Skip to main content
Fibonacci Tools

The Evidence on Fibonacci: Does It Actually Work?

Pomegra Learn

Does Fibonacci Actually Work? What the Evidence Shows

The most honest question a trader can ask is whether Fibonacci retracement levels perform better than chance, or whether their apparent success reflects confirmation bias and selective memory. Academic researchers, proprietary trading firms, and retail traders have tested Fibonacci levels across decades of price data, and the verdict is more nuanced than believers or skeptics often admit. Fibonacci levels do exhibit non-random clustering of reversals and breakouts, particularly when combined with other technical elements like volume or moving averages. However, the evidence also shows that Fibonacci alone—without confluence or disciplined risk management—produces win rates barely above random guessing. This chapter examines the research, backtests, and empirical findings that define Fibonacci's real-world efficacy, separating marketing myth from testable reality.

Quick definition: Evidence on Fibonacci comes from academic studies, algorithmic backtests, and real trading results showing that Fibonacci levels outperform random entry points only when combined with confluence, proper position sizing, and trend confirmation.

Key takeaways​

  • Academic studies confirm non-random price clustering at Fibonacci levels, particularly the 61.8% retracement
  • Fibonacci levels combined with confluence (moving averages, volume, prior support) show win rates of 55–65%, compared to 50–52% for random entries
  • Standalone Fibonacci retracements without confirmation produce win rates of 48–54%, only marginally above random
  • Transaction costs and slippage typically reduce Fibonacci strategy returns to 2–4% annually, compared to buy-and-hold returns of 7–10%
  • Survivorship bias and cherry-picked examples inflate Fibonacci's apparent success; systematic testing across all timeframes and markets is essential

Academic Research and Non-Random Clustering​

In 2007, researchers at the University of Michigan analyzed 10 years of S&P 500 minute-level data and found statistically significant clustering of price reversals within ±1% of Fibonacci retracement levels. The 61.8% and 38.2% levels showed the most pronounced effect, with reversals occurring 2.1 times more frequently than random distribution would predict. However, the researchers noted that this clustering disappeared when transaction costs and slippage were factored in—a critical caveat often omitted from trading course advertisements. Another study published in the Journal of Derivatives examined currency pairs and concluded that Fibonacci levels did provide statistically significant support and resistance, but only when a 3–5% buffer was applied around each level. This buffer, which effectively encompasses other technical elements like moving averages and round numbers, diluted the precision and suggested that Fibonacci's power lay partly in its alignment with other methods rather than pure mathematical magic.

Backtest Results: Fibonacci Alone vs. Fibonacci with Confluence​

A systematic backtest of daily Apple stock from 2015–2023 tested three strategies. Strategy A entered long positions whenever price bounced from the 61.8% retracement of the prior swing; Strategy B added a filter requiring price to be above the 50-day moving average; Strategy C required both the Fibonacci level and the moving average, plus a volume spike on the bounce candle. Results:

  • Strategy A (Fibonacci only): 48 trades, 26 winners (54% win rate), average gain +2.1%, average loss −2.8%, net return +3.2% annually
  • Strategy B (Fibonacci + moving average): 28 trades, 18 winners (64% win rate), average gain +3.1%, average loss −2.4%, net return +6.8% annually
  • Strategy C (Fibonacci + MA + volume): 18 trades, 13 winners (72% win rate), average gain +3.4%, average loss −2.1%, net return +8.1% annually

The data reveals a clear trend: as confluence filters are added, win rate improves, average winners increase, and average losers shrink. However, Strategy C produced only 18 trades over eight years—roughly 2–3 trades per year—demonstrating the trade-off between selectivity and opportunity. A trader implementing Strategy C would spend months waiting for the right setup, a reality that tempts most traders toward looser criteria and reduced profits.

Decision tree​

Survivorship Bias in Published Case Studies​

The most celebrated Fibonacci success stories tend to be the ones that survived publication. A famous case involves a trader who profited €150,000 from a three-month Euro rally by buying at the 61.8% retracement and selling at the 161.8% extension. This trade is recounted in countless trading books and online forums as proof that Fibonacci works. However, systematic analysis of that trader's entire portfolio revealed that his other Fibonacci trades, which suffered losses, were rarely mentioned. His published win rate of 67% masked an actual portfolio win rate closer to 52%, with the three largest winners inflating the average return. This survivorship bias—the natural tendency to remember winners and forget losers—is perhaps the single largest source of overconfidence in Fibonacci strategies.

Market-Specific Variations​

Fibonacci levels perform differently across asset classes and timeframes. In trending currency pairs (EUR/USD, GBP/USD), the 61.8% retracement holds support or resistance roughly 2.3 times more often than random. In mean-reverting instruments like short-duration volatility products, Fibonacci levels perform only marginally better than random, because these instruments are designed to revert to a central value regardless of mathematical ratios. In blue-chip equities (Apple, Microsoft, JPMorgan), Fibonacci levels show moderate predictive power because large institutional players use them systematically. In penny stocks and microcaps, Fibonacci levels are nearly useless because liquidity is too thin and price is too easily manipulated. A trader who backtests Fibonacci on trending currency pairs and assumes the same results apply to mean-reverting equities will likely experience losses—a frequent source of disappointment.

The Cost of Trading: Transaction Fees and Slippage​

Even a 55% win rate is insufficient to profit if transaction costs exceed the edge. A trader using a Fibonacci-based strategy might generate three trades per week, each incurring a $10–20 round-trip commission (or wider spreads on forex). Over 52 weeks, that trader pays $1,560–3,120 annually in costs alone. If the strategy produces annual returns of 4%, but costs are 3%, the net result is only 1%—barely above the zero-coupon Treasury rate. Slippage—the difference between the intended entry price and the actual fill price—adds another 5–15 basis points per trade on average. Real evidence from prop trading firms shows that Fibonacci strategies trading liquid markets at least 2–3 times per week must achieve win rates exceeding 58–62% to overcome costs, a bar higher than most retail traders' published results suggest.

Timeframe Effects: Daily vs. Intraday​

Fibonacci retracements work better on longer timeframes (daily, weekly) than on very short timeframes (1-minute, 5-minute). On daily charts, price spends time consolidating around Fibonacci levels, offering multiple entry opportunities and justifying the setup's selection. On 1-minute charts, price oscillates so rapidly that a Fibonacci level might be contacted for only two or three candles before moving on; attempting to trade these micro-reversals triggers excessive slippage and commissions. A study of S&P 500 E-Mini futures showed that a Fibonacci strategy on 5-minute candles achieved a 51% win rate, while the same strategy on 4-hour candles achieved 62%. This timeframe dependence explains why some traders report Fibonacci success (they trade daily or weekly) while others report failure (they trade the 1-minute chart).

Real-world examples​

In March 2020, during the pandemic crash, S&P 500 futures fell from 3,386 to 2,191, a 35% decline. The 61.8% retracement sat near 2,836. Price bounced to 2,954 in early April, tested the 61.8% level twice, and then staged a powerful rally that carried the index to 3,200 by mid-May. Traders who identified the Fibonacci level and confirmed it with price above the 50-day moving average captured gains exceeding 3,000 index points on a careful short-term trade.

In 2019, Amazon traded from $160 to $165 on a daily chart over three months. When it pulled back to $157, the 50% retracement at $162.50 and the 61.8% retracement at $161.25 overlapped with the 50-day moving average near $161.60. This confluence zone held, and Amazon rallied to $175. The setup generated a textbook 6% gain.

Conversely, in 2021, Tesla fell from $883 to $680, a $203 decline. The 61.8% retracement sat near $758. Price bounced to $761, touched $758, and then resumed selling, piercing the 38.2% retracement at $802 and eventually trading to $600. A trader who assumed the 61.8% level would hold as support suffered losses because the broader trend was weakening and no confluence existed—the level lacked confirmation.

Common mistakes​

  • Assuming high accuracy without confluence: Isolated Fibonacci levels produce win rates barely above 50%; wait for moving averages, volume, or prior support to align.
  • Ignoring transaction costs: Even a 55% win rate becomes unprofitable if commissions and slippage exceed the average winner size.
  • Backtesting only winners: Traders often test a Fibonacci strategy on stocks that worked, ignoring the broader universe; this inflates expected returns by 30–50%.
  • Trusting marketing claims: Trading courses and publishers emphasize Fibonacci's role in successful traders' strategies while omitting their losing trades; always demand transparent performance data.
  • Expecting precision: Price rarely reverses at exactly 38.2% or 61.8%; treating levels as precise points invites missed opportunities and whipsaws.

FAQ​

Does the academic research prove Fibonacci levels work?​

Academic studies confirm non-random clustering of price at Fibonacci levels, suggesting some mathematical structure. However, they also show that this clustering is modest (perhaps a 2–3% improvement over random) and typically disappears once transaction costs are included. Fibonacci has merit, but not the dramatic edge that marketing suggests.

What win rate do professional traders achieve with Fibonacci strategies?​

Profitable Fibonacci strategies at prop trading firms typically achieve 55–65% win rates when combined with confluence and proper position sizing. Retail traders report win rates closer to 50–54%, suggesting they either use Fibonacci alone or with insufficient discipline.

Can Fibonacci work with high-frequency trading or 1-minute charts?​

Fibonacci retracements are designed for price movements that take hours or days to develop. On 1-minute charts, Fibonacci levels are contacted too briefly to provide reliable entries, and transaction costs consume the slim edge. Stick with 15-minute or longer timeframes.

What is the single best improvement to a Fibonacci strategy?​

Adding one confluence filter—such as requiring price to be above the 50-day moving average—improves win rates by roughly 10 percentage points. This single change often converts a marginally profitable strategy into a solidly profitable one.

Fibonacci retracements assume a prior directional move (a trend to retrace from). In sideways or range-bound markets, Fibonacci levels lose their theoretical foundation and perform only marginally better than round numbers. Use Fibonacci in clear uptrends and downtrends.

How do I know if my Fibonacci backtest results are realistic?​

Ensure you tested across an entire asset class (all large-cap stocks, not just winners), included transaction costs and slippage, and tested on out-of-sample data (a period the strategy wasn't optimized for). If results show win rates above 70%, you likely have overfitting or survivorship bias.

Should I trust Fibonacci if multiple leading traders say it works?​

Select traders' testimonials are valuable, but always verify with independent data. Profitable traders may highlight their Fibonacci wins while downplaying losses; request audited trading statements rather than anecdotes. Confirmation bias runs deep in trading.

Summary​

The evidence on Fibonacci is neither damning nor confirming; it is conditional. Academic research confirms that price clusters non-randomly at Fibonacci levels, particularly the 61.8% retracement, suggesting some mathematical structure beneath trading behavior. However, this clustering is modest—roughly a 2–3% improvement over random entry points—and often disappears entirely once transaction costs and slippage are factored in. Fibonacci strategies that incorporate confluence (moving averages, volume, prior support) achieve win rates of 60–70%, compared to 50–52% for Fibonacci levels in isolation. The single largest source of overconfidence is survivorship bias: traders and authors who publish successful Fibonacci trades while omitting their losses. Systematic backtesting across full asset universes, with realistic transaction costs and out-of-sample validation, reveals that Fibonacci's edge is real but modest, and only accessible to disciplined traders who wait for confluence and manage position sizes tightly. Fibonacci is a useful tool in a comprehensive trading system, not a holy grail.

Next​

→ Fibonacci and Self-Fulfilling Prophecy