Trading Strategy Backtesting Benchmark Report

This report compares published Stingray replay artifacts using a common evidence checklist. It is a benchmark of the backtesting workflow, not a ranking of profitable strategies and not a representation of live account performance.

Report version: 2026-07-18. Results remain tied to the stated rule, source, interval, window, and replay semantics. New evidence should be published as a new version rather than silently replacing these rows.

Published benchmark table

Example	Window	Eligible fires	Published outcome	Baseline or control	Main limitation
BTC funding-rate squeeze	2026-04-20 to 2026-05-18	37 after a 1h cooldown	24h average +0.41%; median +0.55%; 62% positive	Non-overlapping 24h illustration: +8.3% versus BTC buy-and-hold +4.9%	Short window; outcome is not fee- or execution-adjusted P&L
BTC RSI/Bollinger/volume breakout	2025-04 to 2026-04	109 after a 1h cooldown	Average +0.03% at 1h, -0.26% at 4h, +0.17% at 24h	The weak 4h result acts as evidence against the loose breakout thesis	Signal replay, not a full entry/exit portfolio simulation

The first example shows a condition worth further out-of-sample monitoring. The second demonstrates the equally valuable outcome: a plausible breakout rule can look weak once its fires are counted consistently.

What is held constant

Every benchmark entry should publish:

• The original thesis and a readable version of the typed rule
• Market, venue, source, interval, and replay window
• Warmup and coverage notes
• Raw, distinct, and cooldown-filtered fires where available
• Forward-return horizons and sample sizes
• Baseline or control comparison where it is meaningful
• Explicit exclusions for fees, slippage, sizing, leverage, and execution

The full contract is documented in the Backtesting Methodology.

What the figures mean

Forward returns measure price changes after eligible fires. They do not prove that a trader could enter at the sampled price, hold exactly to the horizon, or realize the published average after costs.

That distinction is consistent with the SEC’s investor guidance: backtested performance is hypothetical and does not reflect actual performance. See the official Investor Bulletin on performance claims.

Benchmark 1: funding-rate squeeze

The rule combined negative Hyperliquid BTC funding with at least +0.5% Binance BTCUSDT momentum over 60 minutes. The replay used five-minute evaluation and a one-hour cooldown.

It produced 150 raw matching windows and 37 cooldown-filtered alerts. This gap is important: persistent truth is not the same as 150 independent opportunities.

Hyperliquid’s official documentation says funding is paid hourly and provides a historical funding endpoint. The source contract therefore needs to preserve the hourly funding observation while the combined rule is evaluated. See Funding and the perpetuals information endpoint.

Read the full funding-rate example.

Benchmark 2: breakout rule

The rule combined RSI above 65, a touch or break of the upper Bollinger Band, and volume at least twice its 24-hour average on the one-hour chart.

After cooldown, it fired 109 times, or about once every 3.3 days. The negative average at four hours and only modest positive averages at one and 24 hours are useful rejection evidence. Adding more automation would not repair an underspecified edge.

Read the full breakout example.

How to cite this report

When quoting a result, include the rule name, window, eligible-fire count, horizon, and the fact that it is a historical signal replay. Do not shorten a forward-return observation into a claim that the strategy “returned” the same amount live.

For the machine-readable source list, use Strategy Examples JSON. For guided practice, use the No-Code Backtesting Tutorial.