fastquant

What it is

fastquant is a thin Python wrapper over Backtrader whose pitch is a backtest in three lines: pull a DataFrame, call backtest("smac", df, fast_period=15, slow_period=40), read the final portfolio value. Nine strategy aliases ship with it — rsi, smac, emac, macd, bbands, buynhold, sentiment, custom and ternary — plus a multi mode that ORs several together. Pass a range instead of a number for any parameter and it grid-searches the combinations and returns them as a sorted DataFrame.

Everything underneath is Backtrader: the event loop, the sizers, the analyzers and the matplotlib plot. What fastquant adds is the defaults, and the defaults are what to check.

Data & coverage

get_stock_data calls yfinance for daily Yahoo bars and falls back to the Philippine Stock Exchange through the phisix API; get_crypto_data calls ccxt, Binance by default, at 1w, 1d, 1h or 1m. It ships no data of its own — it wraps other people's free endpoints, which is where the rot shows first.

Integrations

Backtrader, pandas in and out, ccxt for crypto. backtest(channel="slack") posts each signal to a Slack incoming webhook read from SLACK_URL, and an SMTP email helper does the same. The custom strategy takes a column of model output, which is how the examples wire in Prophet forecasts.

Limitations

• Dormant. Last commit 15 September 2023; last release January 2023; 77 open issues.
• The PyPI build does not import on pandas 2 or later. The fix has sat on master, unreleased, since June 2023.
• get_yahoo_data renames an Adj Close column and expects flat columns. Current yfinance returns neither: download() now defaults to auto_adjust=True and a MultiIndex.
• The licence is ambiguous. The LICENSE file shipped in both the repo and the wheel is MIT, but setup.py's classifier is GPLv3, so that is what PyPI advertises. Nobody is left to resolve it, which matters if you plan to reuse the code.
• The primary phisix endpoint, 1.phisix-api.appspot.com, returned 503 on 13 September 2026. The code's documented fallback to phisix-api2.appspot.com still answers, current and dated quotes alike, so the PSE path survives. The sentiment helper that scrapes a businesstimes.com.sg search URL does not: that URL now 404s.
• Cheat-on-close is forced on for every run. Orders fill at the close of the bar that generated the signal, which the source comment itself calls "technically impossible". Commission defaults to zero.
• Daily bars, one instrument, no live or paper trading, no intrabar fills.

Alternatives

Backtrader itself, if you want the engine without the wrapper. Backtesting.py for the same "few lines" ergonomics on a maintained codebase, VectorBT for vectorised parameter sweeps, NautilusTrader if the goal resembles production.