Strategy research
Test the strategy before it trades.
An operator turns the idea into code you can read, backtests it with costs, checks it for look-ahead and survivorship, and gives a verdict. Your desk decides what trades.
- ISO 27001
- GDPR
- Every run auditable
Hold the tech-sector ETF when it closes above its 200-day average, otherwise hold the Treasury ETF. Rebalance monthly, 2021 to 2025.
- signal = close > 200-day average
- weights shift one bar before returns
- costs: 5 bps per side on every change
- rebalance: first session of the month
- CAGR
- 24.1%
- Sharpe
- 1.62
- Max drawdown
- −14%
- CAGR
- 11.8%
- Sharpe
- 0.84
- Max drawdown
- −21%
- Look-ahead and same-bar executionFAIL
- Transaction costsWARN
- SurvivorshipPASS
- Parameter sensitivityWARN
- Sample lengthWARN
Signal and trade used the same close. Positions now shift one bar. None in the original. 5 bps per side added.
Waiting for the checks
Half the edge goes once the trade waits a bar and pays costs. Test the lookback on another period before anyone sizes it.
Share the run report in #quant-researchWaiting for a yes
Illustrative run. Research, not investment advice. Hypothetical past performance does not predict future results.
- Minutesfrom a strategy idea to a tested backtest
- Up to 90%less time on routine portfolio tasks
- 10–70%more capacity per person on the desk
Enterprises running operators on Alomana
What the operator does with one strategy idea
The idea
Turns a plain-English rule into code your team can read
The data
Pulls daily bars, corporate actions and filings for US equities and ETFs
The backtest
Runs it with a one-bar delay, costs on every change and a benchmark
The checks
Marks look-ahead, survivorship, costs and data snooping PASS, WARN or FAIL
The audit
Reviews a backtest someone hands you and says what it cannot support
The thesis
Argues against the idea and names how it would fail
How the operator tests a strategy
Connect
Market data over MCP, filings, your research files.
Run the test
Code, costs, checks and a verdict.
Decide
Your desk reads the verdict. Trading stays with you.
Research, not execution.
The operator stops at the verdict. Orders stay with your desk and your execution stack.
- Rules
- Code
- Backtest
- Checks
- Verdict
- Orders: your desk
- Built for research: placing orders is outside its job
- US equities and ETFs, daily data
- Every number computed in code, never guessed
- Every analysis ends: research, not investment advice
Nothing leaves the desk without a yes.
- Shares and writes wait for a person's yes
- A declined action is not retried
- Every run logged: input, output, tools, user
- Measured on
- Ideas tested per week
- Backtests sent back for revision
- Time from idea to verdict
- Hours rebuilding spreadsheets
- Works with
- Alpaca market data via MCP
- SEC filings and issuer sources
- Snowflake
- PostgreSQL
- SharePoint
- Google Drive
- Slack
- Your data vendor via MCP
- Trusted for
- ISO 27001
- GDPR
- Never trains on your data
- Single-tenant for enterprise
A custom operator typically goes live in up to 4 weeks.
Questions research desks ask
What is a strategy research operator?
An autonomous agent that does the research half of systematic trading. It turns a rule into code, backtests it with costs and a one-bar delay, checks it for look-ahead, survivorship and data snooping, and gives a verdict your desk can read and rerun.
Does it place trades?
No. It is built for research: it works from market data, filings and your files, and stops at the verdict. Orders stay with your desk and your execution stack.
Which markets and data does it cover?
US equities and ETFs on daily bars, through a market-data connection to Alpaca over MCP, plus SEC filings and issuer sources. A custom operator can work on your own price and trade history in Snowflake or PostgreSQL.
How does it stop a backtest from flattering itself?
Every signal acts one bar later, costs come off every position change, and each run marks look-ahead, survivorship, costs, data snooping and sample length as PASS, WARN or FAIL. A strong result with a failed check does not pass.
Can it check a backtest someone else built?
Yes. Give it the rules, the code or the reported results. It rebuilds what it can, marks what it cannot assess, and says which claims the evidence does not support.
Is our research data private?
Alomana is ISO 27001 certified and GDPR-compliant, and your data is never used to train AI models. Enterprise deployments run single-tenant, in a dedicated instance.
Bring one strategy you trust.
Talk it over with sales: what the operator needs from your systems, and how it goes live.
Alomana is a secure enterprise AI platform where autonomous agents, called operators, run business processes end to end. Every run is fully auditable.