Noon Barbari

Open data · CC BY 4.0

Research & open data

We publish the raw data behind our research so anyone can check the numbers, rerun the analysis or build on it. Free to reuse with attribution.

Crypto strategy curve-fitting dataset (July 2026)

Every parameter combination of 10 classic strategy templates, backtested on 20 crypto pairs on daily bars since 2021 — each scored twice: on the first 70% of history (in-sample) and on the final 30% the configuration never saw (out-of-sample). One row per engine run.

11,440
engine runs
5,720
parameter configurations
10 × 20
templates × coins
70/30
chronological IS/OOS split

Downloads

Findings and methodology are in the write-up: We backtested 5,720 strategy configurations — most of what looks good is curve-fit

Schema

ColumnTypeDescription
templatestringStrategy template (10 house templates, default-parameter grids)
coinstringSpot pair on Binance, e.g. BTC/USDT (20 coins)
combo_idintParameter combination index within the template's grid
windowstring"is" = in-sample (first 70%), "oos" = out-of-sample (last 30%)
sharpefloatAnnualised Sharpe ratio of daily returns in the window
pnl_pctfloatNet return % in the window (from $10,000, standard fees)
max_dd_pctfloatMaximum drawdown % in the window
tradesintClosed round-trip trades in the window
params_jsonjsonThe exact parameter overrides for this combination

License & citation

The dataset is released under Creative Commons Attribution 4.0 (CC BY 4.0) — use it in articles, papers or products, just link back. Suggested citation:

Noon Barbari (2026). Crypto strategy curve-fitting dataset: 11,440 backtest runs across 10 strategy templates and 20 coins with a 70/30 chronological split. https://noonbarbari.xyz/en/research

@misc{noonbarbari2026curvefit,
  author = {Noon Barbari},
  title  = {Crypto strategy curve-fitting dataset: 11,440 backtest runs
            across 10 strategy templates and 20 coins},
  year   = {2026},
  url    = {https://noonbarbari.xyz/en/research},
  note   = {CC BY 4.0}
}

The dataset describes hypothetical backtests over the stated past period. It is research data, not investment advice, and past performance does not guarantee future results.