Where are the arbitrage opportunities for on-chain stock perpetual contracts?

Author: Zhou, ChainCatcher

Recently, a review of SK Hynix's cross-market arbitrage has been circulating on X. A trader recounted making over $600,000 from the cross-market price spread of Hynix since June. The positions and profits in the review cannot be verified from public data, but the market structure and mechanisms he described have no major flaws.

Meanwhile, Changxin Technology will be listed on the A-share Sci-Tech Innovation Board next Monday, and the Pre-IPO perpetual contracts on-chain have already been released. This is the largest IPO project in the capital market recently, following the Hynix ADR, and many investors are discussing whether Hynix's strategy can be applied to Changxin.

He Profited from Rule Differences

Today, SK Hynix has at least five price sets operating simultaneously—Korean common stock, Nasdaq ADR, Hong Kong leveraged ETF, stock perpetuals on centralized exchanges, and HIP-3 contracts on Hyperliquid. Their anchoring objects, settlement currencies, and trading hours differ, and GodpanSen took advantage of the price differences in between.

First, let's look at how he specifically operated. According to the post, he really focused on Hynix after the contract was launched on Binance in June. Before that, there was only one contract curve on Hyperliquid, and after Binance entered, there were two prices to compare.

First, he mentioned that one weekend he found that the Hynix contract on Binance was $30 more expensive than on Hyperliquid. He judged that the price difference came from the different rules for collecting funding fees, and as long as the funding fee paid before closing was less than $30, he would profit. He built a position of 1,000 shares, and on Monday, as the common stock opened, the price difference converged, netting him a profit of $15,000 after deducting the funding fee.

Next was the main position. Users in the crypto space generally find it difficult to buy Korean common stock; retail investors can only go long on the contract side. He claimed that at times, the single stock premium peaked at over $40. So he bought Korean spot stock using his IB account while shorting the high-premium contract, offsetting the directions without betting on the stock price. He waited for the price difference to revert while collecting the funding fees paid by the longs to the shorts, earning over $120,000 in this segment.

The third point was utilizing the different mechanisms for collecting funding fees across platforms. He stated that the prices on three platforms consistently showed Binance higher than OKX, and OKX higher than Hyperliquid. After backtesting, he found that except for periods of significant price movements before and after trading hours, OKX's weekly funding fee was nearly 1 percentage point higher than Binance's. Thus, when the prices were close, he moved his position to OKX, earning $170,000 in funding fees, estimating that the same position left on Binance would only yield around $100,000.

He attributed this to OKX not using its own algorithms and directly taking a proportional index from Hyperliquid and Binance. This point was not explained in OKX's official documentation, which only stated that the index was composed of multiple price sources, with some components smoothed using EMA.

The fourth point was betting that the rules would change, effectively exploiting a rule bug. During the days when Hynix's common stock plummeted, the Binance contract was over $40 higher than the common stock. According to normal calculations, the 8-hour funding fee rate should be above 1%, but Binance set a single limit that locked the rate at 0.5%, while Hyperliquid settled every hour, pulling the price difference to $30.

GodpanSen judged that this state wouldn't last long, shorting Binance and going long on Hyperliquid, with a position close to $10 million and an average price difference of about $25 per share. He claimed that by that afternoon, Binance changed the funding fee to be collected every 4 hours, causing the price difference to fall, and he closed the position in two transactions, bringing in over $150,000.

However, this segment seems inconsistent with Binance's public announcement. According to Binance's announcement on June 1, the funding fee rate limits for these three perpetual contracts were set at plus or minus 2% when they launched, and only narrowed to plus or minus 0.50% on July 15 at 00:15, while also changing the settlement frequency from 8 hours to 4 hours. He described a situation where the rate was locked at 0.5% before the rule change, which does not align with the announcement; what he encountered was more likely an actual cap rather than the announced limit.

Finally, regarding the leveraged ETF. He mentioned that on the Friday when the Korean stock market was closed but the Hong Kong stock market was open, the 2x long Hynix ETF listed in Hong Kong dropped over 20 points in a single day, equivalent to the common stock dropping over 10%, while the crypto contracts only fell by 5% during the same period. He bought the discounted ETF while shorting the Hynix contract on Binance, waiting for the Korean stock market to open on Monday.

He calculated the hedge ratio as 100 shares of ETF to 1.19 shares of common stock, but actually hedged at a 1 to 1 ratio, intentionally leaving about 20% of a naked long exposure. In the end, he bought 480,000 shares of ETF and shorted 5,000 shares of the contract. On Monday, the common stock only dropped 5% and then rose, allowing him to close the position and take away over $200,000.

Finally, he mentioned that when buying Korean spot stock through IB, he borrowed Korean won for convenience, while the underlying crypto contracts are pegged to the dollar value of the common stock, effectively exposing him to the exchange rate between the won and the dollar. He opened the position just as the won hit a recent low, and over the past month, the won appreciated significantly, resulting in a loss of $60,000 just from foreign exchange when closing the position.

The Mechanism is Perfect, but the Operation is Difficult

In summary, his operations mainly involved three techniques.

First, moving the price difference between two contracts. The same underlying asset has price differences across different exchanges, primarily due to the different rules for collecting funding fees. As long as the funding fees paid during the holding period are less than the price difference, one can profit. These positions are both contracts, and the trader is actually hedging time rather than assets.

Second, buying spot and shorting contracts, holding neutrally to collect funding fees. However, directionally neutral does not equate to risk-neutral. Funding fees are variable; being positive long-term is merely a result of retail investors going long. Once the short side gains dominance, the rate can turn negative, and the shorts become the ones paying instead of receiving.

Third, capturing mismatches during market closures. Hynix exists simultaneously in Korean common stock, Hong Kong leveraged ETFs, Nasdaq ADRs, and 24-hour trading crypto contracts, with their opening and closing times not overlapping. The premise for this type of operation is that prices indeed revert rather than continue moving in the same direction, and additionally, leveraged ETFs incur daily rebalancing losses, leading to greater tracking errors the longer they are held, making them suitable only for short windows like crossing a weekend.

This mechanism seems perfect, but in practice, it is extremely difficult to implement.

Executing this strategy requires not only access to various trading channels—Korean stock accounts, foreign exchange pathways, broker quotas, and margin accounts across multiple exchanges—but also a keen sensitivity to rule differences, including each platform's funding fee settlement cycles, rate limits, index compilation methods, and when these differences will be magnified.

Moreover, his last two points have already moved away from pure price difference capture; one bet on whether Binance would change the funding fee limit, and another bet on whether the Hong Kong ETF discount would revert on Monday, both requiring directional judgments about price movements and platform behaviors.

GodpanSen stated that arbitrage can only use 20-30% of the position size, and the capital scale that can support this set of actions is itself a filter.

Many users expressed that even if they followed along, it was difficult to capture the full profit. Furthermore, the actual risk differences between various arbitrage methods are significant, especially when incorporating rule changes and directional judgments, which already carry a strong game-theoretic nature.

The Same High Premium, What’s Different About Changxin?

However, the market is always profit-driven, and many investors are starting to turn their attention to the upcoming listing of Changxin Technology.

Currently, Hyperliquid has launched the Changxin perpetual contract CXMT-USDC, which, as of the time of writing, is reported at $6.3896, equivalent to about 43.26 RMB, down about 25% from its peak, but still around five times the issue price. Based on a total share capital of 66.881 billion shares after issuance, the implied market value is approximately 2.94 trillion RMB.

The premium is an order of magnitude higher than that of Hynix at the time, but the arbitrage space may not necessarily be the same.

Hynix exists simultaneously in Korean common stock, Nasdaq ADR, Hong Kong leveraged ETF, and contracts across multiple exchanges. Each of GodpanSen's three methods requires at least two curves: moving the price difference needs two contracts, neutral charging requires both spot and contracts, and capturing mismatches requires ETFs and contracts.

In contrast, before its listing, Changxin only has one curve on Hyperliquid, and multiple on-chain positions have already begun to establish short positions. These positions lack a spot leg to hedge against, betting instead on the price converging downward after listing, not arbitraging between two prices.

After listing, the 500,000 RMB asset threshold for the Sci-Tech Innovation Board combined with QFII quota restrictions will block the vast majority of overseas investors from accessing the common stock, creating a barrier similar to that faced by crypto users who cannot buy Korean common stock.

The real difference lies in the price fluctuation limits. According to the Shanghai Stock Exchange rules, new stocks on the Sci-Tech Innovation Board do not have price fluctuation limits for the first five trading days, and from the sixth trading day onward, the limit is set at ±20%. Once the common stock hits the price limit, the convergence mechanism will be directly cut off.

Another point is that in the arbitrage mechanism of going long on common stock and shorting contracts, Changxin involves RMB against USD and USDC, with the RMB not being freely convertible, and there are onshore-offshore price differences, as well as quota and remittance restrictions for QFII funds.

As for the contract price differences between exchanges, this layer can be replicated, provided that after listing, various exchanges gradually launch Changxin's perpetual stocks, resulting in the emergence of the second and third curves. The mismatch in trading hours is the same.

However, this strategy has now been made public and even widely circulated, which means that the information gap has become thinner, and the speed at which the price differences are compressed has also increased.

Are Market Opportunities Increasing?

With the diversification and fragmentation of the market, these seemingly easily accessible arbitrage opportunities appear to have increased, but in reality, they place a heavier burden on traders' understanding and execution capabilities.

Once the same asset is split into common stock, depositary receipts, leveraged ETFs, and contracts across multiple platforms, the differences in index algorithms, settlement cycles, rate limits, and trading hours will continuously create price differences.

The gap between seeing a price difference and capturing it involves cross-market accounts, channels, margin dispatch, and risk control execution; any missing link can distort the action.

It is worth considering that these trades seem to earn certain money, but the risks are almost all hidden outside of the price.

Arbitrage in the early stages largely involved betting on the rules themselves. When new markets are just emerging, the rules are often not yet fully refined, and price differences arise from the roughness of the mechanism rather than pricing errors. Discussions about Polymarket arbitrage over the past two years also belong to this category.

These opportunities have a clear half-life; as mechanisms are completed, market depth increases, and more participants join, the price differences that can be captured will become thinner. For professional traders, rule differences may signify arbitrage opportunities. For ordinary retail investors, the same rule differences may become sources of risk.

On the other hand, leverage is an invisible killer. Yesterday, economist Fu Peng from New Fire Technology also mentioned that capital market pricing reflects expectations and will significantly lead real fundamentals. Current realities like capacity shortages and full orders cannot deduce sustained stock price increases because stock prices trade on the future.

He noted that many young traders in the Korean market made substantial profits one day and faced significant losses the next, with the issue not lying in corporate operations or supply-demand in the industry, but rather in the excessive accumulation of leverage in the market.

This point also holds true for arbitrageurs.