Trading Intelligence & Confluence Engine
A deterministic trading-intelligence demonstration combining weighted signals, risk gates and auditable reason logs.
The Trading Intelligence & Confluence Engine demonstrates how several market inputs can be converted into one deterministic decision state while preserving explicit risk gates and an auditable explanation of why an alert was produced, withheld, or blocked.
MFXG Internal R&D Demonstration — synthetic or controlled inputs; not client work or live investment performance.
The problem
A trading workflow may combine trend, momentum, market context, execution conditions and trader-defined rules. If those inputs are combined informally, it can be difficult to reconstruct why one setup was accepted and another was rejected.
The demonstration turns that reasoning into explicit software logic.
Decision contract
| Input or control | Function | Result |
|---|---|---|
| Direction | Gives each configured signal a positive, negative or neutral contribution | Supports long, short or neutral classification |
| Confidence | Scales how strongly a signal contributes | Lower-confidence observations contribute less |
| Weight | Defines the relative importance of configured inputs | Produces deterministic scoring |
| Spread | Execution-risk gate | Can block an otherwise directional state |
| Reason log | Stores the inputs and gate outcome | Makes the resulting state auditable |
Recorded decision examples
The controlled evidence includes three deliberately different outcomes.
| Instrument label | Score | Direction | Final state | Risk gate |
|---|---|---|---|---|
| ES | 0.762667 | Long | Alert | Passed |
| NQ | 0.247333 | Neutral | Insufficient confluence | Passed |
| GC | -0.756667 | Short | Blocked by risk gate | Failed |
The GC example is important. Its directional inputs agree on a short state, but the configured spread is 18 basis points against a 12-basis-point maximum. The system therefore blocks the state rather than treating directional agreement as sufficient.
Why the reason log matters
A decision system should make rejection states as observable as alert states. The evidence records each signal's direction, confidence and weight, followed by any risk-gate explanation. That structure can support later review without relying on memory or a screenshot of the final signal.
What this demonstrates
The demonstration shows deterministic signal combination, explicit gating, neutral-state handling, and structured reason logs.
It does not place orders. It does not establish profitability, brokerage compatibility, or a recommendation to trade ES, NQ, GC, or any other instrument.
Explore the working demonstration
Open the Trading Intelligence & Confluence Engine to inspect the controlled decision examples.
For related work, see Trading Technology & Software Development.