Data: every backtest, free to download
All 31,407 results behind this site, under a Creative Commons Attribution 4.0 licence. Use them, check them, build on them; credit SpreadAtlas with a link.
Files
- spreadatlas-backtests-2026-10.zip 1015 KB
Every setup on every bar length, CSV with a README
- backtests-m1.json 111 KB
1-minute bars, compact JSON used by the explorer
- backtests-m3.json 417 KB
3-minute bars, compact JSON used by the explorer
- backtests-m5.json 410 KB
5-minute bars, compact JSON used by the explorer
- backtests-m10.json 404 KB
10-minute bars, compact JSON used by the explorer
- backtests-m15.json 395 KB
15-minute bars, compact JSON used by the explorer
- backtests-m30.json 397 KB
30-minute bars, compact JSON used by the explorer
- backtests-m60.json 393 KB
1-hour bars, compact JSON used by the explorer
- backtests-m240.json 382 KB
4-hour bars, compact JSON used by the explorer
Columns in the CSV
timeframe_min | Bar length in minutes: 1, 3, 5, 10, 15, 30, 60 or 240. |
|---|---|
type | single (one indicator) or combo (with a filter or exit rule). |
family | Indicator rule family. |
trigger | The indicator rule and its settings. |
filters | Filters added, separated by semicolons. |
exit | Exit rule added, or 'signal only'. |
trades | Number of trades, all coins together. |
win_rate | Share of trades that made money after costs (0 to 1). |
cagr_after_fee_and_slip | Yearly return after the fee and one tick per trade (0.10 = 10%). |
max_drawdown | Worst fall from a previous high, after fee and slippage. |
sharpe | Annualised Sharpe ratio after fee and slippage. |
train_cagr, test_cagr | Yearly return in the first and the second period, after fee and slippage. |
worst_coin_cagr | Yearly return of the weakest coin. |
cagr_fee_only, train_cagr_fee_only, test_cagr_fee_only | The same returns with the fee only. |
passed_two_period_test | 1 if the setup passed the two-period test, otherwise 0. |
In the CSV, returns are fractions (0.10 means 10%). In the JSON files used by the explorer they are percentages. How the numbers were produced, and their limits, are on the methodology page.
How to cite
SpreadAtlas (2026). Crypto indicator backtests on Upbit, October 2026. https://spreadatlas.com/datasets/. CC BY 4.0.
What it is not
These are backtests on past prices with modelled costs. They are not trading advice, and they flatter the past in ways described on the methodology page. Exchange price data belongs to the exchange; we publish only our derived results.