Taking advantage of the Korea–US price spread in Hynix to give Binance AI Pro a try!
People have been trading the price spread between Hynix shares in Korea and the US, along with funding-rate arbitrage, for over half a year. Over the past three or four days, the spread has fluctuated less and less, and quite a few people in the group are stuck in positions and starting to complain loudly. My take is that when everyone else is complaining, it’s often a good time to open a position—and my cost basis is a little lower than theirs—so I wanted to try one and see how it went. Around the same time, people on X were discussing Binance AI Pro, and I wanted to see whether it could put a complete strategy into action in one go.
Hedging itself has a few hurdles: how to roughly balance the positions on both sides, how far the spread needs to deviate before entering, how to pause or reduce the position when funding rates get too high, and how to ensure both sides are closed together when losses hit a threshold, so no unhedged position is left behind.
To avoid wasting my Binance AI Tokens (points), I first worked with Grok to clarify the strategy logic, risk parameters, funding-cost controls, and rules for closing both sides at the same time. Then I fed everything to Binance AI in one go. With just one conversation, it generated a visual workflow and launched it in the demo account; after that, I switched to live trading.
The order was placed successfully: a conditional order was placed on one side first, and the other side opened at the same time. The stop-loss was automatically set to close both sides in full, leaving no unhedged position. But after opening, the premium widened a bit more, and I’m currently at an unrealized loss (including fees). The biggest drawback is the high funding cost on the long Korean-stock contract side, estimated at around 60 USDT per day for a position of roughly 20,000 USDT. Binance currently doesn’t let you buy the underlying shares, so the hedge has to be between contracts. Hopefully, they’ll eventually let users buy the actual shares directly. I’ll post updates on the stop-loss and closing status later.
Key points for the prompt, if you’d like to test it:
1/ Account risk controls: Total margin of 10,000 USDT, 2x leverage, notional value capped at around 20,000 USDT, isolated margin; maximum loss on either side around 1,000 USDT; if a stop-loss is triggered, both legs must be fully closed, leaving no naked exposure.
2/ Strategy logic: Go long SKHYNIXUSDT (tracks the Korean shares, relatively cheaper) and short SKHYUSDT (tracks the ADR, relatively more expensive), matching by equal notional value rather than mechanically matching the number of shares at a 10:1 ratio.
3/ Entry signal: Price-ratio z-score, normalized on a rolling basis using 100 one-hour candles. Enter only when the spread deviates clearly from its norm, to avoid paying funding costs for nothing when the spread has already converged.
4/ Funding-cost gate: Set a threshold based on the actual daily USDT cost of both legs, accounting for their different settlement intervals. If costs are too high, pause or reduce the position.
5/ Stop-loss execution: If either leg or the combined position reaches the loss threshold, use reduce-only trigger orders to close both sides in full at the same time.
6/ Other: Generate the visual workflow and code first, run a simulation before going live, and explain the opening quantity, notional value, margin usage, conditions for closing both sides, and main risks. Don’t open a position if there isn’t enough data.
This is for informational purposes only and does not constitute investment advice.
BitHappy, evolving all the time
People have been trading the price spread between Hynix shares in Korea and the US, along with funding-rate arbitrage, for over half a year. Over the past three or four days, the spread has fluctuated less and less, and quite a few people in the group are stuck in positions and starting to complain loudly. My take is that when everyone else is complaining, it’s often a good time to open a position—and my cost basis is a little lower than theirs—so I wanted to try one and see how it went. Around the same time, people on X were discussing Binance AI Pro, and I wanted to see whether it could put a complete strategy into action in one go.
Hedging itself has a few hurdles: how to roughly balance the positions on both sides, how far the spread needs to deviate before entering, how to pause or reduce the position when funding rates get too high, and how to ensure both sides are closed together when losses hit a threshold, so no unhedged position is left behind.
To avoid wasting my Binance AI Tokens (points), I first worked with Grok to clarify the strategy logic, risk parameters, funding-cost controls, and rules for closing both sides at the same time. Then I fed everything to Binance AI in one go. With just one conversation, it generated a visual workflow and launched it in the demo account; after that, I switched to live trading.
The order was placed successfully: a conditional order was placed on one side first, and the other side opened at the same time. The stop-loss was automatically set to close both sides in full, leaving no unhedged position. But after opening, the premium widened a bit more, and I’m currently at an unrealized loss (including fees). The biggest drawback is the high funding cost on the long Korean-stock contract side, estimated at around 60 USDT per day for a position of roughly 20,000 USDT. Binance currently doesn’t let you buy the underlying shares, so the hedge has to be between contracts. Hopefully, they’ll eventually let users buy the actual shares directly. I’ll post updates on the stop-loss and closing status later.
Key points for the prompt, if you’d like to test it:
1/ Account risk controls: Total margin of 10,000 USDT, 2x leverage, notional value capped at around 20,000 USDT, isolated margin; maximum loss on either side around 1,000 USDT; if a stop-loss is triggered, both legs must be fully closed, leaving no naked exposure.
2/ Strategy logic: Go long SKHYNIXUSDT (tracks the Korean shares, relatively cheaper) and short SKHYUSDT (tracks the ADR, relatively more expensive), matching by equal notional value rather than mechanically matching the number of shares at a 10:1 ratio.
3/ Entry signal: Price-ratio z-score, normalized on a rolling basis using 100 one-hour candles. Enter only when the spread deviates clearly from its norm, to avoid paying funding costs for nothing when the spread has already converged.
4/ Funding-cost gate: Set a threshold based on the actual daily USDT cost of both legs, accounting for their different settlement intervals. If costs are too high, pause or reduce the position.
5/ Stop-loss execution: If either leg or the combined position reaches the loss threshold, use reduce-only trigger orders to close both sides in full at the same time.
6/ Other: Generate the visual workflow and code first, run a simulation before going live, and explain the opening quantity, notional value, margin usage, conditions for closing both sides, and main risks. Don’t open a position if there isn’t enough data.
This is for informational purposes only and does not constitute investment advice.
BitHappy, evolving all the time