Scalping Research Branch
The Scalping research branch studies short-duration market opportunities where spread, volatility, movement speed and current market structure are especially important.
The methodology has changed significantly during development, but the research objective remains the same: identify short-term trading conditions in which the available market movement is large enough relative to spread, risk and execution cost.
The current research does not start by choosing one fixed currency pair. Instead, the system first scans multiple instruments, identifies suitable market conditions and estimates direction before a paper-trading funnel is allowed to evaluate a possible entry.
Development of the Scalping Method
FX Scalping A/B V1 and V1.1
The first active versions were FX Scalping A/B V1 and FX Scalping A/B V1.1.
These versions tested short-term entry logic on predefined instruments. V1 used a global maximum of one open paper position, while V1.1 introduced controlled re-entry and allowed a second position only under defined conditions.
The versions produced a meaningful number of profitable trades, but the overall structure was not strong enough. The main weakness was not simply the percentage of losing trades. The losing trades were too large relative to the gains and repeatedly outweighed otherwise successful entries.
This showed that improving only entry thresholds or individual exit parameters would probably not solve the main problem. The research architecture itself needed to change.
V1 and V1.1 were therefore removed from active operation and preserved as earlier stages of the Scalping methodology.
Transition to Dynamic Instrument Selection
The next development stage changed the order of the decision process.
The earlier approach effectively followed this structure:
preselected instrument → search for entry → manage trade
The modernised approach uses a different sequence:
scan the market → select suitable instrument → estimate direction → evaluate paper entry
This distinction is important for scalping because a currency pair that is normally suitable for short-term trading may temporarily have an excessive spread, insufficient volatility or poor movement quality.
Dynamic selection allows the system to reject such instruments before the entry stage.
FX Dynamic Scalping Selector V1.2.1
FX Dynamic Scalping Selector V1.2.1 became the scanning and direction layer of the modernised Scalping architecture.
Its purpose is to observe a wider group of currency pairs and determine which instruments currently satisfy basic short-term market conditions.
The selector evaluates factors including:
- current bid-ask spread;
- ATR and recent volatility;
- movement speed;
- impulse quality;
- available short-term movement potential;
- probable BUY or SELL direction;
- the relationship between potential movement and trading cost.
V1.2.1 does not independently represent a complete trading robot. It acts as the upstream market-selection layer for later paper research.
Current status: Active Scan
Mode: PAPER_SIM_ONLY
orders_sent: NO
FX Dynamic Scalping Selector V1.3.2 Paper Funnel
V1.3.2 introduced a paper funnel designed to convert a smaller subset of selector signals into simulated trades.
The version was technically safe and remained completely paper-only, but the Week 32 audit showed that its filtering structure was too restrictive.
During the reviewed week:
- the raw selector produced 549 YES signals;
- the reconstructed V1.3.2 candidate set contained 128 rows;
- only 41 closed paper trades were produced.
For a scalping research line, this represented a substantial reduction in usable signal flow.
The result did not show that the upstream selector was inactive. Instead, it suggested that the paper funnel was discarding too many potentially useful observations before they could become trades.
V1.3.2 therefore became an important intermediate research stage, but it was not retained as the active funnel.
Current status: Disabled / Archived Research Stage
Week 32 Throughput Review
During the Week 32 audit, several broader filtering scenarios were reconstructed from the available selector data.
The most useful result came from the scenario called All Symbols NY Only Spread2.
This configuration produced 446 rows during the week and used a much simpler structure:
- all symbols available from the selector;
- NY-session observations only;
- maximum spread of 2.0 pips.
The significantly higher throughput indicated that a broader NY-session structure could provide a better research base than the heavily filtered V1.3.2 funnel.
This observation became the basis for V1.4A.
FX Dynamic Scalping Selector V1.4A Paper Funnel
The current active paper candidate is FX Dynamic Scalping Selector V1.4A Paper Funnel.
V1.4A was reconstructed from the V1.3.2 framework but deliberately removes several restrictions that had reduced signal throughput.
The current V1.4A funnel:
- uses all symbols supplied by the selector;
- accepts NY_OVERLAP and NY_LATE sessions;
- requires spread <= 2.0 pips;
- requires a signal that resolves clearly to BUY or SELL;
- does not use the previous JPY-only restriction;
- does not use the previous move_60m filter;
- does not use the previous last-body filter.
The objective is not to maximise trade count without control. The purpose is to restore enough signal throughput for the research to produce a meaningful statistical sample while retaining basic market-quality restrictions.
Current Research Architecture
After the Week 32 consolidation, the active Scalping setup consists of only two scheduled research processes.
- FX Dynamic Scalping Selector V1.2.1 Scan
- FX Dynamic Scalping Selector V1.4A Paper Funnel
The structure can be summarised as:
market scan → instrument selection → direction assessment → V1.4A paper funnel → closed paper results
V1.2.1 provides the broader market observations. V1.4A then applies the current NY-session funnel and creates paper trades when its conditions are satisfied.
Current Research Objective
The immediate objective is to determine whether V1.4A can generate enough trades to evaluate the Scalping methodology properly without returning to the weak risk structure seen in the original A/B versions.
The next observation period will therefore focus on:
- number of available selector signals;
- number of signals accepted by V1.4A;
- paper trade frequency;
- positive and negative R distribution;
- whether losses remain controlled relative to profitable trades;
- whether the broader NY-session rules create a sufficiently stable sample.
A higher number of trades alone will not be treated as success. The broader funnel must also demonstrate an acceptable relationship between opportunity frequency, losses and positive R.
Version Status
- FX Scalping A/B V1 — archived earlier Scalping implementation;
- FX Scalping A/B V1.1 — archived earlier Scalping implementation;
- FX Dynamic Scalping Selector V1.2.1 — active scan and direction layer;
- FX Dynamic Scalping Selector V1.3.2 Paper Funnel — disabled / archived intermediate stage;
- FX Dynamic Scalping Selector V1.4A Paper Funnel — current active paper research candidate.
Current Testing Status
Methodology: Scalping
Active candidate: FX Dynamic Scalping Selector V1.4A Paper Funnel
Mode: PAPER_SIM_ONLY
orders_sent: NO
Demo trading: No
Live trading: No
V1.4A is not a finished trading robot and is not a Demo Candidate at this stage. It is a paper-only research version being used to determine whether the broader Dynamic Scalping architecture can produce enough useful trades for further development.
Research Conclusion
The Scalping branch has progressed from fixed-pair entry testing to a two-stage architecture based on dynamic market selection and a separate paper funnel.
The earlier A/B versions demonstrated weaknesses in risk structure. V1.3.2 demonstrated that excessive filtering could make the research sample too narrow. V1.4A now tests a broader compromise: preserve basic market-quality controls while allowing enough NY-session signal flow to produce meaningful paper statistics.
The next development decision will depend on the observed trade sample rather than on the design assumptions alone.
Risk notice: The Scalping systems described on this page are experimental research versions. Paper-simulation results do not guarantee future performance and do not constitute financial or investment advice.
Weekly Report 32 — Research Consolidation and Dynamic Scalping V1.4A