Products / Strategy Lab
Inside the Terminal
Test an idea before you trust it.
The Strategy Lab replays a strategy through years of real history one rebalance at a time, showing it only what was public that day, charging costs on every trade, and grading the result with nine evidence tests.
- Point-in-time replay
- Nine evidence tests
- No code needed
What it is
Strategy Lab, explained.
A strategy in the Lab is a short pipeline: a universe of stocks, one step that scores them, a rule for turning scores into weights, and costs. The engine walks history, shows each decision only the rows that were public that day (a filing counts from the day after it was filed), prices trades a day later, and reports the result net of costs against SPY and against holding the whole universe, split into in-sample and out-of-sample.
The Lab lives inside TellHawk Terminal under My Tools. Single-ticker strategies can also be run in BaaS.
Who it is for
People with an idea
Momentum, reversal, low volatility or insider buying: pick one off the shelf, or describe your own, and see what honest testing says.
Skeptical investors
The verdict banner and the nine tests are designed so a weak result is as visible as a strong one.
Learners
Seven starter strategies, each with its real verdict, show what the evidence looks like. Most do not pass every test.
Key features
What it does today.
Describe a strategy, read it back in English
Choose a step from the shelf (momentum, reversal, low volatility, insider buying, news tone), one of the 41 chart-rule strategies, a machine-learning ranker, currency patterns, or your own read-only SQL query. The Lab writes the strategy back as a plain-English sentence.
- No code needed; SQL and machine-learning steps are there if you want them
- Fees of 1 bp and slippage of 5 bp by default, both editable
- Risk overlays (trend, volatility target, fixed) move part of the book to cash and never add leverage
Point-in-time, so it cannot peek
Every table a step can read is replaced, for each decision date, by a view of only the rows public that day. A step may be a single read-only SELECT; writes, PRAGMA, ATTACH and stacked statements are refused. Short interest is hidden for 10 days after settlement.
- Look-ahead tests run against the real database
- Each rule is proven causal before it runs: adding later bars never changes an earlier signal
Nine evidence tests, and a luck check
A result is judged on nine plain checks, a deflated Sharpe that is shrunk for how many versions you tried, and a bootstrap that asks whether luck could explain it. A stress test re-prices the strategy under changed assumptions.
- Out-of-sample tile and verdict banner on every run
- Year-by-year and rolling views
- Promising ideas can be frozen and tracked forward on data that did not exist yet
Scoreboard and library
The Scoreboard ranks every strategy by its evidence; the library holds published strategies replayed at their published settings.
Starters
Seven read-only worked examples, each with its idea, what it tests and how to read the result. Open a copy to change one.
Blend and compare
Combine finished runs into a blend and compare them side by side.
Share with the team
Send a strategy and the summary of its latest replay to the TellHawk team with a note. It is never published or added to anyone's Lab automatically.
Taking an idea further
A Lab strategy is a portfolio-level pipeline, so it is not deployed or published directly. To publish or automate the same idea, describe it as a single-ticker MA-crossover, RSI or rule-builder strategy and take it through BaaS, the Marketplace and Trader.
A few real runs, honestly
These are real walk-forward runs over real history, split into 4 folds that are scored one at a time and checked against a random-entry baseline with the same trade count and costs (0.05% per side). Beating random entry is a low bar: it is not the same as beating the market.
| Run | Period | Folds beating random entry | Folds beating buy-and-hold |
|---|---|---|---|
| NVDA · MA Crossover (20/50) | Sep 2016 to Sep 2026 | 4 of 4 | 0 of 4 |
| AAPL · RSI Mean-Reversion (14) | Sep 2016 to Sep 2026 | 4 of 4 | 1 of 4 |
| TSLA · MA Crossover (10/30) | Sep 2016 to Sep 2026 | 2 of 4 | 1 of 4 |
| NVDA · RSI Mean-Reversion (14) | Sep 2016 to Sep 2026 | 1 of 4 | 0 of 4 |
Backtests replay history after estimated costs; they are not a forecast and do not predict future returns. Run your own in the Strategy Lab or BaaS.
How it works
4 steps, from first look to result.
A real screen capture of Strategy Lab.
Pick a starting point
Open a starter, a shelf step or your own SQL, and set the universe, costs and rebalance period.
Run the replay
The Lab replays history decision by decision and shows progress live.
Read the verdict
Check the out-of-sample result, the nine evidence tests and the luck check before the headline return.
Stress it and track it
Run the stress test, then freeze today's picks to watch them on new data.
Which tool do I use?
Four tools, four different jobs.
They overlap on purpose, but each answers a different question. Here is how they differ and how they fit together.
Research workbench · you are here
Strategy Lab
- Best for
- Deep research on a portfolio-level idea.
- How it tests
- Replays a whole universe of stocks through history one rebalance at a time, showing each decision only what was public that day, net of costs.
- You get
- Nine evidence tests, a luck check, a deflated Sharpe and a stress test.
- Does it trade?
- Never. It tests ideas; it places no orders.
- Access
- Inside TellHawk Terminal. Free: 5 backtests a day.
Quick check
TellHawk BaaS
- Best for
- A fast check of a simple idea on one symbol.
- How it tests
- Walk-forward folds for a moving-average crossover, RSI or rule-builder strategy on a single ticker.
- You get
- Fold-by-fold scores and the four-lens validation gate.
- Does it trade?
- Never. A passed result can be sent to Trader.
- Access
- Sign-in required. Free: 5 backtests a day. It is also the engine the Marketplace uses to re-test a strategy when it is published.
Find and share
TellHawk Strategy Marketplace
- Best for
- Finding or sharing a strategy that has already passed the gate.
- How it tests
- Every listing was re-run on the server through the same walk-forward backtest and gate before it was listed.
- You get
- Profit factor, Sharpe, win rate, folds passed and the gate result for each listing.
- Does it trade?
- Never. Run in BaaS and Deploy to Trader open those tools with the strategy pre-filled.
- Access
- Free to browse and to publish. Paid listings are priced by their authors.
Run it on your account
TellHawk Trader
- Best for
- Running a validated strategy on your own account.
- How it tests
- Automation runs a gate-passed moving-average, RSI or rule-builder strategy on a connected paper account with the supported brokerage. A stop-loss is required and every order passes the same pre-trade risk check.
- You get
- Pause or stop at any time; every run is logged.
- Does it trade?
- Automation trades on paper only. Live orders are placed by you, with a confirmation step.
- Access
- Free and Pro plans. One brokerage is supported today.
- Research in the Lab, or check fast in BaaS
- Find or publish it in the Marketplace
- Run it in Trader, on paper
The route works for single-ticker moving-average, RSI and rule-builder strategies, the types the server can re-run and verify. A Lab strategy is a portfolio-level pipeline, so it is not published or deployed directly: describe the same idea as one of those types to take it through BaaS, the Marketplace and Trader. A backtest replays history and does not predict the future.
Limits
What to know before you use it.
- Survivorship: the warehouse holds essentially no delisted companies, so results are flattered by an unknown amount. The Lab charges a stated allowance (100 bp a year by default), which is a guess, not a measurement.
- Costs are assumptions; shorts assume borrow is available; dividends are credited but not reinvested; there is no calendar-date rebalancing or open-price execution.
- Point-in-time fundamentals are not available yet, and intraday strategies (opening-range breakout, VWAP) cannot be tested.
- A backtest replays history and does not predict the future. Research only, not investment advice.
Pricing
Start free.
Everything is free during the beta.
Free
- 5 backtests a day
- Blends of up to 3 strategies
- Starters, Scoreboard, library and the full evidence tests
Pro
- Unlimited backtests and blends
- Everything in Terminal Pro
The Lab is part of TellHawk Terminal and uses the same plans. Everything is free during the beta.
Full plan comparison on the pricing page.
Related
Works well with.
Strategy Lab
Describe a rule in plain steps, or pick one off the shelf, and replay it through years of real history before you trust it.
Research only, not investment advice. Backtests replay history and do not predict the future.