nanoback 0.5.3

C++20 event-driven multi-asset backtesting engine with Python APIs for analytics, WFO/Monte Carlo validation, corporate actions, tick replay, and execution realism.

BACKTESTER (nanoback)

BACKTESTER is a C++20 event-driven multi-asset backtesting engine with Python bindings, packaged as nanoback. It focuses on realistic execution simulation, data correctness, and statistically valid research workflows.

Highlights (v0.5.x)

• Fast C++ core with Python APIs
• Bar-mode and tick-mode simulation paths
• Corporate actions support: split, dividend, spinoff, delisting
• Smart execution realism primitives:
  • multi-venue routing model (fees, volume share, fill curves)
  • signal/order/fill latency modeling + adverse selection penalties
• Derivatives primitives:
  • instrument model for equity, options, futures, FX forwards
  • option expiry settlement, futures roll events, margin liquidation path
• Research validity stack:
  • analytics (Sharpe, Sortino, CAGR, drawdown, attribution)
  • parameter sweeps and heatmaps
  • walk-forward optimization
  • Monte Carlo shuffle/block-bootstrap stress tests
• Deterministic audit ledger and snapshot/resume support

Repository Layout

• include/nanoback: C++ public headers
• cpp: core engine and Python bindings
• python/nanoback: Python API, loaders, analytics, research modules
• benchmarks: performance and regression checks
• examples: runnable usage examples
• tests: regression and functional test suite

Quickstart

cd C:\Users\TAPESH\Documents\BACKTESTEER
python -m venv .venv
.\.venv\Scripts\Activate.ps1
python -m pip install -e .[dev]
python -m pytest

Core Usage

Basic Bar Backtest

import numpy as np
import nanoback as nb

result = nb.run_backtest(
    timestamps=np.array([1, 2, 3, 4], dtype=np.int64),
    prices=np.array([100.0, 101.0, 99.0, 102.0], dtype=np.float64),
    signals=np.array([1, 1, 0, -1], dtype=np.int64),
    config=nb.BacktestConfig(max_position=2),
)

print(result.summary())

Tick Replay

import nanoback as nb

config = nb.BacktestConfig()
config.data_mode = nb.DataMode.TICK

result = nb.run_backtest_ticks(
    timestamp_ns=timestamp_ns,
    asset=asset_idx,
    price=price,
    size=size,
    side=side,  # TickSide.BID / ASK / TRADE
    target_positions=targets,
    cols=n_assets,
    config=config,
    symbols=symbols,
)

Corporate Actions

import nanoback as nb

config = nb.BacktestConfig()
config.corporate_actions = nb.load_corporate_actions_csv("corp_actions.csv", symbol_to_asset)

Statistical Validation

import nanoback as nb

sweep = nb.Sweep(data)
grid = nb.ParamGrid({"lookback": [5, 10, 20], "max_position": [1, 2]})
res = sweep.run(strategy, grid, n_jobs=1, compiled=True)

wfo = nb.WalkForward(n_splits=6, train_frac=0.7, anchored=True)
wfo_res = wfo.run(data, strategy, grid, compiled=True)

mc = nb.MonteCarlo.from_backtest(result)
mc_res = mc.run(n_sims=5000, method="block_bootstrap", block_size=20)

Data Loaders

• load_csv, load_parquet for bar data
• load_ticks_parquet for tick event replay
• load_corporate_actions_csv for corporate action ingestion
• load_yahoo_adjusted for adjusted prices + suspicious jump warning checks

Benchmarks

Latency benchmark (CI-style):

.\.venv\Scripts\python.exe benchmarks\benchmark_engine.py --max-seconds 0.50 --min-fills 1000 --ci-mode

Stress benchmark (large shapes):

.\.venv\Scripts\python.exe benchmarks\benchmark_engine.py --mode stress --profile xlarge

Useful benchmark flags:

• --baseline / --update-baseline
• --regression-factor
• --stage-regression-factor
• --pnl-tolerance
• --fill-count-tolerance
• --memory-guard-gb
• --record-history --history-file ...

Notes

• This is a research and simulation platform, not a production OMS/EMS.
• Performance and determinism are prioritized, but you should still validate strategy-specific assumptions before live deployment.