How have crypto technical indicator signals performed?

Current source-backed snapshot

A familiar indicator formula does not prove an edge. This young study begins in September 2026. Its main estimate keeps only the earliest event for each symbol inside each rolling 24-hour window, reports excluded overlaps, and retains the all-events estimate as a sensitivity check. Results remain exploratory even when a group reaches n≥20: this is an auditable description of a short recorded sample, not a strategy-performance claim.

Input documentation: CoinNudge Market Events sample · CoinNudge API documentation · CoinNudge methodology

Observation window: Published Market Events v1 history since September 2026

Calculation cadence: Auto-refresh about every 21600 seconds

Calculation version: published-event-outcomes-2.1

Download current dataset: JSON · CSV Raw values retain the dataset's published precision. Free fair-use limit: 60 requests per minute per IP, shared across all Research JSON and CSV endpoints.

Observation evidence

First 30 of 917 observations. JSON and CSV contain all published evidence. CSV dataset_section distinguishes summary and observation rows.

History begins in September 2026 and is not yet a long-horizon performance claim. Main estimates keep the earliest event per symbol in each rolling 24-hour window. The control baseline is the median of each retained event's same-symbol control median, weighted once per control-covered retained event; it is not matched by volatility, hour or regime. The deterministic circular bootstrap resamples consecutive UTC-day clusters in blocks of 3 for 1,000 iterations, preserving all retained symbols inside each sampled day.

How to read this page

• What happened after CoinNudge's recorded technical events, by indicator and direction?
• History starts in September 2026, reference prices exclude fees and fills, and same-symbol controls are not matched by volatility, hour or regime. Treat every result as exploratory until the calendar history matures.
• Inspect the table before relying on the summary.
• JSON and CSV preserve the published evidence.

What can this page tell you quickly?

Question answered

What happened after CoinNudge's recorded technical events, by indicator and direction?

Measured scope

Market Events v1 history since September 2026

Update schedule

Recalculated about every 6 hours from stored source observations.

Do not infer

History starts in September 2026, reference prices exclude fees and fills, and same-symbol controls are not matched by volatility, hour or regime. Treat every result as exploratory until the calendar history matures.

Why does indicator popularity not establish profitability?

RSI, MACD, EMA and Bollinger formulas are public and widely used. Performance depends on market, timeframe, trigger definition, execution and sample selection. A page that only explains the formula cannot answer what happened after a specific recorded event.

CoinNudge therefore binds outcomes to immutable event rows and exposes the row-level evidence. The table is useful only within its stated markets and observation period.

How are direction-adjusted return, MFE and MAE different?

Direction-adjusted return makes an upward event positive when price rises and a downward event positive when price falls. MFE is the best reference-price excursion in the event direction; MAE is the worst adverse excursion during the same window.

Neither excursion is an executed trade. A trader may not enter at the reference price or exit at an extreme, and fees, spread, latency and slippage are absent.

Why mark groups below 20 as exploratory?

Percentages from five or ten observations can move dramatically after one new outcome. A visible threshold prevents a small group from looking as reliable as a mature sample.

Even n≥20 does not create independence because event windows can overlap and crypto regimes cluster. Use the downloadable rows for stronger controls and uncertainty estimates.

How can I verify and cite this result?

Download the page JSON for the complete snapshot and CSV for tabular evidence. Record the API dataset key `technical-indicator-performance`, the returned calculation_version, observation interval and named source alongside any quoted number. The visible table rounds values for readability; the downloads retain the published precision.

The stable URL remains unchanged when the data refreshes. A new calculation time does not prove every input changed, so compare source-observed timestamps and the stated market events v1 history since september 2026. Missing rows remain missing rather than becoming zero or a favorable estimate.

What is measured, and what is not?

Method and data boundary

Read immutable Market Events v1 releases and deduplicate by row_id. For each indicator and direction, retain complete events in time order and keep only the earliest same-symbol event in each rolling 24-hour window. Publish raw N, overlap exclusions, non-overlap N, direction-adjusted median and positive share. The deterministic circular bootstrap resamples consecutive UTC-day clusters in blocks of three for 1,000 iterations, keeping all retained symbols in each sampled day together. The baseline is the median of each retained event's same-symbol control median, weighted once per control-covered event; publish event-minus-baseline and its interval without calling it causal. Full raw_event_ledger publishes every retained/excluded event with reason, excluded_by_row_id, elapsed hours and 1-based within-group order. raw_event_hash is SHA-256 of ledger-order UTF-8 row_ids joined by LF without trailing LF. Seed label is indicator|direction; difference seed XORs 0xA5A5A5A5. Three-day clusters span all calendar days including empty days and preserve all same-day symbols. Require 20 finite values and 5 calendar days; an empty draw suppresses CI. Deterministic contract: RNG xorshift32 uses unsigned 32-bit x ^= (x<<13)&0xffffffff; x ^= x>>17; x ^= (x<<5)&0xffffffff. Seed is first four SHA-256 bytes of UTF-8 seed_label, unsigned big-endian; zero initializes as 1. Advance RNG once per block; start=state modulo calendar-day count; circular consecutive days, concatenate and truncate to calendar-day count. Carry state across 1,000 iterations. Median averages the middle two for even N. CI uses sorted estimates and linear interpolation at h=(N-1)*p for p=.025/.975. Numeric seeds are public in JSON/CSV.

History starts in September 2026, reference prices exclude fees and fills, and same-symbol controls are not matched by volatility, hour or regime. Treat every result as exploratory until the calendar history matures.

Read the complete CoinNudge methodology

Sources and verification

• CoinNudge Market Events samplePublished event, context and forward-outcome fields.
• CoinNudge API documentationFree Research endpoints and paid data-delivery boundaries.
• CoinNudge methodologyOutcome and missingness contracts.

Frequently asked questions

Want this condition monitored for you?

See every Pro alert category, qualifying rule and Telegram delivery example before subscribing.