The Fund Hit a Record High. The Money Left Anyway.

The net value recovered. The money left anyway.

At the end of January 2026, the flagship product at Shanghai-based Banxia Investment reached a cumulative net value of 8.618 — the highest in the fund’s history, according to third-party fund data platform Simuwang.

It had been climbing for seven months. The rebound started in June 2025 and ran all the way to that high.

That is normally when money stays. What happened instead is that money kept leaving, and the record high did not stop it.

Banxia had already dropped off China’s list of ten-billion-yuan private funds in early 2025, falling into the 5–10 billion yuan band. By April 7, 2026, filings with the Asset Management Association of China showed it in the 2–5 billion yuan band. (AMAC discloses assets under management in brackets rather than exact figures.)

At its peak the firm ran more than ten billion yuan.

Investors were not scared off by falling net value. They left as it climbed back.

What they were waiting for was not her

Chinese financial media covering the redemptions in May 2026 kept returning to one explanation: these investors were not waiting for a comeback. They were waiting to break even. Once the net value recovered to around their cost, or a little above it, they redeemed.

The same reporting offered two more. One is that China now has far more large private funds than it did three years ago, so money that gets back to par has somewhere better to go. The other is that when a founder’s personal brand is fused to the firm’s, the spotlight amplifies attention on the way up and amplifies redemptions on the way down.

The third explanation is the one worth sitting with. What it says, underneath, is that some of these people were not buying a fund.

The money came in at the top

To see how high the top was, look at the filing dates.

One analysis of Wind filing data found that of the 64 products Banxia has registered, 49 were filed between 2021 and April 2023 — roughly three quarters of them, inside a window of about two years.

The peak in net value came on April 21, 2023, at 7.764.

So the money arrived pressed up against the high. As one industry figure put it in Chinese media, a good number of Banxia’s clients subscribed at elevated net values and ran straight into the drawdown.

That drawdown lasted two years. By May 30, 2025, cumulative net value was 5.1 — down 34.31% from the April 2023 peak. Redemptions were processed normally throughout; this was a loss on paper, not a lock-up.

What happened next is not common. On May 4, 2023, Banxia announced that from May 8 it would stop accepting subscriptions from new clients. Existing investors could still add money, redemptions continued as usual, and the notice said the firm would reopen fundraising when market conditions allowed.

Li Bei’s stated reason: if assets grew much further, several instruments the strategy relied on would run into exchange position limits, which would affect how the strategy could be implemented and its long-run returns. She said she would rather stay in the second tier by size and keep first-tier performance than slide into second-tier performance.

In an industry paid on assets under management, turning new money away is rare. That belongs in her column.

She was not lucky

This matters, because the easy version of this story is that she caught three good years. She did not.

Take 2018 — a bear market in commodities, a bear market in stocks, a mild bull market in bonds, a year that suited very few playbooks. She held no equities, put about 70% into rate bonds as a base and used the remaining 30% in options and futures to short commodities, and finished the year up close to 20%.

In 2020, the Banxia Steady Mixed Macro Hedge product returned 258.24%, the best among domestic macro hedge funds that year. In 2021, the same product returned 94.97%, according to Simuwang data cited in Chinese financial media.

That is the firm’s highest-volatility product; the average across Banxia’s products in 2020 was 99.99%. Every net value and loss figure in this piece refers to that same product, good years and bad.

In 2022 the CSI 300 fell 21.63% and she finished slightly positive.

And even counting the two bad years since: as of May 22, 2026, that product’s cumulative net value was 6.919 — an annualised return of 27.38% since inception.

None of this is padded. Macro hedge strategies make money from rotation across asset classes, and she has repeatedly shown she can find that path in very different environments.

In mid-2022 Banxia crossed ten billion yuan and Li Bei became the first woman in China to run a private fund that size. The nickname that has followed her for years — simu monü, roughly “the private-fund sorceress” — belongs to that period too. She is among the most fluent users of social media in the industry; her WeChat posts routinely clear 100,000 reads, and she writes about her own life as well as her market views.

So a person got turned into an oracle.

Here is where the causation got reversed. What she had demonstrated was that a framework can make money across many kinds of years. What people heard was that this person does not get things wrong.

The first sentence is about a method. The second is about a human being. They look almost identical, and the distance between them is the entire risk.

Both letters had her name in the title

After 2023 the calls stopped landing, and investors and finance bloggers started using the word fanzhi — “reverse indicator” — about her. Whatever she likes, you avoid. By May 2025 there was a widely shared piece whose title was simply “Why Did Li Bei Become a Reverse Indicator?”

Worth noting: that piece concluded the label was unfair to her.

But “sorceress” and “reverse indicator” came from the same crowd.

The numbers were genuinely poor. As of the end of 2025, two representative products showed three-year returns of 3.28% and −1.47%, against 19.33% for the CSI 300 over the same period.

In June 2025 she published a long self-assessment. Its title: “What Did I Miss? What Did I Get Wrong?”

On December 18, 2025, she published an open letter. The title was “Why You Should Allocate to Banxia Now — Bet on Li Bei.”

Eight days later, on December 26, she launched a paid in-person course, “Learning to Invest From Zero,” priced at 3,888 yuan a session and 12,888 yuan for four, capped at 200 people.

It went around Chinese finance circles fast. The next day she responded: she did not need the money, and the most knowledge-monetisation brings in is a few million a year. She said she would take no fee and play no part in handling the funds, and that she would put up seed money for a charitable fund which, together with part of the course proceeds contributed by the platform, would go toward education support.

In February 2026 came the second open letter. The title was shorter: “Allocate to Banxia — Bet on Li Bei.”

Not “allocate to a macro hedge strategy.” Not “allocate to Banxia’s method.” Bet on Li Bei.

At that point the loop closes. The market crowned her, and she accepted the crown and printed it in a headline.

As of June 18

On May 8, 2026, that product’s net value was 8.11.

By the end of May, 7.

On June 18, 5.6.

In the single week from June 15 to June 18, it fell 15.03%.

As of June 18, the product was down 24.5% for the year, past the 20.49% it lost across all of 2023. The CSI 300 was up 6.73% over the same stretch.

Her view has not changed. She still likes Chinese property, saying a once-in-a-decade turning point is essentially confirmed. She still thinks AI is a bubble that will break. After the drawdown she told investors that, on risk-control grounds, she had trimmed positions over the previous week and cleared holdings she was not highly confident in.

She may turn out to be right. Neither the property call nor the AI call is settled yet.

But that is a different question.

Three years ago she was the one who shut the door to new clients, on the grounds that too much money would drag on returns. Now, inside a single year, she has written two letters asking people to come back, with her own name in both titles.

She has also said that if investors want to redeem and wait on the sidelines, she respects that.

The assets keep falling.

Whether any one person will keep being right is not knowable in advance. Only one thing is knowable in advance: whether your money went in for a method, or for a name.

While net value is rising, those two look the same. The difference shows up on the day it falls.