Skip to main content

Glossary

Quant terms used on FinLab AI

Understand the numbers before they impress you. Fourteen terms, plain language.

Terms are ordered to mirror the FinLab AI backtest report — 1–2 sentences each, no jargon stacking.

CAGR (Compound Annual Growth Rate)
Total strategy return expressed as a single compounded annual rate. A 5-year cumulative return of 100% works out to a CAGR of about 14.9%. 完整說明與台股實例 →
Sharpe Ratio
Excess return per unit of risk (volatility). > 1 is solid, > 2 is excellent; the absolute number depends on the risk-free rate assumption used in the calculation.
Max Drawdown
The largest peak-to-trough decline in equity over the period. -30% means at the worst point your capital was down by a third — the key metric for whether you can sleep at night.
Backtest
Simulating a strategy on historical data. A great backtest doesn't promise future returns, but a bad backtest almost always means a worse future.
Win Rate
The share of trades that ended profitably. High win rate doesn't equal high returns — a strategy that loses big and wins small can have a great win rate and still lose money. 完整說明與台股實例 →
Turnover
How frequently the portfolio is rotated. 200% annual turnover means the whole portfolio is swapped roughly twice a year; high turnover eats into returns through fees and slippage. 完整說明與台股實例 →
Slippage
The gap between your intended price and the price you actually fill at. FinLab backtests build in fees and transaction tax but no slippage assumption; raise fee_ratio to stress-test costs so you don't overestimate strategy performance. 完整說明與台股實例 →
Alpha
Return earned above what the broader market explains. Alpha = 5% means the strategy beat the market-driven return by 5 percentage points. 完整說明與台股實例 →
Beta
How tightly a strategy moves with the market. Beta = 1 tracks the market, < 1 is more defensive, > 1 is more aggressive. 完整說明與台股實例 →
F-Score (Piotroski F-Score)
A 0–9 score across nine financial-health checks. Higher scores indicate cleaner, more conservative financials — a common quality filter for value strategies.
Momentum
The empirical tendency for recent winners to keep winning. Momentum strategies typically buy stocks that have outperformed over 3–12 months, rebalanced monthly or quarterly.
Mean Reversion
The empirical tendency for stretched moves to retrace. Mean-reversion strategies fade recent oversold names and tend to complement momentum strategies.
Quant Trading
Generating trades from statistical models and code rather than discretionary judgment. FinLab AI focuses on transparent, reproducible quant research.
Factor / Indicator
A quantifiable signal used to select stocks — e.g. P/E, YoY revenue growth, RSI. FinLab AI ships 900+ U.S. and Taiwan factors you can combine.
Look-ahead Bias (Point-in-Time)
Using information in a backtest that wasn't actually available at the time — e.g. fundamentals not yet aligned to their filing date, or values that were only revised later. It inflates backtested performance and can't be reproduced in live trading. Using point-in-time data aligned to the filing date avoids it; FinLab aligns its financial data to the filing date by default. 完整說明與台股實例 →