On the morning of December 8, 2005, a young company called J-Com listed on the Tokyo Stock Exchange. It was a small recruitment firm with exactly 14,500 shares outstanding. Within minutes of the open, a trader at Mizuho Securities set out to sell one share at ¥610,000.
He typed it the other way round: an order to sell 610,000 shares at ¥1 each.
Six hundred and ten thousand shares — of a company that had fourteen and a half thousand in existence. An offer to sell 42 times the entire company, for one yen per share.
Every safeguard fails at once
The trading system flashed a warning. The trader dismissed it — warnings pop up all the time; clicking through them was routine. The order hit the book, crushed the price limit-down, and started filling against a wall of delighted buyers.
Mizuho realised within minutes and tried to cancel. And here the story turns from embarrassing to historic: the Tokyo Stock Exchange's system refused the cancellation. A software flaw kept the order alive while Mizuho frantically bought back shares that, in aggregate, did not exist. The firm was legally short hundreds of thousands of phantom shares and had to settle in cash. The loss Mizuho booked: about ¥40.7 billion — roughly $340 million — for one transposed number. Years later the courts found the exchange 70% to blame for the failed cancellation and ordered it to pay Mizuho about ¥10.7 billion, so the final net figure was nearer ¥30 billion. Heads rolled far beyond the trading desk; the TSE's president resigned over the failed cancellation. Several trading firms that had gleefully bought the ¥1 shares later returned their profits under public pressure.
"That could never be me" — except it is, weekly, in miniature
It is tempting to file this under exotic institutional failure. But look at the anatomy: a routine action performed hundreds of times, done once in reverse under time pressure; a warning dismissed because warnings are always dismissed; and a small error becoming a catastrophe because the system allowed the position to exist at all.
Retail traders re-enact this constantly at smaller scale: the sell that was meant to be a buy; the 10-lot that was meant to be a 1-lot; the market order sent to the wrong contract month at 2 a.m. The lesson is not "be more careful" — the Mizuho trader was a professional having a normal busy morning. Human error rates never go to zero. The lesson is that process, not attention, is the safeguard: order previews read aloud, size limits hard-coded at the broker, one deliberate breath between building an order and sending it. This is the same philosophy as the printed pre-trade checklist from Lesson 1 of the course — the point of external process is that it works on exactly the days your attention doesn't.
What this story teaches
- Attention fails; process holds. Any protection that depends on you being sharp will eventually meet a morning when you are not.
- Never dismiss a warning you haven't read. Alert fatigue was the real killer here — the system did try to stop him.
- Hard-code your limits. A broker-side maximum order size is free insurance against your own fingers.
- Speed is optional; accuracy is not. Almost no retail trade is so urgent that three extra seconds of verification costs you the opportunity.