← Back to Blog · 2026-10-04 · 8 min read · Strategy Case Study

Picture this: it's late 2022. Lithium carbonate — the white powder powering every EV battery on the planet — has been on a two-year vertical run. Spot prices in China have climbed from roughly 50,000 yuan per ton in early 2021 to somewhere around 600,000 yuan per ton by November 2022. Every dip has been bought. Every "this is the top" call has been wrong. If you were a mean reversion trader, the playbook was obvious: fade the spikes, buy the pullbacks, collect.

Then the market didn't pull back. It collapsed — and it kept collapsing for months. Spot prices fell from that ~600,000 yuan peak to somewhere in the neighborhood of 180,000 yuan per ton by spring 2023. Traders who bought "oversold" conditions all the way down were destroyed in stages, each bounce looking like the bottom, each bottom failing.

Here's the twist that makes this a genuinely useful case study: lithium carbonate futures didn't even exist during the 2022 spot crash. The contract only launched on the Guangzhou Futures Exchange (GFEX) in mid-2023 — after the worst of the collapse. So the futures market opened into a falling, structurally broken commodity, and mean reversion traders who carried over the "buy every dip" habit from the bull era got a brutal education in regime change.

Let's break down what actually happened, why mean reversion failed so spectacularly, and what concrete rules you can take from it into your own trading of Chinese commodity futures.

A Quick Primer: What You're Actually Trading

Before the lessons, the specs — because trading a market you can't describe is gambling with extra steps.

Note that last row. A 1-ton contract with a 20-yuan tick is small — accessible to retail — but the daily price limits mean a gap against you can move fast. Compare that to the workhorses most international traders start with in China: rebar on the Shanghai Futures Exchange (10 tons per lot, 1-yuan tick) and iron ore on the Dalian Commodity Exchange (100 tons per lot, 0.5-yuan tick). Those contracts have night sessions and deep liquidity; lithium carbonate is newer, more speculative, and structurally more violent.

That context matters for everything that follows.

What Actually Happened: A Structural Collapse, Not a Correction

The 2021–2022 lithium rally was driven by a genuine demand explosion — EV adoption in China roughly doubled year after year — colliding with supply that takes years to bring online. Classic bottleneck economics. Spot prices went parabolic.

But parabolic markets built on a temporary supply shortage contain the seeds of their own destruction. By late 2022:

• New supply from Australian mines and Chinese lepidolite operations was ramping hard
• EV subsidies in China were winding down, pulling demand forward and then yanking it away
• Downstream battery makers had stocked up aggressively and suddenly stopped buying — destocking hit all at once
• Prices had reached levels where substitution, thrifting, and demand destruction all became rational

When the turn came, it wasn't a dip in an uptrend. The entire pricing regime changed. Every rally was a chance for producers to hedge and holders to exit, not a resumption of the trend. That's the environment where mean reversion dies.

And when GFEX launched lithium carbonate futures in mid-2023, the contract listed into this post-peak world. Traders who assumed "new contract, hot commodity, dips get bought" were fighting the last war. The futures price ground lower through 2023 and into 2024 as the supply wave kept arriving — a slow bleed punctuated by sharp countertrend rallies that mean reversion systems interpreted as "the bottom is in." It rarely was.

Why Mean Reversion Fails in Regime Change

Mean reversion is not a bad strategy. It's a strategy with a specific habitat. Let's be precise about where it works and where it doesn't.

Where mean reversion works

• Range-bound, two-way markets: Think rebar oscillating around a stable cost curve, or iron ore trading in a policy-managed band. Prices anchor to production costs and inventory cycles.
• Liquidity-driven dislocations: When a move is caused by positioning flushes rather than fundamental repricing — the 2020 oil crash into negative territory was an extreme example, where the collapse was substantially a delivery-mechanics and storage-crisis event layered on a demand shock. Reversions after forced-liquidation spirals are often tradeable.
• Markets with hard anchors: Cost of production, state intervention, or physical parity give price a floor and ceiling to revert toward.

Where it fails

• Structural supply/demand shifts: When the anchor itself is moving — a cost curve collapsing because new supply is flooding in — "oversold" is just "correctly priced, one step behind."
• Parabolic tops: The same verticality that generates your reversion signals also signals regime fragility. The 2021 thermal coal episode in China is instructive: prices went vertical, the state intervened directly with production increases and price controls, and the market snapped violently downward. A mean reversion trader fading the spike upward got run over; one buying the post-intervention crash without understanding the policy shift also got hurt.
• New, thin contracts: Early-stage futures markets attract speculative flows that don't care about fundamentals. Signals are noisier, gaps are more common, and the "mean" itself is unstable.

The lithium case is the cleanest recent example of all three failure modes stacked on top of each other.

The Core Lesson: Separate Volatility From Regime

Here's the mental model I'd tattoo on every aspiring trader's monitor: mean reversion assumes the mean is stable. Your first job is to test that assumption, not to trade the signal.

A z-score of -2 on a 20-day lookback tells you price is stretched relative to recent history. It tells you nothing about whether "recent history" is still relevant. In lithium's collapse, price was "oversold" on almost any lookback for months — because the mean itself was falling faster than any short-term oscillator could reset.

Practical regime filters worth considering:

• Long-horizon slope: If the 60-day (or longer) moving average slope exceeds some threshold — say, a daily percentage decline your instrument historically sustains — disable long-side reversion entries. You're in a bear regime; fade rallies or stand aside.
• Realized volatility expansion: When annualized realized vol doubles or triples versus its own trailing baseline, the market is repricing something. Mean reversion edge historically shrinks in high-vol regimes across most commodities.
• Fundamental anchor check: For Chinese commodities, this means watching inventory data, cost curves, and — critically — policy. Beijing's interventions in coal, and its industrial policy swings in sectors like steel and EVs, can override any technical signal. If you trade rebar or iron ore, you're trading a market where policy is a first-order variable, not a tail risk.

Practical Rules for Applying This to China Futures

Let's turn the case study into a checklist you can actually run.

Rule 1: Classify the market before you classify the signal

Before any mean reversion trade in Chinese commodity futures, answer one question in writing: what is the anchor price is reverting to, and is it stable? For rebar, it's a mix of iron ore/coke costs and construction demand. For lithium carbonate, in 2023–2024 the anchor was a falling supply-driven cost curve. If you can't name a stable anchor, you don't have a reversion trade — you have a trend trade you're afraid to admit to.

Rule 2: Never average down on a structural thesis

The lithium collapse punished dip-buyers in stages precisely because each bounce looked like confirmation. A hard rule: if your reversion entry is stopped out and you want to re-enter, the re-entry must be justified by a new signal, not by the old thesis at a "better" price. Two losses on the same idea in the same direction means the regime filter failed — go flat and re-evaluate.

Rule 3: Size for the gap, not the tick

With daily price limits in the high single digits and no night session on GFEX lithium contracts, your true risk is the limit-down gap, not your stop. If a limit move against your full position would exceed your daily risk budget, cut the size until it wouldn't. This applies doubly to newer contracts and to any position held through Chinese policy announcement windows.

Rule 4: Respect the calendar and the session structure

China futures markets have session structures and holiday calendars (Golden Week, Lunar New Year) that create long closures. A reversion position carried through a week-long holiday is a bet on information you can't react to. Either flatten before major closures or size down as if the gap risk in Rule 3 just doubled.

Rule 5: Demand asymmetry from countertrend trades

Fading a collapse can work — but only with defined risk and a target that justifies it. A reasonable framework: risk a fixed fraction per trade (many traders use something in the 0.5–1% range of account equity), target at least 2x the risk, and use a time stop. If a reversion trade hasn't worked within N bars — pick N from your backtest, not your mood — the reversion isn't reverting. Exit. In a regime like lithium's 2023 downtrend, time stops alone would have saved most dip-buyers from the slow bleed.

How to Pressure-Test This Before Risking Money

Everything above is hypothesis until it survives contact with real data. A few concrete steps:

• Backtest across regimes, not just the fun ones. Test your reversion logic on Chinese commodity futures through 2020 (oil crash spillover), 2021 (thermal coal intervention), and 2023–2024 (lithium's grind lower). If your equity curve only survives the quiet years, you've learned your strategy's habitat — which is genuinely useful knowledge.
• Model the microstructure honestly. Include realistic margin, tick value (20 yuan per tick on a 1-ton LC contract adds up fast on volatile days), slippage on limit days, and holiday gaps.
• Forward-test on live data under evaluation conditions. This is where most traders skip a step that matters. Running your rules under a fixed risk framework, with real market data and real drawdown consequences, exposes the gap between your backtest discipline and your actual discipline. That's exactly what a structured futures evaluation is for — and it's why we built XS Select around it: you can test a China futures system on real historical and live data, with evaluations starting from $29, before you ever put meaningful capital at risk.

The Takeaway

Lithium carbonate's collapse wasn't a failure of mean reversion as a concept. It was a failure to notice that the concept's core assumption — a stable mean — had quietly died somewhere between a spot peak near 600,000 yuan and a futures contract listing into a bear market.

The traders who survive in Chinese commodity futures long-term aren't the ones with the cleverest oscillator. They're the ones who classify the regime first, size for the gap not the tick, respect policy as a first-order force, and validate everything against real data before real money is on the line.

Mean reversion is a tool. Lithium taught the market — expensively, in public — what happens when you bring a range tool to a regime war. Test your system accordingly.