← Back to Blog · 2026-09-17 · 7 min read · Strategy Case Study

Picture this: it's late 2023, and you're watching lithium carbonate futures on the Guangzhou Futures Exchange bleed lower week after week. Your RSI(2) strategy — the one that printed money fading dips in rebar and iron ore — keeps telling you the market is "oversold." You buy. It keeps falling. You buy again. It falls again. By the time the market finally stabilizes, your drawdown has eaten through two months of profits.

If that scenario feels painfully familiar, this case study is for you. We're going to dissect what happened to classic mean reversion strategies during the lithium carbonate collapse, why a strategy that works beautifully in range-bound Chinese commodity futures fell apart in a structural bear market, and — most importantly — what filters you can add so it doesn't happen to you again.

The Setup: One of the Sharpest Commodity Collapses of the Decade

First, the context, because the lithium carbonate story is genuinely unusual even by China futures standards.

Lithium carbonate — the key ingredient in EV batteries — went through one of the most dramatic boom-bust cycles in modern commodity history. During the 2021–2022 EV boom, spot prices in China climbed to extraordinary levels, peaking around roughly 500,000–600,000 yuan per ton in late 2022. Then the wheels came off. Chinese lithium supply expanded rapidly, battery-grade capacity caught up with demand, and prices entered a sustained, multi-stage collapse through 2023 and 2024, eventually falling by well over 80% from the peak.

Here's the critical detail for traders: lithium carbonate futures only listed on the Guangzhou Futures Exchange (GFEX) in mid-2023 — after the peak, while the downtrend was already underway. So unlike traders in the 2021 thermal coal rally or the 2020 oil crash, futures traders never got to trade the euphoria phase. From day one, the listed contract lived inside a structural bear market punctuated by violent, short-lived relief rallies.

That's a brutal environment for mean reversion. And it's exactly why it's worth studying.

Know Your Instrument: Lithium Carbonate Contract Specs

Before we talk strategy, let's ground ourselves in the actual contract. Many global traders assume all Chinese commodity futures look like iron ore — big contracts, night sessions, deep liquidity. GFEX broke that mold.

Three things jump out. The 1-ton multiplier means even small retail accounts can express a view, but transaction costs are proportionally heavy, so scalping-grade mean reversion gets expensive fast. The no-night-session structure means every gap open is a fresh risk event — your stop-loss on a position held overnight is really a suggestion, not a guarantee. And the exchange's willingness to adjust limits and margins during volatility spikes means your backtest assumptions about slippage need a wide safety margin.

Why Mean Reversion Looked So Tempting Here

Let's be fair to the strategy. In its early listing months, lithium carbonate futures were a mean reversion trader's dream on paper:

• Extreme two-way volatility. The contract swung in double-digit percentage moves within weeks — far more than mature contracts like rebar or copper on the SHFE.
• Sharp mean-reverting intraday and multi-day behavior. After each flush lower, a bounce would arrive within days, rewarding anyone who bought exhaustion candles.
• Heavy speculative participation. A new contract with a hot narrative attracts fast money, and fast money overextends — which is precisely the fuel mean reversion feeds on.

A simple textbook setup might look like this:

• Calculate a 2-period RSI on the daily continuous contract
• Enter long when RSI(2) drops below 10
• Exit when RSI(2) closes above 50, or on a fixed 5–8 day time stop
• Optional: skip entries if price is more than 2 ATRs below the 20-day moving average (a "stretch" filter)

Run that on the first few months of GFEX data and it looks respectable. The bounces were real. The problem is that the sample was tiny, and the underlying regime — a market grinding down roughly 80%+ from spot highs over the broader 2023–2024 period — was quietly loading a trap.

Where It Broke: Oversold Is Not a Floor

Here's the uncomfortable truth the crash exposed: in a structural bear market, "oversold" just means "cheap relative to a price level that's about to stop mattering."

Walk through the mechanics. Lithium carbonate's decline wasn't a panic liquidation like the 2020 oil crash — it was a slow, persistent repricing as supply overwhelmed demand. Each leg down followed a pattern:

• Price grinds lower for two to four weeks
• Mean reversion signals fire — RSI(2) below 10, price tagging the lower Bollinger Band, consecutive down closes
• A bounce arrives... lasting two or three days, retracing maybe a third of the prior leg
• Then the downtrend resumes, and the bounce entry sits underwater for weeks

Notice the failure mode. The strategy didn't blow up in one catastrophic day. It bled. Small wins on the bounces, larger losses when the bounce failed, repeated dozens of times. With a 1-ton contract and meaningful tick costs, the math gets worse with every round trip. This is the classic death-by-a-thousand-cuts that makes trend-following regimes so hostile to fading strategies.

Mean reversion doesn't fail because the signals are wrong. It fails because the strategy's edge — the assumption that price returns to a stable center — simply doesn't exist when the center itself is falling.

Compare this to a market like rebar, where multi-month ranges are common and the "center" genuinely holds. That's why so many traders who cut their teeth on Chinese commodity futures like rebar and iron ore got humbled the moment they applied the same logic to lithium carbonate. The strategy didn't change. The regime did.

The Filters That Would Have Saved You

So how do you keep a mean reversion edge while refusing to fight a freight train? Three practical layers, in order of importance.

1. A hard trend filter — non-negotiable

The single highest-impact change: only take long mean reversion entries when the higher-timeframe trend isn't against you. Concretely:

• Compute the 50-day and 200-day moving averages on the daily continuous contract
• If the 50-day is below the 200-day and both are sloping down, long entries are disabled
• Alternatively, use a simple regime proxy: if price has made a lower low over the past 20 sessions while the 20-day MA slope is negative beyond a threshold, stand down

Applied honestly, this filter would have blocked most long fade entries during lithium carbonate's main descent legs. Yes, you'd miss some bounces. You'd also skip the majority of the bleed. Over a multi-quarter downtrend, that trade is overwhelmingly favorable.

2. Asymmetric engagement: fade with the regime, not against it

Mean reversion isn't only "buy dips." In a persistent downtrend, the higher-expectancy version is fading the bounces — shorting overbought conditions back toward the mean. A 2-period RSI above 90, or a stretch of three-plus up-closes into a falling 20-day MA, gave repeated short entries during the crash that aligned with the dominant flow. Same statistical family, opposite direction, radically different survival rate.

3. Size for gap risk, not just stop distance

Because GFEX has no night session, overnight risk is structural. If your stop is 1.5% away but the contract can gap 3–4% on an open after a policy headline or a big move in related battery-materials markets, your true risk per trade is the gap, not the stop. Practical fix: cap position size so that a full limit-move against you costs no more than your normal stop-loss loss. That usually means running smaller than your backtest suggests — and accepting it.

Taking This Beyond Lithium: A Checklist for Chinese Commodity Futures

The lithium carbonate crash is a case study, but the lessons generalize across China futures. Before deploying any mean reversion system on Chinese commodity futures — whether it's rebar on the SHFE, iron ore on the DCE, or a newer GFEX product — run it through this checklist:

• Regime test: Split your backtest into trending and range-bound periods. If the strategy only survives the range, you don't have a strategy — you have a regime bet.
• Cost stress test: Newer Chinese contracts often carry wider effective spreads relative to tick value. Double your assumed slippage and see if the edge survives.
• Session awareness: No night session? Model overnight gaps explicitly. Deep night-session markets like iron ore behave very differently.
• Exchange notice awareness: Chinese exchanges actively adjust margins, limits, and fees during volatility events. Build margin headroom into your sizing so an exchange action doesn't force a liquidation at the worst moment.
• Correlation check: Lithium carbonate didn't crash in a vacuum — broader sentiment across Chinese commodity futures soured at times. Know whether your mean reversion book is secretly one big short-China-demand bet.

None of this is exotic. It's the unglamorous work that separates traders who survive regime changes from traders who become cautionary tales.

Final Thoughts: Test It Before the Next Lithium Happens

The lithium carbonate crash will not be the last structural repricing in Chinese commodity futures. China's push into new energy, supply-side policy shifts, and the sheer speed of capacity build-outs guarantee more boom-bust cycles in metals, chemicals, and agricultural products. Mean reversion will keep working — in the right regime. The traders who last are the ones who know, in real time, which regime they're in.

Reading about it is one thing. Watching your own equity curve bleed through a simulated crash is another entirely. If you want to pressure-test your system against real Chinese futures data — lithium carbonate included — you can run it through a real-data China futures evaluation on XS Select, with challenges starting from $29. No promises about outcomes, just an honest way to find out whether your mean reversion rules can survive a market that refuses to mean-revert.

Trade the regime, not the signal. See you in the next case study.