Why mean reversion wins more often than trend following

If you ask a hundred retail traders how often their strategy wins, the ones running mean-reversion setups will say 65 to 80 percent and the ones running trend-following setups will say 30 to 45 percent. They're both telling the truth. They're also describing two strategies that can have identical long-run expected value. The win-rate gap doesn't come from one being better than the other. It comes from the shape of the bet each one takes.

Mean reversion bets that price will snap back toward an average after stretching too far. Most of the time it does — small moves are noise, and noise reverts. So mean-reversion strategies win frequently and pocket small gains. The losses, when they happen, are the moves that didn't revert — moments when an extension turned into a real trend, and the strategy caught the wrong side of it. Few losses, but each one is bigger than a typical win.

Trend following bets the opposite: that a move that has started will continue. Most of the time it doesn't — most starts are head-fakes that pull back into the noise. So trend strategies lose frequently and book small losses. The wins, when they happen, are the rare moves that turned into actual trends, and those moves are large enough to pay for everything in between.

This post walks the structural reason for the win-rate gap, the academic research on both styles (which you'll be surprised to find supports both as real edges, not myths), what each style genuinely costs in the wrong market regime, and how to think about combining them. The interactive at the end runs a simulation of both strategies under matched expectancy so you can see the equity curves diverge in shape while converging in result.

The two archetypes, properly stated

Mean reversion in its purest form: identify when price has moved more than X standard deviations from a moving average, take a position betting it returns, exit when it does. The classic academic example is Bollinger-band reversion, but the family includes pairs trading, statistical arbitrage, RSI-extreme buying, and any strategy that buys lower lows or sells higher highs on the assumption that "extreme" prices are temporary.

The earliest formal academic evidence for mean reversion as a real market phenomenon is the DeBondt and Thaler 1985 paper in the Journal of Finance, which showed that the worst-performing stocks over a 3-5 year window outperformed the best-performing stocks over the subsequent 3-5 years — a long-horizon reversal effect. Subsequent research by Lo and MacKinlay extended this to shorter timeframes and confirmed that short-term reversion is a measurable feature of most equity markets.

Trend following in its purest form: identify when price has broken out of a recent range or moving-average band, take a position in the direction of the breakout, hold until the trend reverses or a trailing stop fires. The family includes momentum strategies, breakout systems, and the famous Turtle Trading rules from the 1980s. Most CTAs (commodity trading advisors) and managed-futures funds run some flavour of trend following.

The foundational academic paper here is Jegadeesh and Titman 1993) in the Journal of Finance, which showed that stocks with strong returns over the past 3-12 months continued to outperform stocks with weak returns over the same horizon. The momentum effect is one of the most robust anomalies in financial economics — it shows up across markets, across decades, and across asset classes. Cliff Asness's later AQR research extended momentum to bonds, currencies, and commodities and found similar effects.

Both are real. Both have empirical support. Both have produced multi-decade track records for the funds that run them well. The interesting question isn't "which is better?" — it's "why do they have such different win rates while delivering similar long-run returns?"

The reason for the win-rate gap

The mathematics is the same as the stop-placement geometry. Win rate is a function of where you put your exit relative to your entry, not a function of how good the strategy is.

A mean-reversion trade exits at the mean — typically a small distance from entry. The stop, by definition, has to be wide enough to contain the noise that's already pushed price to the entry point in the first place. That makes the target close and the stop far. Geometric expected win rate: high.

A trend-following trade exits at the end of the trend — by design, a large distance from entry. The stop is set just outside the breakout level so it triggers fast on a fakeout. That makes the target far and the stop close. Geometric expected win rate: low.

You're looking at the win-rate-vs-R:R seesaw, with mean reversion sitting at one end and trend following at the other. The math is symmetric: a mean-reversion strategy with a 75% win rate and 1:3 reward-to-risk has the same gross expectancy as a trend strategy with a 35% win rate and 3:1 reward-to-risk. Both produce the same expected value per trade. Both look like completely different beasts on a P&L curve.

This is also why the simple advice "high win rates are better" is wrong as a universal rule. A high win rate combined with a small R:R is structurally vulnerable to costs and to the rare losing trades that aren't small. A low win rate combined with a large R:R is structurally vulnerable to long losing streaks and to traders' discipline failing during them. They have different failure modes; neither is universally safer.

The shape of the curves

The mean-reversion equity curve is mostly smooth and upward-sloping, with occasional cliff-like drawdowns when reversion fails (a "trend break" — the move you bet would reverse instead kept going). The trend-following equity curve is mostly flat or grinding downward, punctuated by sharp upward jumps when a real trend gets caught. Both end at the same +18% in this stylised example. The journey is unrecognisable as the same destination.

The structural distinction is in the skewness of the return distribution. Mean reversion has negative skew — the tail risk is on the left, the rare events are bad. Trend following has positive skew — the tail capture is on the right, the rare events are good. This is why combining them in a portfolio is genuinely useful: their tail events tend to fire in opposite regimes, so the combined skew is closer to symmetric.

When each one works (and when each one breaks)

The honest framing of when each style underperforms its long-run average:

Mean reversion fails in trending markets. When the market is in a real trend, every "extension" you sell into keeps extending. Every "extreme" you buy keeps getting more extreme. A typical mean-reversion strategy in a strong-trend regime can give back 6-12 months of small gains in a few weeks of being on the wrong side. The 2017 BTC bull run, the 2020 March crash, the 2024 SOL run — these are the events that punish mean reversion specifically. Any mean-reversion trader who's been at it more than two years has lived through one of them.

Trend following fails in ranging markets. When the market is bounded between clear levels, every breakout fakes out before the trend the strategy needs to capture. The strategy enters at the top of ranges, gets stopped out, enters at the bottom of ranges, gets stopped out, and bleeds slowly through small losses. The 2015-2016 BTC sideways year, the 2018-2019 crypto winter range, the 2023 mid-year chop — these are the regimes that are catastrophic for trend strategies. Most trend funds have at least one historical drawdown of 20-30% during a sustained ranging period.

The corollary is useful: if you can identify the current regime with any reliability, you can adapt. If the market looks trending, lean trend; if it looks ranging, lean reversion. The catch is that regime detection is itself a hard problem — most academic methods (Hurst exponent, Markov regime-switching models) only confirm the regime after it's been in place for a while, by which point the regime is already shifting again. The retail-friendly version of regime detection — "is the market making higher highs and higher lows?" — works surprisingly well as a rough filter, but even that doesn't catch the transitions cleanly.

The simulation

The defaults run a 75%-win-rate mean-reversion strategy and its 25%-win-rate trend mirror, both with a +5 percentage-point edge above the geometric coin-flip baseline. The expected return per trade is essentially identical, but the equity curves will look completely different — the green mean-reversion line will mostly grind upward with sharp drops, the gold trend-following line will mostly grind sideways or down with sharp jumps. Re-roll the simulation a few times and notice that final returns vary widely for both styles even though the underlying probabilities haven't changed. That's small-sample variance — same point as the win-rate sample-size post.

The most interesting experiment: drop the edge to 0. Both strategies become break-even in expectation, and you'll see the mean-reversion curve still looks better most of the time — long stretches of small positive trades — until the inevitable drawdown that erases the gains. Visual smoothness is not the same as actual edge.

Why retail traders gravitate to one and pros to the other

Mean reversion has a strong psychological advantage for retail traders. The high win rate produces frequent small dopamine hits. The losses, when they come, are infrequent enough that they don't dominate memory. The equity curve looks clean. It's pleasant to live with.

Trend following has the opposite profile. The losing streaks last for months. The wins are irregular. There's no week-to-week feedback that the strategy is working, only the once-a-quarter event that makes the entire cycle pay. Most retail traders quit a trend strategy during one of the inevitable drawdowns and miss the recovery.

This is the structural reason most professional managed-futures funds run trend-following despite the lower win rates — they have institutional capital that can wait through drawdowns, they have the analytical infrastructure to size trades correctly during the rare large-payoff events, and they have client agreements that protect them from being fired during the sideways stretches. Retail traders have none of those. The strategy that fits institutional money structure-wise is exactly the strategy that retail traders are most likely to abandon.

That doesn't mean retail traders can't do trend following — Turtle-style and managed-futures-replication retail systems do work — but they require specific psychological skills that mean-reversion doesn't, mainly tolerance for long underperformance with the conviction the strategy will pay off.

For more on the broader behavioural framing of why high win rates feel safer than they are, the why-traders-lose-money post covers the full set of cognitive biases that make this exact mistake easy to fall into. To see how the same expectancy plays out under each style's variance shape, the annualized return projection tool runs the Monte-Carlo for any inputs you plug in.

Combining both

The diversification case is real. A portfolio that allocates capital to both mean-reversion and trend-following strategies has lower drawdowns than either alone, because the regimes that hurt one tend to help the other. AQR has published research showing the rolling correlation between momentum and value (a mean-reversion analogue) is reliably negative across decades, which is the technical version of "they cover each other's bad regimes."

The implementation is straightforward in principle: split capital, run both, rebalance occasionally. The catch is that the running-both requires more capital than running one, more cognitive bandwidth, more discipline, and more separate edges to maintain. Most retail traders are better served by picking one and running it well than by running two badly.

If forced to pick: mean reversion is generally easier for retail traders to execute consistently, has less drawdown variance, and has a long-term track record from the literature. Trend following has higher long-run returns in some studies but is much harder to stay with through the dry periods. Either choice is defensible. Neither is "objectively better."