TYO AI Research Lab
AIMT5MCP
AI-assisted research for the next generation of algorithmic trading. This page describes how AI is actually used at TYO and what is still being explored — each item carries a status, and nothing here claims that AI predicts markets or guarantees results.
Status legend
Every technology on this page is marked with where it actually stands. "Live" means in daily use at TYO today; everything else is research in progress and is labelled as such.
01AI research architecture
How information flows when AI joins the research loop. The AI layer reads and analyses; trading decisions remain in the explicit, auditable logic of each EA.
- MetaTrader 5 Market data, execution, Strategy Tester
- MCP Model Context Protocol — a standard interface between AI agents and tools
- TYO AI Research Layer Where models read reports, code and data
- Market Analysis Structure, sessions, volatility regimes
- EA Analysis Reading strategy code and trade records
- Backtesting Tick-level runs in the Strategy Tester
- Risk Analysis Drawdown, streaks, exposure, stacking
- Candidate Strategy A hypothesis worth testing — never a promise
02AI development loop
The cycle a strategy idea travels. AI accelerates the steps between idea and evidence; the Strategy Tester remains the judge, and most candidates fail.
- Strategy Idea A market observation written as a rule
- EA Implemented in MQL5
- Compile Deterministic build
- Backtest Tick-level test in MetaTrader 5
- AI Analysis Reading the full report and deal list
- Improvement Hypothesis What the failure suggests changing
- Candidate EA The next version
- Validation Out-of-sample checks before anything ships
03What AI actually does here
AI is a research accelerator at TYO — not a signal source, not a forecaster, and not a guarantee of anything.
-
Research Live
Reading market structure, literature and prior tests
-
Development assistance Live
Drafting and refactoring MQL5, structuring test matrices
-
Backtest analysis Live
Parsing full tester reports — every trade, not just the summary. The analytics on this site are produced this way
-
EA comparison Live
Cross-system metrics, rankings and risk placement
-
Market regime analysis Research
Classifying conditions to test strategies against
-
Parameter research Research
Exploring input spaces systematically instead of by hand
AI does not predict markets, and no model here is presented as a forecast of future prices.
04MCP × MetaTrader 5
Model Context Protocol is an open standard that lets an AI agent talk to tools through a defined interface. TYO is exploring what a disciplined bridge between an AI agent and MetaTrader 5 makes possible for research.
-
Report reading via MCP Experimental
Serving tester reports and trade records to an AI agent as structured data
-
Backtest orchestration Research
An agent queuing Strategy Tester runs and collecting results
-
Live account telemetry Research
Read-only account state for monitoring research
-
AI Portfolio Manager Research
Combining systems by evidence — see the Portfolio Lab direction
Nothing on this page is a product. Items marked Research or Experimental may change or be abandoned, and no AI capability here predicts prices or guarantees outcomes.
05Research output
The measurable product of this pipeline is the public research log. Every entry states whether AI was involved, what it did, and that a human reviewed it — the same entries the log itself renders, so this section is checkable.
- 26Experiments
- 20AI-assisted (hybrid)
- 6Human only
- 0Fully AI
Every AI-assisted entry in the log has been reviewed and approved by a human. Entries generated by AI without human review are withheld by the build pipeline — a safety state, not a style choice.
Systems we have shipped.
Each Expert Advisor is documented with its concept, logic, risk model and test conditions. No black boxes, no unexplained results.