A 1% positive overnight return in Bitcoin or Ethereum predicts a 0.4% drop in the VIX during the next U.S. trading session. One signal, eight years of data, a Sharpe of 2.36 net of costs
Interesting piece. I tried replicating the strategy and could only reproduce strong results when using same-day Yahoo daily BTC/ETH returns, which appears to introduce lookahead bias for a 9:30am VIXY trade, since the crypto daily bar is not complete until later in the day. When I shifted the Yahoo daily crypto returns so they were actually known before the U.S. open, the edge largely disappeared. I also tested clean 5-minute spot BTC/ETH data for the later period using both the true 16:00-09:30 overnight window and the 16:00-20:00 “Yahoo UTC close” proxy, and neither reproduced the reported performance. Curious how you handled the Yahoo daily bar timing in the 2018–2024 proxy period.
Interesting piece. I tried replicating the strategy and could only reproduce strong results when using same-day Yahoo daily BTC/ETH returns, which appears to introduce lookahead bias for a 9:30am VIXY trade, since the crypto daily bar is not complete until later in the day. When I shifted the Yahoo daily crypto returns so they were actually known before the U.S. open, the edge largely disappeared. I also tested clean 5-minute spot BTC/ETH data for the later period using both the true 16:00-09:30 overnight window and the 16:00-20:00 “Yahoo UTC close” proxy, and neither reproduced the reported performance. Curious how you handled the Yahoo daily bar timing in the 2018–2024 proxy period.
Great point. Let me run some tests!