Strategy Lab Overview

The Strategy Lab is where strategies get tested and, more importantly, where most of them should be thrown out. Each step is a chance to discard a weak idea before it costs you anything.

Where strategies come from

• Strategies: the strategies that ship with TigerTradeBot, and the community strategies you’ve added.
• Strategy Designer (admin): build your own with a visual rule builder for entries, exits and filters, or describe an idea to the AI assistant, which drafts the rules and critiques them. You don’t need to write code.

Write down what a strategy believes about the market before you test it. If you can’t state the idea in one sentence, the backtest results won’t mean much either.

Backtest

The Lab runs strategies through TigerTradeBot’s backtester over historical data:

• Stocks: Schwab price history, or TAAPI.io if you’ve configured a key.
• Crypto: Alpaca, or public exchange data (Coinbase, Binance, Kraken) that needs no key.
• Options: there’s no historical data for individual option contracts, so backtest the underlying instead.

A single backtest is a first look, not a result.

Optimize

Optimize sweeps parameter ranges on a grid; AI Optimize iterates on the parameters with your configured AI model. Optimization History keeps past runs. The goal is not the single best number: it’s a plateau, a region where many nearby parameter values perform similarly. A sharp, isolated peak is a classic overfitting signature.

The optimizer flags that for you:

• An Overfit risk badge (low, medium or high), based on stability, trade count, a suspiciously high win rate, parameters sitting at the edge of their range, and sharp score cliffs.
• A Stability badge and a Show stable results only filter.
• Parameter Sensitivity, which shows how much results move when each parameter changes.

Winning parameters are never applied automatically. Promote to Preset saves them as a named preset after you’ve reviewed them.

Compare

Compare Presets puts saved parameter sets for a strategy side by side, and Compare checks a result against a baseline. If a strategy doesn’t clearly justify its complexity after costs, stop here.

Validate

• Walk-Forward Validation splits your date range into calendar segments (four by default), backtests the same parameters on each one, and shows whether performance holds up or decays over time. It also compares in-sample segments with the held-back final segment (a 70/30 split by default). It doesn’t re-optimize inside each window, so run it after optimizing. Background: What is walk-forward testing?
• Multi-Symbol Validation runs the same preset across several symbols, so you can tell an edge from a lucky ticker.
• A Robust rating (robust, moderate or fragile) summarizes these checks.

Note: Passing validation is evidence for further testing, not proof of future results. Every number in the Lab comes from the past.

Then paper trade

Run the preset in a bot instance in Paper mode, in real time, on the built-in simulator. Let it run for weeks and compare its behavior with the backtest. Only then consider Live, which needs a license that allows live trading.

The honest summary

Most strategies fail somewhere along the way. That’s the process working. Trading involves substantial risk of loss and is not suitable for every investor; the Lab exists so you find out what fails on paper instead of live.