Every selloff in the last 7 years followed the same pattern. I tracked which funds buy fear - and whether it works.

COVID crash. 2022 rate hikes. Q2 2025 corrections. Every time, the same thing happens: retail panics, institutions split. Some sell. Some double down.

I track 500 largest US institutional 13F filers by AUM - not just what they hold now, but every position they've opened and closed since 2018. Win or loss, entry to exit.

Here's how 20 well-known funds actually perform across all their positions:

|Fund|Win Rate|Avg P&L|Positions|Avg Hold| |:-|:-|:-|:-|:-| |Berkshire Hathaway|65.1%|+31.9%|152|7.5 qtrs| |Soroban Capital|64.0%|+22.5%|186|3.4 qtrs| |Elliott Management|63.2%|+69.3%|155|3.6 qtrs| |D1 Capital Partners|56.9%|+11.9%|522|2.6 qtrs| |Citadel Advisors|54.6%|+4.3%|32,940|3.7 qtrs| |Soros Fund Mgmt|53.4%|+15.5%|1,741|2.2 qtrs| |Viking Global|53.3%|+9.9%|859|2.9 qtrs| |Lone Pine Capital|51.6%|+23.3%|314|3.3 qtrs| |Point72|51.0%|+5.2%|12,911|2.3 qtrs| |Third Point|50.0%|+11.7%|458|2.6 qtrs| |Bridgewater|49.5%|+3.0%|5,005|3.5 qtrs| |Two Sigma|49.4%|+4.7%|16,058|3.4 qtrs| |Baupost Group|48.3%|+28.1%|209|3.9 qtrs| |Renaissance Tech|48.1%|+6.4%|17,692|3.8 qtrs| |Millennium Mgmt|47.8%|+3.3%|18,144|4.2 qtrs| |SoftBank|46.4%|+27.8%|110|3.1 qtrs| |Coatue Management|45.2%|+13.2%|1,521|2.2 qtrs| |Dragoneer|40.2%|+6.1%|261|3.8 qtrs| |Tiger Global|39.8%|+11.1%|397|5.1 qtrs| |ARK Invest|34.3%|-4.3%|875|5.9 qtrs|

Win rate alone doesn't mean much. Look at Baupost: 48.3% win rate, +28.1% avg P&L. Klarman's whole thesis is buying at enough of a discount that even when he's wrong, the losses stay small. A 48% win rate with +28% average means the upside is absorbing the downside. SoftBank is similar - loses more than it wins, but avg P&L is +27.8%. Elliott at +69.3% avg P&L is the activist premium - they don't just buy stocks, they force changes.

ARK is the one fund where both columns are red. 34.3% win rate AND -4.3% avg P&L. The only negative average on this list. They hold losers for 5.9 quarters before cutting. Berkshire holds for 7.5 quarters - same patience, opposite results.

Now here's what some of these funds did during Q4 2025's selloff:

|Fund|Stock|Shares Added|Q4 Drop|Portfolio Weight| |:-|:-|:-|:-|:-| |D1 Capital|Sea Limited|+169%|-29%|4.2%| |Durable Capital|DoorDash|+114%|-16%|7.7%| |HHLR Advisors|Alibaba|+65%|-20%|25.6%| |Soroban Capital|Meta|+61%|-10%|6.0%| |Altimeter Capital|Coupang|+56%|-27%|5.6%| |Dragoneer|Coupang|+47%|-27%|11.1%|

These aren't token buys. HHLR (Hillhouse Capital) put 25.6% of a multi-billion dollar portfolio into a single Chinese stock during a 20% drop. Two separate funds - Altimeter and Dragoneer - independently loaded up on Coupang during the same 27% decline.

But here's the thing worth knowing: Dragoneer has a 40.2% all-time win rate. They buy fear often. They're wrong more often than they're right. Not every contrarian move is a smart one.

Every major selloff in the dataset - COVID crash, 2022 rate hikes - recovered within months. Q4 2025's dip buys are still playing out, and with Iran pushing oil past $100, those positions are being tested right now. But historically, the only way buying institutional-level dips doesn't work is if you're betting on the permanent decline of Western capital markets. And if that's your thesis, your portfolio is the least of your concerns.

Caveats: 13F data has a 45-day lag and quarterly granularity. Entry/exit prices are estimated from cost basis data, not exact transactions. Renaissance's 13F covers RIEF/RIDA, not Medallion. Soros Fund run by Dawn Fitzpatrick since 2017. ARK's 13F partly reflects fund flows. Citadel includes market-making. 2018-present was mostly bullish.

All data from public SEC 13F-HR filings on EDGAR. I built a tool that tracks every entry and exit across largest 500 US funds.