← Back to Blog · 2026-10-08 · 7 min read · Strategy Case Study

Picture this: it's late 2022, and lithium carbonate — the white powder powering every EV battery on the planet — is trading around 600,000 RMB per ton. Analysts are calling for a million. Then, within months, the price collapses by more than half. If you had a short position anywhere in that move, you made a year's P&L in a quarter. If you were a mean-reversion trader trying to "buy the dip" every few days, you probably got run over. Repeatedly.

This is one of the cleanest natural experiments in recent Chinese commodity futures history: a market that flipped from euphoric uptrend to relentless downtrend in under a year. It's a perfect stress test for the oldest question in systematic trading — does your edge survive a regime change?

One honest caveat before we start: lithium carbonate futures only listed on the Guangzhou Futures Exchange (GFEX) in mid-2023, so the futures market itself only captured the tail end of the crash. But the spot market collapse defined the futures trajectory from day one, and the listed contract spent its first year in a grinding, bouncy decline that tells us everything we need to know. Let's break it down.

What Actually Happened: A Two-Act Collapse

Act one was the spot market. Lithium carbonate roughly tripled through 2021 as EV demand exploded and Chinese battery makers stockpiled aggressively. It kept climbing through 2022 and peaked around 600,000 RMB per ton in late 2022. Then the story inverted: Chinese EV subsidies were being phased out, battery makers were sitting on bloated inventories, and new supply capacity was coming online. By spring 2023, spot had crashed below 200,000 RMB per ton — a decline of well over 60% in a matter of months. There was a sharp rebound mid-year, and then the grind lower resumed, with prices drifting toward 100,000 RMB per ton and below as 2023 turned into 2024.

Act two was the futures market. GFEX listed lithium carbonate futures in July 2023, listing near the 200,000+ level. The contract immediately inherited a market in structural oversupply. The first months saw violent two-way volatility — an early sell-off, a meaningful counter-rally into the autumn, then a persistent downtrend that dominated the rest of the year and beyond.

That shape matters. This wasn't a clean one-way slide like the 2020 oil crash. It was a downtrend punctuated by vicious relief rallies — the exact environment where naive systems of both types get hurt.

Know Your Instrument: GFEX Lithium Carbonate Contract Specs

If you're used to trading rebar or iron ore on the Shanghai and Dalian exchanges, GFEX's flagship contract has some quirks worth internalizing before you backtest anything:

The 1-ton multiplier is the sleeper detail here. It makes the contract accessible to small retail accounts, but it also means transaction costs and slippage are proportionally larger than in heavyweight contracts. Any system you run on lithium needs to respect that friction — especially mean-reversion systems, which trade frequently by design.

Also note: GFEX launched without some of the night-session depth that Dalian and Shanghai offer, so your session-based logic needs adjusting versus, say, an iron ore night-trading setup.

Setup A: The Trend-Following System

Let's define a deliberately simple trend system so we're comparing apples to apples:

• Entry: Go short when the 20-day moving average crosses below the 60-day moving average. Flip long on the opposite cross.
• Stop: Initial stop at 3× the 14-day ATR from entry.
• Exit: Trail the stop at 3× ATR, or exit on an opposite MA cross.
• Sizing: Risk a fixed 1% of equity per trade.

How does this fare through the lithium collapse?

The good news: the big move pays. Once the 20-day MA rolled below the 60-day in the aftermath of the late-2022 spot peak (and again decisively after the mid-2023 futures listing), the system catches a downtrend that extends for months. With 3× ATR trailing stops, you sit through bounces without getting shaken out, and each leg lower adds to the gain. One or two well-held short trades can account for the vast majority of the system's annual P&L — classic trend-following distribution.

The bad news: the whipsaw tax is brutal. Lithium's downtrend was not smooth. The mid-2023 counter-rally was strong enough to flip a moving-average system long, then reverse it short again at a loss. In a market with 4% daily limits and ATR readings that expand violently, a 3× ATR stop can mean losing 10%+ of the position's notional in a single adverse session. Expect a string of small-to-medium losses punctuated by one or two large wins. If your backtest shows a 35–45% win rate with a fat right tail, that's not a bug — that's the strategy.

The trend system's job on lithium wasn't to be right often. It was to survive the bounces long enough to be present for the collapse.

Setup B: The Mean-Reversion System

Now the other side of the ring:

• Entry: Buy when price closes below the lower 2-standard-deviation Bollinger Band on a 20-day lookback AND the 2-period RSI drops under 10. Sell short on the mirror-image overbought condition.
• Exit: Close when price reverts to the 20-day middle band, or on a 2× ATR disaster stop.
• Sizing: Same 1% risk per trade.

On paper, this looks great on lithium. A crashing market produces constant oversold readings, and the mid-2023 rebound handed the long side a genuine gift — buying the first capitulation low near the futures' early lows caught a double-digit percentage bounce.

But here's the problem: in a persistent downtrend, mean reversion is structurally fighting the tape. Every "oversold" buy in the second half of 2023 was a knife-catch in a market where each relief rally failed at a lower high. The system wins small when it wins (exit at the middle band), and loses big when the dip keeps dipping. With 1-ton lots and frequent entries, commissions and slippage compound the bleed. The disaster stop saves you from ruin on any single trade, but the equity curve takes the shape of a slow staircase down, interrupted by the occasional bouncy plateau.

There's a deeper lesson: mean-reversion systems implicitly assume the price distribution is stationary. Lithium in 2022–2023 was anything but — a structural repricing driven by subsidy removal, inventory gluts, and new mine supply. No RSI reading tells you that the "fair value" you're reverting toward has itself collapsed by half.

Head-to-Head: What the Crash Actually Teaches

Put the two equity curves side by side and the pattern is textbook:

• Trend following: choppy, losing stretches during the bounces; one or two outsized winners; positive skew. The system's survival depends entirely on its ability to hold through adverse moves without widening stops or "taking profits early."
• Mean reversion: steady small wins during range-bound stretches, then a cluster of losses during trend legs. On lithium, the mid-2023 rebound rescued the P&L temporarily; the resumption of the downtrend erased it.
• Drawdown character: the trend system's drawdowns are frequent but shallow relative to its winner; the mean-reversion system's drawdowns are slow, grinding, and psychologically harder to sit through because there's no big winner coming to bail you out.

Neither system is "better." Lithium in this period was simply a trend regime, and trend systems get paid in trend regimes. Run the same mean-reversion system on a genuinely range-bound Chinese commodity — think agricultural products in quiet years, or rebar in a consolidation phase — and the ranking flips. The edge isn't in the indicator. It's in matching the system to the market's regime.

Practical Takeaways for Your Own System

Before you touch lithium carbonate futures or any other Chinese commodity futures contract, steal these five rules from the 2022–2023 case:

• Add a regime filter. Something as simple as "only take mean-reversion longs when the 60-day MA slope is flat or rising" would have kept the dip-buyer out of most of lithium's damage.
• Respect limit-up/limit-down mechanics. A ±4% daily limit that can expand means your backtest must simulate days when you simply cannot exit. Trend systems with wide ATR stops handle this; tight-stop systems don't.
• Watch the calendar spreads. In a market in structural oversupply, deferred contracts often behave differently from the front month. Backtesting a single continuous series without checking roll behavior can distort results.
• Size for the volatility, not the notional. The 1-ton contract invites oversizing. Fixed-fractional risk (1% or less per trade) is what lets a trend system survive long enough to catch its one big winner.
• Backtest across regimes, not just the exciting ones. A system that only looks good on 2022–2023 lithium is a system fitted to one story. Test it on the 2021 thermal coal rally and intervention, on the 2020 oil crash, on boring rebar years. Consistency across regimes is the real signal.

Test It on Real Data Before You Risk Real Money

The gap between "this would have worked" and "this works" is where most retail trading careers end. The only way to close it is to run your rules against real, exchange-grade Chinese futures data — including the limit days, the rolls, the session quirks — and see the honest equity curve, drawdowns and all.

That's exactly why we built XS Select. As a new platform purpose-built for evaluating futures traders on real China market data, we let you put your trend, mean-reversion, or hybrid system through structured evaluations starting from $29 — no inflated promises, just your rules versus real markets. If lithium's crash taught us anything, it's that the market doesn't care about your backtest's beauty. Prove your system before the next regime change proves it for you.