backtest_report#

Labels the output of the BacktestReport node. It takes a backtest engine's [equity, pnl, position, cost] output (the four positional columns that BacktestPriceTarget and the other backtest engines emit) and returns (running, summary).

  • running is a dict of numpy arrays: the four engine columns plus the BacktestReport node's drawdown (dollar), cum_cost, turnover (units traded), trades (count), max_drawdown (running worst), sharpe (running), and equity_held (equity carried across skipped bars). Each is a causal series whose last value is the summary.

  • summary is a dict of the final statistics: total_pnl, max_drawdown, total_cost, turnover, num_trades, and sharpe.

The aggregation runs in the C++ BacktestReport node, so a pure-C++ user gets the same statistics by calling that node directly. This wrapper only labels its columns and reads the last row, and needs no pandas. Wrap running in a pandas.DataFrame yourself if you want a frame.

Signature#

backtest_report(values, index=None)

values is the (T, 4) array a backtest engine emits. index optionally adds an "index" array to running for labeling the rows.

Example#

import numpy as np
from screamer import BacktestPriceTarget, backtest_report

price = 100 + np.cumsum(np.random.default_rng(0).standard_normal(500) * 0.3)
signal = np.sign(np.random.default_rng(1).standard_normal(500))

running, summary = backtest_report(BacktestPriceTarget(spread=0.0005)(signal, price))
for name, value in summary.items():
    print(f"{name:13s} {value:10.4f}")
print("\nrunning columns:", list(running))

total_pnl       -14.3679
max_drawdown    -17.6393
total_cost       13.1233
turnover        529.0000
num_trades      265.0000
sharpe           -0.0941

running columns: ['equity', 'pnl', 'position', 'cost', 'drawdown', 'cum_cost', 'turnover', 'trades', 'max_drawdown', 'sharpe']
On this page