deptho

Deptho is the tick-level reality check for any backtest. Bring signals from any backtester or custom code, and Deptho replays them against months of real trades from Binance's public archive, entirely on a local machine. The free library shows funding and fees by position; Deptho Pro handles tick data bigger than memory.

pip install deptho
python -m deptho.demo               # 90 days of BTCUSDT perp data from Binance's archive, a sample strategy, a report
python -m deptho.demo --synthetic   # same, offline

A tearsheet that shows what costs took

import deptho as dp

pnl = dp.pnl_from_positions(prices, positions, funding=funding_rates, fee_bps=4)
dp.tearsheet(pnl=pnl, benchmark=prices.pct_change(), title="BTC perp momentum").show()

• Net vs gross equity, so you see exactly what costs took
• Cumulative cost drag broken out into funding, fees and slippage
• Funding charged to the position actually held at each funding time
• Underwater (drawdown) curve, rolling Sharpe and Sortino, rolling alpha against a benchmark
• Slippage distribution and slippage vs order size (pass fills=)
• report.save("report.html") writes one self-contained HTML file that works offline
• report.fill_gap_card("card") writes one shareable image (SVG, plus PNG with pip install "deptho[png]"): the bar-close return, the return with slippage measured by replaying the orders, and the gap between them in points

Cost attribution is lookahead-safe by default: a position decided at a bar's close earns from the next bar.

Inputs

Tick trades come from Binance's public archive (dp.datasets.read_aggtrades) or from Tardis.dev trades files you've downloaded (dp.datasets.read_tardis_trades), in the same shape either way.

Large files

dp.load(path) reads CSV or Parquet with pandas. It first measures a small sample, and if the full file won't fit in memory it raises dp.MemoryWallError with an estimate, instead of letting pandas exhaust RAM and kill your kernel.

Privacy

Deptho runs entirely on your machine and collects no telemetry. The only network call is a data download you start yourself, which goes to Binance's public archive and prints the host first.

Deptho Pro

Pro is for when the free tools run out: replaying orders against every trade across months of tick data.

• Sync months of Binance USD-M perpetual trades into one local, memory-mapped dataset that grows a day at a time and repairs itself after an interruption.
• Replay a table of orders (time, side, quantity) from any backtester against every trade print, lookahead-safe by default. Fills are modeled from trade prints; order book depth is coming.
• Sweep many order sets over the same dataset in tens of MB of memory.
• License checked offline. Nothing is uploaded.

Benchmarks, method and pricing: deptho.io. Turn off in-library Pro mentions with dp.config.upgrade_prompts = False or DEPTHO_NO_PROMPTS=1.

Links

• Website and benchmarks: https://deptho.io
• Guides: https://deptho.io/fix/
• Deptho Pro: https://deptho.io/pro.html

License: Apache-2.0. The card's PNG uses the bundled Space Grotesk and JetBrains Mono fonts (SIL Open Font License 1.1, in deptho/fonts/). Market data comes from Binance's public archive (data.binance.vision); Deptho is not affiliated with Binance.