Do Elliott waves work? Separate three different tests

Distinguish reproducibility, structural agreement and trading performance before evaluating a wave hypothesis.

Student Wyckoff editorial team

Published: 23 Oct 2026 (UTC+3)

Русский · English · Deutsch · Español · Français · Italiano · 繁體中文 · 日本語 · 한국어

The question of whether a method works needs a stated task. Producing a repeatable interpretation, agreeing on a model and earning a trading return are different outcomes. A successful test of one does not establish the others.

Test reproducibility

Preserve the data, range and relevant settings, then repeat the procedure. Record whether the same conditions give the same reviewable result. This tests consistency under stated inputs. It does not establish that the interpretation correctly describes a complete pattern or forecasts later prices.

Test structural agreement

Ask reviewers to apply the same model definitions to the same information. Record which rules they checked and where they disagree. A disagreement may come from missing evidence or differing assumptions. Agreement itself also does not prove a future price outcome.

Test performance under a separate protocol

A performance study needs decisions fixed before the observations being evaluated, a defined sample and accounting for costs and execution. Keep failed and unresolved cases. Do not infer profitability from a set of charts selected after the outcomes became known.

Match the conclusion to the test

Report which question was tested, the data used and the limits. This article presents no measured win rate or software backtest. A small usability exercise can support a usability conclusion; a broader performance wave hypothesis requires its own evidence rather than a stronger label on the same exercise.

Two reviewers agree on one historical chart. Is a profitable strategy established?

No. That observation concerns agreement on that case. A properly designed performance evaluation is a separate task.

Two research papers: different tasks and limits

Marañon and Kumral (2018, Resources Policy 59, 125–138) study gold, silver, copper and a metal-price index, using Monte Carlo on detected waves. The publisher’s accessible abstract and excerpts give a cautious assessment of EWP for those commodity cycles. The full dataset’s exact dates and N were not established from the material read; this comparison does not itself establish mass-psychology causality.

Jarušek, Volná and Kotyrba (2022, Neural Networks 145, 342–355) combine neural networks, FFT and Elliott for a 60-minute EUR/USD forecast with four currency-pair inputs. The 30000 windows from 2015–2018 are training data, not independent TestN; the publisher’s excerpt explicitly excludes trading fees from the simulation. We read accessible abstract/excerpts, not the full article requiring paid access; complete test N remains unverified. Different tasks and procedures do not yield a universal Elliott accuracy or results for our product. Full titles and DOIs appear below.

Two research tasks: commodity cycles/Monte Carlo and a currency neural-network model/FFT. The comparison shows scope and limitations, not a common accuracy scale.

Can 30000 be described as an independent test sample, with the model’s findings transferred to every Elliott tool?

No. These are selected training windows for a specific model. Transfer requires a verified test protocol, sample, costs and external evaluation; two different publications do not supply that validation.

Reddit and Bitcoin videos: preserve the claim before judging it

A Reddit question or comment records a participant’s view, not a complete trading sample. The linked September 2026 discussion asks others to inspect a proposed count; replies are not an independently adjudicated ledger. For a Bitcoin YouTube analysis, save the exact channel and video URL, upload time, quote source, interval, completed-bar cutoff, original scenario and later edits. Distinguish a video posted before the move from an explanation assembled afterward. Views, likes and confident narration do not measure forecasting accuracy.

An official EWI article updated 11 June 2026 presents a selected Bitcoin video excerpt. It is a publisher’s retrospective presentation, not a denominator containing every original call, revision and unresolved case. Compare the original available information with the stated condition, then retain all qualifying outcomes and execution costs before measuring a trading result. This resource note does not claim that a particular paid archive was watched or endorse a clip’s forecast. Use the separate dated Bitcoin source record for prices, rather than treating comments or video popularity as market data.

Using experience reports, forums, Reddit, YouTube and Wiki

For introductory terminology, the German Wikipedia entry Elliott-Wellen is linked below. It is a changeable overview, not a replacement for the precise rule definition used in a count. The linked Reddit thread dated 25 February 2026 asks about practical experience with the method. It provides a concrete discussion entry point, not a representative survey or an accuracy measurement. Anonymous opinions and claimed profits are not treated as verified results.

For a forum post or YouTube video, save author, original publication time, venue-qualified instrument, interval, data cutoff, main and alternative counts, and invalidation condition. For video, also record the timestamp of the particular chart view and check whether later bars were already visible. Verify channel ownership against the author’s own website. EWI publishes its own promotional videos and articles; identifying that source is not independent approval or a ranking of YouTube channels.

Select an analysis published before the move and freeze its original version. Reveal only bars available at that time, inspect rules and alternatives, then open subsequent bars and record every revision. A post-event chart can illustrate an explanation without proving a prior forecast. Assess method usability, a specific prediction and a trading result after costs separately when reading experience reports; likes and selected winners do not supply a complete denominator.