A strategy that's gold in one market can be worthless in the next. That's not an opinion — it's what you get when you run the exact same strategy across two markets and change nothing else.
We took the classic RSI/SMA crossover — a simple trend-following strategy that goes long when momentum turns and to cash when it rolls over. Same rules, same default parameters, same window (2018 to June 2026), weekly chart. Once on the top 50 cryptocurrencies, once on the top 50 US stocks. The yardstick: average buy-and-hold — what you'd have made simply buying and holding, averaged across all entry points.
The result is almost indecently clear.
The headline
| Metric (RSI/SMA, 1W, 2018–2026) | Crypto top 50 | S&P top 50 |
|---|---|---|
| Avg strategy CAGR | 15.6 % | 7.5 % |
| Median strategy CAGR | 11.9 % | 4.6 % |
| Avg buy-and-hold CAGR | −5.0 % | 16.7 % |
| Beats buy & hold | 40 / 50 (80 %) | 2 / 50 (4 %) |
| Avg max drawdown | −68.8 % | −34.0 % |
| Avg Sharpe (annualized) | 0.07 | 0.27 |
| Avg trades per asset | 43 | 78 |
Same strategy. In one market it beats buy-and-hold on 4 out of 5 assets. In the other, on exactly two out of fifty. On stocks it doesn't lose narrowly — it gets crushed by plain buy & hold (16.7 % vs. 7.5 % CAGR).
Why it wins on crypto
Crypto is a rollercoaster with built-in catastrophes. Look at the buy-and-hold figure: averaged across all entry points, simply holding the top-50 coins since 2018 was negative on average (−5 %). Many alts never recovered after their peak. The average max drawdown of −69 % tells the rest.
In a market like that, a strategy that goes to cash in time isn't a return-chaser — it's insurance. And that's exactly where it earns its keep. Broken down by market regime:
| Regime (crypto) | Assets | Avg diff vs. B&H | Outperformers |
|---|---|---|---|
| Bearish (B&H < 0 %) | 30 | +27.6 pp | 28 |
| Sideways (0 to +10 %) | 5 | +23.5 pp | 4 |
| Bullish (B&H > +10 %) | 15 | +5.6 pp | 8 |
30 of the 50 coins were net losers over the window if you just held them. On 28 of those, trend-following limited the damage — by almost 28 percentage points of CAGR on average. That's the whole trick: in a market that regularly costs you 70–90 %, you win by not taking the biggest crashes in full.
The spread is still brutal. Best asset relative to its own buy & hold: a Fantom token at +165 percentage points. Worst: a −88-pp own goal where the strategy did worse than just holding. Trend-following on crypto isn't a calm business — it's just the lesser evil.
Why it bombs the S&P 500
Blue-chip stocks are the opposite of crypto: a stubborn, persistent uptrend with comparatively mild pullbacks (avg max drawdown −34 % vs. −69 % for crypto). Averaged across all entries, just holding returned +16.7 % CAGR. Step out of that and you don't dodge a catastrophe — you miss the recovery.
And that's precisely what happens. Broken down:
| Regime (stocks) | Assets | Avg diff vs. B&H | Outperformers |
|---|---|---|---|
| Bullish (B&H > +10 %) | 36 | −10.5 pp | 1 |
| Sideways (0 to +10 %) | 13 | −7.0 pp | 0 |
| Bearish (B&H < 0 %) | 1 | +4.3 pp | 1 |
36 of the 50 stocks were in an uptrend — and on 35 of those the strategy actively destroyed return, an average of 10.5 percentage points of CAGR. Stepping out was most expensive on exactly the biggest compounders: on a semiconductor heavyweight the in-and-out cost −29 percentage points versus simply holding, on a pharma name −22, on the big AI names double digits each. You can't "time" a secular winner — every cash trip during a pullback is a missed continuation.
Only two stocks beat their buy & hold at all, and both narrowly.
The uncomfortable nuance
Two things you have to say out loud, or the story is too clean:
First — the 80 % is partly a crypto artifact. The high hit-rate on crypto isn't because the strategy is brilliant; it's because the benchmark is on the floor. Buy & hold on crypto was negative on average. It's easy to beat something that loses itself. On stocks the benchmark is +16.7 % — a much higher bar.
Second — risk-adjusted, it flips. Average Sharpe is actually higher on stocks (0.27 vs. 0.07). Per unit of risk, the stock version delivers more; it's only beaten by the even-better buy & hold. The crypto version "wins" against a miserable benchmark but is wild in absolute terms and barely risk-efficient. "Beats buy & hold" and "is a good strategy" are not the same thing.
What this actually means
This isn't a strategy failure. It's a frequency mismatch. Trend-following lives off large, sustained moves in one direction — and off markets where stepping out before the crash pays. Crypto delivers both in abundance.
Stocks, on daily and weekly scales, are more mean-reverting: they overshoot to the downside and recover, rather than running clean medium-term trends. The real momentum effect in stocks lives on the 3-to-12-month scale, not the weekly chart. A weekly trend-follower is therefore in the wrong frequency bucket — it sells pullbacks that are statistically closer to buying opportunities.
Which leaves the obvious question on the table: so which strategies are built for stocks? The answer isn't "optimize trend-following," it's "switch categories" — toward mean-reversion. That's the material for the next posts: RSI(2), Capitulation Finder, Triple Strike Reversal — strategies that bet not on the trend but on the return to the mean. We're testing them on this exact S&P universe right now.
Limits of this analysis
- Survivorship bias: We used today's top 50 — the survivors. Dead coins and fallen stocks are missing, which makes both sides look kinder than they were. This only gets clean with a point-in-time universe.
- One strategy, one window, one interval. This illustrates the frequency argument; it's not proof that "trend-following never works on stocks." Other parameters or windows shift the numbers.
- Pre-cost. The engine runs without fees and slippage. At 78 trades per stock, that would drag the stock results further down — against the strategy, not for it.
- Under 30 trades is anecdote. Both sides average well above that (43 and 78), but don't over-read individual low-trade assets.
What Backtesting Arena adds here
You can reproduce this comparison on the platform itself: the same strategy, both asset classes, the same window, with avg buy-and-hold always shown alongside as the yardstick. We surface the drawdowns and the phases where a strategy loses, not just the ones where it shines — even when, as here, it's our own trend-following strategies that look bad on stocks. That's the point: evaluating a strategy means knowing which market and which market phase it works in — and which it doesn't.
FAQ
Does this mean trend-following is useless for stocks in general? No. On the weekly scale with this mechanic, yes — it sells pullbacks that mostly recover. But momentum effects in stocks do exist, just on longer scales (3–12 months) and with different mechanics (cross-sectional ranking rather than single-name crossovers). That's a different strategy family.
Why does the strategy beat so many coins if you say it isn't "good"? Because the bar is so low. Buy & hold on the top-50 coins since 2018 was negative on average. Beating something that loses is easy. Risk-adjusted (Sharpe), the crypto version is actually weaker than the stock version.
Is the comparison fair when crypto is so much more volatile? That's exactly the point: the same strategy meets two completely different market characters. The volatility is the reason trend-following adds value on crypto (catastrophe protection) and not on stocks (cash drag in an uptrend). We average across all entry points so no single lucky or unlucky date drives the picture.
Would fees flip the result? They'd sharpen it, not flip it. The stock version trades almost twice as often (78 vs. 43) and already trails before costs — with fees it trails further.
Why only the top 50 and not the whole market? Readability and data quality. Today's top 50 are also survivorship-biased — a point-in-time universe would make both sides look more realistic (and worse). We name that as a limit openly.
What's next? We're testing mean-reversion strategies (RSI(2), Capitulation Finder, Triple Strike Reversal) on the same S&P universe — the category built for the daily/weekly behavior of stocks. Results to follow in their own posts.