How to Detect Market Cycles: A Beginner’s Guide

If you have ever noticed that a market seems to make lows every few weeks or peaks at regular intervals, you have intuitively observed a market cycle. The question is: is what you are seeing real, or is it pattern matching on noise?

This guide walks through the process of detecting genuine market cycles using quantitative methods. No programming required—but understanding the concepts will help you interpret results from any cycle detection tool.

Step 1: Understand What You Are Looking For

A market cycle is a recurring oscillation in price that appears with statistical regularity. It has three measurable properties:

• Period — The length of one complete cycle (trough to trough or peak to peak), measured in bars. A 40-day cycle means price tends to oscillate from low to high and back to low every 40 trading days.
• Amplitude — The strength of the oscillation. A high-amplitude cycle creates large swings; a low-amplitude cycle creates barely noticeable undulations.
• Phase — Where you currently are within the cycle. Are you near a trough (potential low), a peak (potential high), or somewhere in between?

Detecting a cycle means determining all three properties from historical data.

Step 2: Prepare Your Data

Before detecting cycles, the data must be detrended. Markets have underlying trends (upward over the long term for stocks, for example). If you search for cycles in trended data, the algorithm will find the trend itself, not the oscillations around it.

Common detrending methods include:

• First-differencing — Subtract each bar's close from the previous bar's close. Simple and effective for highlighting short-term oscillations.
• Linear detrending — Fit a straight line to the data and subtract it. Good when the trend is approximately linear.
• Hodrick-Prescott filter — A more sophisticated filter that removes the smooth trend component while preserving cyclical variation. FractalCycles offers this as an option.

The choice of detrending method affects which cycles you detect. First-differencing emphasizes shorter cycles; HP filtering preserves medium-term structure. When in doubt, try multiple methods and see which cycles appear consistently.

Step 3: Run Spectral Analysis

Spectral analysis examines the detrended data in the frequency domain. Instead of asking "what happened at each point in time?" it asks "which periodicities are present in this data?"

The output is a power spectrum — a chart showing spectral power (y-axis) against period length (x-axis). Peaks in the power spectrum indicate frequencies where the data oscillates more strongly than expected by chance.

FractalCycles uses the Goertzel algorithm for this step, which is more efficient and flexible than the traditional FFT (Fast Fourier Transform) for trading applications. It evaluates spectral power at specific periods rather than computing the entire frequency spectrum.

Step 4: Validate with the Bartels Test

This is the step most beginners skip—and the most important one. Every dataset will produce peaks in the power spectrum, even purely random data. The question is whether those peaks represent genuine structure or noise artifacts.

The Bartels significance test answers this. It divides your data into segments equal to the candidate cycle length and measures the consistency of the pattern across segments. A cycle that appears consistently across all segments scores high; one that appears in some segments but not others scores low.

Interpretation guidelines:

• Below 50% — The cycle is more likely noise than signal. Do not rely on it.
• 50-70% — Marginal significance. The cycle may be real but treat it with caution.
• Above 70% — Statistically significant. There is strong evidence this cycle is genuine.
• Above 85% — Highly significant. Very strong cyclical structure.

Step 5: Examine the Results

After spectral analysis and Bartels validation, you will have a list of validated cycles ranked by significance and amplitude. What to look for:

• Dominant cycle — The cycle with the highest spectral power that also passes significance testing. This is the primary rhythm in the data.
• Harmonic relationships — Look for cycles whose periods form roughly 2:1 ratios. A 20-day and 40-day cycle pair is a common harmonic relationship that suggests genuine nested structure.
• Consistency across timeframes — A cycle that appears on daily and weekly charts (at proportional periods) is more trustworthy than one that appears on only a single timeframe.

Step 6: Build a Composite Projection

Once you have identified validated cycles, the final step is combining them into a composite wave. This sums the individual cycle waves into a single projection that shows when cycles reinforce each other (creating stronger turns) and when they cancel out (creating uncertain periods).

The composite projection extends forward beyond the last data point, providing a structural roadmap of expected cycle turning points. This is not a price prediction—it is a timing framework showing when the cyclical structure favors highs and lows.

Common Beginner Mistakes

1. Skipping detrending — Analyzing raw price data without removing the trend. The algorithm will detect the trend itself rather than oscillations around it.
2. Ignoring significance testing — Treating every spectral peak as a real cycle. Most peaks in most datasets are noise.
3. Using too little data — Trying to detect a 50-day cycle from 100 bars of data. You need at least 8-10 repetitions for reliable detection.
4. Overfitting — Including every detected cycle in the composite. Fewer, stronger cycles produce better forward projections than many marginal ones.
5. Treating projections as predictions — The composite shows structural timing, not guaranteed price movement. Use it as context, not as a standalone trading signal.

Try It Yourself

The fastest way to detect market cycles is with FractalCycles. The platform handles the entire pipeline—detrending, Goertzel spectral analysis, Bartels validation, and composite projection—automatically. You can analyze any stock, ETF, cryptocurrency, commodity, or forex pair.

Start with our free Cycle Period Finder tool to experiment with cycle detection, or create a free account to run full analyses with composite projections.