Backtesting ArenaBacktesting Arena
← Back to blog

We Fixed Our Fibonacci Strategy — and Retired It Anyway

Auditing 40,000 backtests, our Fibonacci strategy came up short. Before cutting it we asked: is it our config? A better exit made it 20 points better — and we retired it anyway. Why "beats Buy & Hold" isn't the bar.

Backtesting Arena·July 1, 2026·4 min read·47 views
We Fixed Our Fibonacci Strategy — and Retired It Anyway

Auditing 40,000 backtests, our Fibonacci retracement strategy was one of the weakest. The obvious move would be to quietly delete it. Instead we asked the question most platforms skip: is this the strategy — or is it our configuration?

The answer is instructive, because it's both.

The suspect: a fixed target

Our Fibonacci strategy buys the pullback to the 61.8% retracement in an uptrend and sells at a fixed target: the 1.272 extension. That's textbook — and exactly the problem. A fixed target caps the upside. In a crypto bull market where Buy & Hold just keeps running, a fixed-target strategy exits while the market runs away.

So we tested it instead of asserting it.

The repair test

Same entry logic, different exit: a trailing stop (ATR chandelier) instead of the fixed target — "let winners run." Across eight liquid crypto pairs, daily bars from 2018, versus Buy & Hold (pre-cost):

VariantMedian ΔCAGR vs B&Hbeats B&H
v1 (fixed 1.272 target)−29.5 pp2 of 8
Trailing exit−9.6 pp3 of 8

The fixed exit really was a bottleneck — about 20 percentage points sat in it alone. The skepticism was warranted.

Why we're retiring it anyway

Here's where it gets interesting. Two things argue against keeping it, even with the better exit:

1. Repaired isn't enough. Even with a trailing exit, the strategy stays negative on liquid majors (beats B&H on 3 of 8). Unlike our squeeze strategy, which a correct rebuild flipped into positive territory, Fibonacci doesn't clear the bar.

2. The nominal "beats B&H" is flattered. Across all 61 pairs tested, Fibonacci edges Avg B&H at the median — +1.9 percentage points. Sounds like edge. It isn't: the average B&H of those pairs is −9.7%. The strategy "wins" on dead alt-coins by sitting in cash while Buy & Hold collapses. Losing less is not an edge.

Run-weighted across the full production history, it sits at −9.2% median vs B&H.

The real bar: our own roster

Even the best Fibonacci variant is far below the strategies we already run. On crypto, a simple EMA cross beats Buy & Hold by a median of +24 points, Supertrend by +23 — both with fewer trades. A strategy that at best plays breakeven and needs ~50 trades per run to do it is negative after trading costs — and it costs a slot that a better strategy deserves.

That has become our standard: "beats Buy & Hold" is not the bar. The bar is whether a strategy earns its place against our best. If it doesn't, it's ballast — and ballast dilutes the signals that matter.

What happens to the data

The strategy leaves the backtester, but not the platform: the accumulated backtest data stays visible in Strategy Insights and the Strategy Library, now tagged with the retirement finding and the numbers. The insight is part of the product — anyone who wants to know why Fibonacci pullbacks don't work on crypto finds the answer, with the data.

What Backtesting Arena contributes here

We audit our own strategies as hard as anyone else's. Retiring a strategy that nominally beats Buy & Hold is a hard call — and it keeps the platform honest: we crystallize the few strategies that earn their place instead of maintaining a long menu where half of them net nothing after costs. Fewer strategies, more signal.

FAQ

You're retiring it even though it beats Avg B&H — isn't that contradictory? The nominal +1.9% is flattered by dead alt-coins (whose B&H is −9.7% — "losing less" isn't edge). On liquid majors the strategy loses. And even read positively: breakeven before costs at ~50 trades is negative after fees. "Beats B&H" alone isn't our standard — earning a slot against our best strategies is.

Was it a bug? No. The fixed 1.272-target exit was a weak design choice, not an error — and a trailing exit lifted the result by ~20 points. But even repaired, the strategy doesn't clear the bar.

Can I still see the data? Yes. The backtest data stays in Strategy Insights and the Strategy Library, tagged as retired, with the numbers. It's just no longer offered in the backtester.

What does "wins by sitting in cash" mean? When a coin loses 90% over years, any strategy that goes to cash in between is "better than Buy & Hold" — it just loses less. That's not a usable advantage, it's an artifact of dying assets.

Will it come back? Only if a future version clears the bar against our roster — not as a revival of the same logic.

Try it yourself

Run the backtest with your own parameters and time ranges.

Run backtest →

More on this topic

Strategy

Does the Ichimoku Cloud Work on Crypto? Buying Only Above the Cloud Earned Less on All Ten Large Coins

Backtesting Arenatradingstrategies.work

Ichimoku is usually traded with 9, 26 and 52 and with the rule "only buy above the cloud". Both were tested on crypto from 2018 to October 2026. The cloud rule lowered the result on all ten of today's largest coins, and on bitcoin and ether the traditional numbers did no better than randomly drawn ones. What that means for your settings.

Trend followingBacktestingMethodology+1
Oct 15, 20261 min
Strategy

When should you switch off a trading strategy? Popular kill rules flag good strategies almost as often as bad ones

Backtesting Arenatradingstrategies.work

A popular framework says: cut a strategy's position in half when its 90-day Sharpe drops below the backtest's 5th percentile, pause it at 1.5x the backtest drawdown. Simulated, those rules cut eight in ten healthy strategies to half size within three years and leave four in five strategies without any edge running through their first year. What to check instead.

Risk managementBacktestingMethodology+2
Oct 12, 20261 min
Strategy

Is the Anchored VWAP Strategy Worth It? Checked Against Six Years of Bitcoin

Backtesting Arenatradingstrategies.work

The anchored VWAP strategy draws an average-price line from a starting point you pick, such as the last low, and buys while price stays above it. Its result depends on that starting point. On six years of Bitcoin, none of five common anchor rules placed the line better than chance, and the 52-week low did worse than 99 percent of random anchors. Why that happens, why the charts still look convincing, and what to use instead.

BacktestingMethodologyTrend following+1
Oct 7, 20261 min
Strategy

A Backtest With a 100% Win Rate: We're Testing It — Rules Set Before We Compute

Backtesting Arenatradingstrategies.work

A backtest claiming a 100% win rate and 34 million percent returns is circulating. We're rebuilding it — and setting the rules before we compute.

BacktestingLook-ahead biasMethodology
Jul 14, 20261 min
📬

Don't miss new blog posts

One short email per new post — strategies, backtests, market analysis. No spam, unsubscribe with one click anytime.

By subscribing you accept our privacy policy. We use Resend for delivery. Double opt-in confirmation required.

Comments (0)

Join free to post comments.

Sign up →

No comments yet. Be the first!