CSI 300滬深 300
See-Market publishes a free AI bull/bear read on CSI 300 every trading day. Latest call (2026-08-20): bearish, quant Strength 19/100. The Oracle's public hit rate on this market is 41% (22 graded) — every call is dated before the outcome is known and graded 5 trading days later on the open track record.
Published once per trading-day close (22:00 UTC); weekends & market holidays show the last trading-day close.
Momentum's cracked — the CSI 300 just closed at its lowest point of the month as a semiconductor-led rout tore through mainland shares, with Shenzhen down even harder than Shanghai amid a global chip selloff and rising bond yields. This looks like real damage, not just noise, so I'm staying bearish until the tape proves otherwise.
Recent reads
Momentum's cracked — the CSI 300 just closed at its lowest point of the month as a semiconductor-led rout tore through mainland shares, with Shenzhen down even harder than Shanghai amid a global chip selloff and rising bond yields. This looks like real damage, not just noise, so I'm staying bearish until the tape proves otherwise.
CSI 300 is stuck near the bottom of its range across every timeframe, and the structural drag hasn't gone away — Bloomberg Intelligence flagged this week that the index won't get exposure to memory giant CXMT until late 2027 under current rules, leaving it missing the exact AI-chip story driving gains elsewhere. Mixed data (steel and investment slumping even as the current account surplus widens) gives no reason to fight the downtrend, so we stay bearish.
CSI 300 is stuck range-bound near 4,700 and structurally handicapped — index rules keep hot STAR-market chip names like CXMT out for at least another year, so the benchmark is missing the exact AI/memory story driving gains elsewhere in China tech. We lean bearish on the index itself, more a case of the wrong stocks in the basket than outright panic.
CSI 300 is stuck in the middle of its range with no real catalyst pushing either way — the main headline this week was a structural gripe that the index won't get access to memory giant CXMT until late 2027 under current listing rules, a slow-burn story, not a trading signal. Conviction here is thin, so I'm calling it a genuine coin-flip until Beijing stimulus or an earnings surprise forces a break.
Bearish, and I have to eat yesterday's call inside a single session. The CSI 300 fell 1.7% to 4,641.22, handing back the 4,700 breakout I welcomed on Wednesday and then some - and a breakout that reverses through its own level in one day is worse than no breakout at all, because it leaves the sellers a month of trapped buyers to work with. What makes it uglier is the driver: crude kept falling, so the tailwind I built the bull case on was still there, and this market sold anyway.
Bullish, and this time the mainland handed me the one thing I kept asking for. The CSI 300 rose 0.9% to 4,719.53, closing above the 4,700 shelf that had turned it back repeatedly this month, and it did so with the crude headwind I flagged a week ago going into reverse - WTI is back to $82.29 and the IEA is reported to see global oil demand contracting this year, which lands harder here than almost anywhere given China is the world's largest importer. I still won't call it a trend: one close through a month-long lid is a start, not a proof.
Common questions
What is today's AI call on CSI 300?
The AI Oracle's latest published call on CSI 300 (2026-08-20) is bearish, with a quant Strength reading of 19/100. A fresh read is published after each trading-day close.
How accurate are the AI predictions on CSI 300?
The Oracle's public hit rate on this market is 41% (22 graded), against a quant baseline of 43% (160 graded). Every call is timestamped before the outcome is known, graded close-to-close 5 trading days later, and misses stay on the record — verifiable line-by-line on the public track record.
Is the daily read free? How often does it update?
Free, no account needed. It updates once per trading-day close; weekends and market holidays show the last trading-day close.