Describe and build
Start with an idea or enable Autopilot. An AI research agent develops the hypothesis, an AI coding agent writes the strategy, and verification checks its implementation.
How Stratmill works
How your AI strategy factory works.
Follow one connected lifecycle: idea generation, AI coding, verification, backtesting, optimization, risk review and paper trading. Supported live execution is a separate step that requires your authorization.
See the strategy factory in motion.
Illustrative example · not a live result
Watch AI agents build and review a strategy.
AI agents develop the idea, write code and assess the test results. Review findings can send the strategy back for another iteration.
Watch the path, not just the endpoint.
An illustrative equity curve reveals setbacks along the way. Inspect drawdown and trading costs before interpreting a result.
Keep open positions in the picture.
Compare account equity with closed-trade balance. An open position can change equity before a trade closes.
- Describe
- Build & verify
- Backtest
- Stress-test
- Paper-trade
EquityClosed-trade balanceDrawdown
Net return—
Max drawdown—
Checkpoint—
Test and review
Backtesting and optimization engines produce the evidence. AI analysis and risk agents assess the findings and can send the strategy back for revision.
Admit and observe
Approved work enters paper trading within available slots. Review its behavior on new data and keep real execution separate.
What to bring to your first strategy run.
A useful brief explains the market behavior you expect, the instrument and timeframe, what triggers a trade, how the position exits, and what would invalidate the idea. Start narrowly enough to understand the evidence.
Give it a direction
Describe the edge and its risks.
Review the work
Read reports, trades and decisions.
Decide what comes next
Iterate, abandon or follow approved paper work.
A strategy development pipeline you can follow.
Each stage produces work for the next review. A failed check can send an idea back for revision or end the run.
Define
AI research agent
A hypothesis with a market, rules and risks.
Build
AI coding & verification
Strategy code and implementation checks.
Test
Backtest & optimization
Trades, costs and validation results.
Challenge
AI analysis & risk review
Findings, exceptions and a review decision.
Admit
AI portfolio review
A release decision tied to the strategy version.
Observe
Paper trading
Simulated positions, trades and equity on new data.
The route is a simplified overview. Optimization, iteration and admission depend on the run and its findings.
Know what to open, and why.
Follow a result back to the assumptions and decisions that produced it.
Scroll horizontally to see all columns.
| Open | Look for | What it helps answer |
|---|---|---|
| Hypothesis | Market, timeframe, expected edge and risks. | What is this strategy trying to test? |
| Strategy report | Code access where permitted, checks, trades and performance. | Does the implementation match the idea? |
| Iteration history | Changes, review feedback and earlier attempts. | What changed between experiments? |
| Paper record | Positions, fills and equity after admission. | How is the strategy behaving on incoming data? |
Interactive illustration
See how review narrows the field.
A research process should make it easier to stop weak ideas. Explore a simplified cohort and the evidence needed at each checkpoint.
Ideas reaching each checkpoint
- Share of the original cohort
Start with a testable hypothesis, clear rules and a reason the effect might exist.
Check that the code implements the hypothesis and that the required data is available.
Review costs, out-of-sample results and robustness before deciding whether to continue.
Admission also requires review and an available paper slot. Observe the approved version on new data.
The workflow extends beyond a backtest.
Explore the components around strategy development: data, forecasts, portfolio decisions, release controls and feedback from prior work.
Market data
Review source coverage and point-in-time assumptions before interpreting a result.
Explore market dataAlpha & signals
Evaluate forecast information, costs and overlap before combining signals.
Explore alpha & signalsOptimization & validation
Inspect sensitivity, walk-forward results and a reserved holdout.
Explore optimization & validationPortfolio construction
Compare correlations, weights, risk contributions and effective bets.
Explore portfolio constructionLive trading
Authorize supported execution and monitor orders, positions and reconciliation.
Explore live tradingResearch memory
Available prior outcomes and recorded lessons inform future AI hypotheses and revisions.
Explore research memoryQuestions, answered.
More questions and answersWhere do I start?
Create an account and describe your idea in your workspace. You can inspect public strategies and market pages before signing in.
What do plan limits mean?
Active-run limits control concurrent research work. Paper slots control admitted paper strategies. Monthly credits and daily launch limits are separate constraints.
What should I include in my first research brief?
Describe the market and instrument, the timeframe, the behavior you expect, entry and exit conditions, position sizing and the risks. Explain what evidence would make you reject the idea. A focused question is easier to evaluate than a broad request to find a profitable strategy.
Why might a run be waiting instead of progressing?
Concurrent-run limits, daily launch limits, agent availability, a pause or a pending review can affect progress. Check the run stage and history for context. Launching an idea does not guarantee that every stage starts immediately.
Can I pause or stop my research?
Yes. Owners can pause, resume or stop eligible runs from the strategy detail page. Pausing holds the run for later; stopping ends it. If the strategy has entered paper trading, inspect its position record rather than assuming the control closed every simulated position.
What happens when a check fails?
Depending on the finding, the run can return for revision, wait for another review or be abandoned. Review the verification feedback, analyst findings and iteration history to understand the decision. A failed experiment can still help you refine the original question.
Start with a question worth testing.
Explore published strategies, or open a workspace to build, test and monitor your own with AI agents.
Stratmill is an AI trading strategy platform, not financial advice or a broker. Backtest and paper results are hypothetical. Trading involves risk of loss.