Do popular trading strategies survive rigorous backtesting? We tested 24

Most popular trading strategies fail — even in one of history's greatest bull markets

Between 2010 and 2026, the S&P 500 (via SPY) returned +543%. It was one of the most favorable environments for long-only equity strategies in modern history. And yet, when we ran 24 popular retail trading strategies through a rigorous five-gate validation pipeline, 58% of them failed.

Here is the irony that makes this study worth reading: if a strategy can't survive a decade-plus bull run, it has no business being deployed in a bear market, a sideways market, or any real portfolio. The strategies didn't die because they were overfit. They died because of costs and statistical fragility — two forces that exist in every market regime, good or bad.

The bull-market caveat: why these results are probably generous

Before diving into the data, you need to understand the environment these strategies were tested in. This is not a footnote — it is what makes the study credible.

• SPY returned +543% over the full period (2010–2026).
• 89% of months were within 10% of all-time highs.
• Only 2 months were more than 20% below peak.

This means long-only strategies enjoyed a massive tailwind. If a strategy holds positions for any meaningful duration, a rising tide did much of the work. The 58% attrition rate is arguably inflated by the bull market — results would likely be worse in a bear or sideways market. The honesty is the point: even with the wind at their back, most popular strategies failed.

How we tested: a five-gate survival pipeline

We built a monotonic, cumulative funnel. Every strategy had to pass each gate to advance — there were no second chances, no selective reporting, no parameter re-tuning between gates.

The funnel (cumulative survivors)

Note: the per-gate pass rates above show how many of the strategies that reached each gate survived it — a separate diagnostic. The cumulative column tracks the monotonic decline from 24 to 10. For example, Gate 4 (Monte Carlo) killed 6 strategies outright, dropping the running total from 20 to 14.

The 14 deaths: what killed them

Every failed strategy fell into one of three categories:

Cost-killed: 7 strategies

These strategies were profitable on paper — positive Sharpe, positive returns — right up until we applied a realistic cost model. The edge was real but too thin to survive the friction of actual trading. High-frequency strategies with many trades per year were the most vulnerable.

Statistically fragile: 4 strategies

Monte Carlo simulation on out-of-sample returns revealed a >25% probability of ruin — meaning there was a real, non-trivial chance of catastrophic drawdown even though the average return looked acceptable. These strategies looked safe but carried hidden tail risk.

Never worked: 3 strategies

These strategies failed the very first gate. They were never profitable — not even in a naive in-sample backtest during one of the biggest bull markets in history. If a strategy can't make money in a market that went up 543%, it isn't a strategy.

The worst cost casualties

Three strategies stand out as dramatic examples of cost-driven collapse:

The Stochastic Oscillator is the textbook case: it generates many signals with a very small edge per trade. Each round-trip eats roughly 0.12% plus $2 in costs. Trade often enough and the friction compounds faster than the alpha. An in-sample Sharpe of +0.48 turned into −3.16 — not just unprofitable, but actively destructive.

The key insight: the strategies work — they just can't survive friction

Here is the most important number in this study: median Sharpe degradation from in-sample to out-of-sample was only 11.4%.

That means the strategies genuinely predict future returns. They are not overfit — an overfit strategy would show 40%, 60%, 80% degradation when taken out of sample. Instead, the edge transfers reasonably well. What kills these strategies is not statistical overfitting. It is the cost of acting on the signal.

This reframes the entire problem. The question was never "do these strategies work?" They do. The question is: is the edge large enough to overcome the cost of capturing it? For 14 of 24 popular strategies, the answer was no.

The 10 survivors

These strategies passed every gate — in-sample, out-of-sample, walk-forward, Monte Carlo, and realistic costs — and still showed positive Sharpe ratios at the end:

A few patterns emerge. The strongest survivors — 52-Week High Breakout (cost-adjusted Sharpe +0.68) and Golden/Death Cross (+0.57) — share a common trait: they trade infrequently, hold positions for extended periods, and capture large moves. Less friction per unit of alpha. Meanwhile, strategies like Aggressive MA Cross barely survived (+0.10 after costs) because faster signal generation means more trades and more cost drag.

The exact cost model

Every trade in this study was charged the following, based on realistic retail execution assumptions:

• Commission: $1.00 per trade (flat)
• Spread: 1.0 basis point (0.01%)
• Slippage: 5.0 basis points on entry + 5.0 basis points on exit

That works out to approximately 0.12% plus $2 per complete buy-and-sell cycle. These are not aggressive assumptions — they reflect what a retail trader using a discount broker actually faces. If anything, they are conservative for smaller accounts or less liquid instruments.

Methodology

• Data source: FMP EOD (end-of-day) price API
• Primary instrument: SPY
• ETF basket: 13 sector ETFs (SPY, QQQ, IWM, XLK, XLF, XLE, XLV, XLI, XLP, XLU, XLB, XLRE, XLC)
• Period: January 2010 – June 2026 (4,145 daily bars)
• In-sample: 2010–2021 (70%)
• Out-of-sample: 2022–2026 (30%)
• Walk-forward: 2-year rolling windows with 6-month step increments
• Monte Carlo: 1,000 bootstrap paths on OOS returns (seed = 42, fully reproducible)
• Survival criterion: profitable AND positive risk-adjusted return (Sharpe > 0) at every gate

Limitations and honest caveats

No study is complete without disclosing what it doesn't capture:

1. Bull-market bias. As detailed above, SPY returned +543% over the test period. This favorable environment likely inflated survival rates. In a prolonged bear or sideways market, fewer strategies would pass — potentially dramatically fewer.
2. ETF universe. The study used ETFs rather than individual stocks. ETFs inherently filter out bankruptcies and delistings, avoiding survivorship bias at the security level — but this also means the results may not generalize to single-name trading, where idiosyncratic risk, liquidity gaps, and corporate actions add another layer of cost and risk.
3. No tax modeling. The cost model accounts for trading friction but not short-term capital gains taxes, which can significantly reduce net returns for high-turnover strategies in taxable accounts.
4. Single-position sizing. Strategies were tested with fixed position sizing — no dynamic risk allocation, Kelly criterion, or volatility targeting. Optimal sizing could improve or worsen individual results.

The takeaway

Most popular trading strategies are not scams or snake oil. The signals work — the edge is real, and it transfers out of sample with only modest degradation. But an edge that exists on paper and an edge that survives the real world are two very different things.

The killers are mundane: $1 per trade, a basis point of spread, five basis points of slippage. Multiply that across hundreds of trades per year and it erodes everything but the strongest, slowest-moving signals. Before you deploy any strategy, ask the question this study forces: is the edge large enough to pay for the cost of capturing it?

If the answer is uncertain in a market that went up 543%, it won't get better when the tide turns.