← Back to General Weekly Reports
Weekly Research Overview
ISO Week 31 produced a mixed but informative set of results across the active trading robot research lines. Some robots continued to collect data without producing a decisive result, several exposed weaknesses that require further work, and one newer approach — AUDJPY Follow Driver — delivered the most notable performance of the week.
All robots operated in PAPER_SIM_ONLY mode. No real or demo orders were sent, and the execution status remained orders_sent = NO. The figures below are taken from the clean weekly report for ISO Week 31.
AUDJPY Follow Driver
The strongest result came from the AUDJPY Follow Driver SELL research line. This method does not analyse AUDJPY in isolation. It also monitors AUDUSD and USDJPY as potential driver instruments.
The underlying idea is that a clear movement in the driver pairs may be followed by a delayed reaction in AUDJPY. The robot attempts to identify this delay and open a paper trade in the direction of the expected continuation.
AUDJPY Follow Driver SELL V0.2
Version V0.2 closed two paper trades and finished the week at approximately +6.19R. This exceeded the current research reference level of +5R per week.
There were no losing closed trades. One position reached its target, while the second was closed after reaching its maximum holding period.
This is an encouraging first result, but it is not yet evidence of stable performance. A sample of two trades is too small to support a reliable conclusion. The version therefore remains under observation.
AUDJPY Follow Driver SELL V0.3
Version V0.3 also finished the week positively. It closed two trades and produced approximately +3.38R, with one profitable trade and one losing trade.
Historical tests made V0.3 an interesting research candidate, but during the current live paper observation period it has so far performed less strongly than V0.2. Both versions will continue to run in parallel because they represent different risk and timing profiles.
V0.2 is the more conservative version. It waits longer before entry, uses a smaller target and exits more quickly. V0.3 enters earlier, holds positions for longer and attempts to capture a larger movement.
EURJPY MACD
The EURJPY MACD line produced a calm positive result. The Base version and V30 each closed two trades and finished at approximately +1.92R.
This was not a breakthrough result, but it was also not a failure. At the current stage, the most useful action is to leave both versions unchanged and allow them to continue collecting live paper statistics.
EURJPY MACD V30.1 finished close to break-even with a small negative result. The outcome is not critical, but there is currently no strong reason to prioritise further development of this version.
GBPUSD MACD Logging Issue
The GBPUSD MACD V Base result cannot yet be evaluated correctly. The journal shows that a paper position was opened and later closed because of the maximum holding-time rule, but the closed-trade file did not produce a valid R result.
This is primarily a technical logging problem rather than a trading conclusion. Until the result is recorded correctly, it would be misleading to classify the robot as profitable or unprofitable.
FX Scalping A/B
The weakest group of the week was the previous FX Scalping A/B implementation.
FX Scalping A/B V1 finished at approximately -4.25R, while V1.1 finished at approximately -3.98R.
The percentage of profitable trades was not exceptionally poor. More than half of the trades were successful. The central problem was the relationship between profit and loss: losing trades were too large and outweighed the gains produced by the winning trades.
This indicates that the earlier scalping logic could identify some valid short-term movements, but its instrument-selection process, exit structure and risk profile were not effective enough. The previous versions will therefore not be promoted further in their existing form.
Development continues within the same Scalping research branch, but through a substantially modernised architecture based on dynamic instrument selection.
Read more: Scalping Research Branch
Inactive Research Lines
Two robots were removed from active observation:
- AUDCAD MACD V Base
- GBPUSD MACD V30
Their scheduled tasks were disabled and their working files were moved out of the active area. The data was preserved for research history, but these versions are no longer part of the current active monitoring group.
Week 31 Decision
The main development signal of the week came from AUDJPY Follow Driver. V0.2 exceeded the +5R weekly reference, while V0.3 also remained positive. However, the trade count is still too small to classify either version as stable.
The current decision is therefore:
- continue paper observation of AUDJPY Follow Driver V0.2 and V0.3;
- prepare V0.2 for a separate broker-demo observation stage;
- continue collecting statistics from the stable EURJPY MACD versions;
- do not develop EURJPY MACD V30.1 at this stage;
- replace the previous Scalping A/B architecture with the modernised Dynamic Scalping Selector approach;
- repair the GBPUSD result-logging process before evaluating that line.
Week 31 did not produce a final trading solution. It produced something more useful for the current research stage: one promising Follow Driver result, several neutral MACD observations, and a clear technical reason to modernise the Scalping methodology.
Research status: Paper simulation only. No live orders were sent.
AUDJPY Follow Driver Research
Scalping Research
EURJPY MACD Research
← Back to Robot Library