 |
The result is promising, but the sample is too small to support a strong conclusion. Group A is therefore treated as a research hypothesis, not a confirmed trading edge.
Selected Group B
SELL + position_pct 90–100 + BODY_WEAK_POS
Group B selected SELL signals located in the upper 90–100 position-percentage area when the entry candle was classified as BODY_WEAK_POS.
This group appeared across more weeks than Group A and therefore provided the broader structural base for the new version.
| Metric | Group B |
|---|---|
| Trades | 12 |
| Winning trades | 10 |
| Losing trades | 2 |
| Win rate | 83.33% |
| Gross +R | +8.0849R |
| Gross -R | -2.1926R |
| Net R | +5.8923R |
| TP exits | 6 |
| SL exits | 2 |
| MAX_HOLD exits | 4 |
Group B was not loss-free, but it combined a high historical win rate with controlled Gross -R and positive total Net R.
Combined A+B Allowlist
The two selected groups were combined into one selective SELL allowlist.
| Metric | A+B Allowlist |
|---|---|
| Trades | 18 |
| Winning trades | 16 |
| Losing trades | 2 |
| Zero-result trades | 0 |
| Win rate | 88.89% |
| Gross +R | +12.6020R |
| Gross -R | -2.1926R |
| Net R | +10.4094R |
| Average R per trade | +0.5783R |
| TP exits | 9 |
| SL exits | 2 |
| MAX_HOLD exits | 7 |
Compared with the full baseline, the selective allowlist reduced the number of admitted trades from 218 to 18.
The purpose of this reduction was not to maximise trade frequency. It was to test whether a smaller set of explicitly selected conditions could preserve the positive part of the V4 history while excluding most of the negative contribution.
Weekly Distribution of the A+B Allowlist
| Week | Trades | Winning trades | Losing trades | Net R |
|---|---|---|---|---|
| Week 24 | 3 | 3 | 0 | +2.3634R |
| Week 25 | 5 | 3 | 2 | +0.4218R |
| Week 26 | 3 | 3 | 0 | +2.0266R |
| Week 27 | 4 | 4 | 0 | +3.7493R |
| Week 28 | 3 | 3 | 0 | +1.8482R |
The allowlist did not produce a negative week in the available deduplicated sample.
However, this result must be interpreted carefully:
- only 18 trades were included;
- weekly trade counts ranged from 3 to 5;
- Group A appeared only in Weeks 27 and 28;
- the results were extracted from the same historical research pool used to select the groups;
- and none of these figures constitutes independent forward confirmation of V4.2.29.
Analysis of the Excluded Trades
The 200 trades excluded from the A+B allowlist produced the following combined result:
| Metric | Excluded Trades |
|---|---|
| Trades | 200 |
| Winning trades | 101 |
| Losing trades | 98 |
| Zero-result trades | 1 |
| Gross +R | +75.4772R |
| Gross -R | -82.9109R |
| Net R | -7.4337R |
| TP exits | 60 |
| SL exits | 67 |
| MAX_HOLD exits | 73 |
This result supported the structural decision.
The broad V4 history did not lack winning trades. The excluded population still contained 101 positive trades and more than +75R in Gross +R.
The problem was that this positive material was accompanied by an even larger negative contribution.
The purpose of the new allowlist is therefore not to claim that every excluded trade is bad. Its purpose is to admit only the groups that currently have enough evidence to justify fresh forward testing.
Version Decision
Decision: REPLACE EURGBP V4.1.28 AS THE ACTIVE V4 RESEARCH VERSION
Previous active version: EURGBP V4.1.28 M5 Body Filter
New active research version: EURGBP V4.2.29 Selective SELL Groups
Priority: Tested Logic
Testing mode: PAPER_SIM_ONLY
EURGBP V4.1.28 will remain available as an audit reference, but its broad signal structure will not continue as the active V4 branch.
The new version represents a structural change:
- from broad signal admission to selective admission;
- from repeated blacklist filters to a defined allowlist;
- from mixed-direction V4 research to selected SELL groups;
- and from incremental V4.1 changes to a new V4.2 research branch.
EURGBP V4.2.29 Logic
EURGBP V4.2.29 is a selective SELL-only paper research model.
Allowed Group A
- Direction: SELL
- Session: NY_LATE_16_20
- Slot type: FIRST_SLOT
Allowed Group B
- Direction: SELL
- Position percentage: 90–100
- Entry body classification: BODY_WEAK_POS
A signal is admitted when it matches Group A or Group B.
All other signals are blocked with:
SIGNAL_BLOCKED_NOT_IN_SELECTIVE_SELL_GROUP
Why V4.2.29 Is Not a Demo Candidate
The historical A+B result is materially stronger than the full broad baseline, but this is not sufficient to classify V4.2.29 as a Demo Candidate.
The selected groups were identified using the same available V4 history on which their results were measured. This creates selection bias and means that the result must be tested on fresh, unseen market data.
Current limitations include:
- the combined sample contains only 18 trades;
- Group A contains only 6 trades;
- Group A appears only during Weeks 27 and 28;
- the strategy may generate very few trades during some weeks;
- the weekly result may remain below the internal +5R target even when positive;
- and no independent forward week has yet confirmed the new allowlist.
V4.2.29 therefore remains in the Tested Logic category.
Week 29 Research Objective
Week 29 will be used as the first fresh forward paper observation period for EURGBP V4.2.29.
The main questions are:
- Will Group A or Group B produce new qualifying signals?
- Will the selected trades remain positive outside the review dataset?
- Will the allowlist produce enough trade frequency for meaningful evaluation?
- Will Gross -R remain controlled?
- Will MAX_HOLD behaviour remain acceptable?
- Will the scheduled task and output files remain operationally stable?
No status upgrade will be made solely because the retrospective A+B result is positive.
Operational Status
EURGBP V4.2.29 was implemented as a separate paper-only research task.
- Mode: PAPER_SIM_ONLY
- orders_sent: NO
- Live orders: No live orders are sent
- Order execution: Disabled
- Output files: Separate V4.2.29 files
- Scheduled task: Separate V4.2.29 task
- Previous task: V4.1.28 disabled after V4.2.29 verification
The active EURGBP V4 research branch is now EURGBP V4.2.29 Selective SELL Groups.
Research Conclusion
EURGBP V4.1.28 completed Week 28 with a negative result of -3.1288R.
The version still produced more than +12R of profitable trade material, but its broad structure also admitted more than -15R of negative trade material. The M5 Body Filter did not reduce this negative side sufficiently.
The subsequent full-depth review found that the broad V4 baseline contained almost no aggregate edge after all admitted trades were combined.
A selective allowlist consisting of Group A and Group B produced a materially stronger retrospective result, but the sample remained small and was derived from the same history used for group selection.
For that reason, the result is treated as a new research hypothesis rather than proof of stable performance.
The Week 28 decision is:
- stop V4.1.28 as the active broad V4 branch;
- retain V4.1.28 as an audit reference;
- activate EURGBP V4.2.29 as Tested Logic;
- run V4.2.29 in PAPER_SIM_ONLY mode;
- and evaluate it using fresh Week 29 forward paper data.
Risk Note
This report documents an internal trading robot research process.
EURGBP V4.1.28 and EURGBP V4.2.29 operate only in PAPER_SIM_ONLY mode. orders_sent remains NO, and no real orders are sent.
The selected A+B groups were identified using historical paper results. Their past performance may be affected by selection bias and does not demonstrate how the new version will perform under future market conditions.
Historical analysis, backtesting, live paper observation and weekly robot reports do not guarantee future performance.
Forex and CFD trading involve significant risk. This report is provided for research and educational purposes only and does not constitute financial advice, investment advice, a recommendation to trade, or a promise of future returns.