arbitrix-core

MIT-licensed open-source backtest engine and cost model from the Arbitrix trading toolkit.

Install

pip install arbitrix-core

Optional extras:

• arbitrix-core[fast] — enables numba JIT for the SL/TP vectorized loop.
• arbitrix-core[docs] — mkdocs build dependencies.
• arbitrix-core[dev] — pytest + ruff.

Quickstart

import pandas as pd

from arbitrix_core import Backtester, BTConfig, Signal, BaseStrategy
from arbitrix_core import costs
from arbitrix_core import load_ohlcv


class SmaCross(BaseStrategy):
    def generate_signals(self, df):
        fast = df["close"].rolling(10).mean()
        slow = df["close"].rolling(30).mean()
        signals = []
        for i in range(1, len(df)):
            if fast.iloc[i - 1] < slow.iloc[i - 1] and fast.iloc[i] > slow.iloc[i]:
                signals.append(Signal(when=df.index[i], action="buy", price=df["close"].iloc[i]))
            elif fast.iloc[i - 1] > slow.iloc[i - 1] and fast.iloc[i] < slow.iloc[i]:
                signals.append(Signal(when=df.index[i], action="sell", price=df["close"].iloc[i]))
        return signals


df = load_ohlcv("eurusd_h1.csv", time_col="datetime")
costs.configure(commission_per_lot=3.0, point_overrides={"EURUSD": 10.0}, allow_provider_lookups=False)
result = Backtester(BTConfig()).run_single(df, SmaCross(), risk_perc=0.01, initial_equity=10_000.0)
print(result.metrics)

Full documentation at https://g14mb0.github.io/arbitrix-core/

Sync to public repo

arbitrix-core is published from the upstream Arbitrix monorepo via subtree split. The private workflow .github/workflows/arbitrix-core-sync.yml force-pushes arbitrix/src/arbitrix_core to https://github.com/G14MB0/arbitrix-core on every push to main or development that touches the subtree, and on workflow_dispatch.

Requirements:

• Repo secret PUBLIC_REPO_TOKEN — fine-grained PAT with Contents: write scope on G14MB0/arbitrix-core.
• Public repo development branch is overwritten on every sync. Do not commit directly to that branch.