The fat finger, and the order you cannot cancel

Graduate · enrolled learners

This lesson opens with The Mizuho J-Com error, 8 December 2005.

What happened

A trader at Mizuho Securities intended to sell one share of a company called J-Com at 610,000 yen, but transposed the price and quantity and entered an order to sell 610,000 shares at 1 yen each, an order for far more shares than existed and at an absurd price. Two failures then compounded. First, the order was accepted: nothing in the entry path rejected a sale of hundreds of thousands of shares that did not exist at a price of one yen, though the order was obviously nonsensical. Second, when Mizuho realized the mistake and tried repeatedly to cancel, the Tokyo Stock Exchange system rejected the cancellations, so the erroneous order could not be pulled back while it filled. Mizuho lost roughly 40 billion yen, about 225 million dollars. Nothing was hacked; a human made a fat-finger error, the systems accepted an order they should have questioned, and then the one action that could have contained it, a cancel, did not work.

The decision point

Order entry is where a human or an algorithm hands an instruction to the market, and it is the place to catch the obviously wrong order, through validation and limits, and to guarantee that a mistake can always be pulled back, through a cancel that works, because an order-entry path that accepts the impossible and then cannot cancel it is two failures that together are catastrophic. Mizuho is the case: a trader meant to sell one share at 610,000 yen and instead sold 610,000 shares at 1 yen, the systems accepted an order for more shares than existed at an absurd price, and then repeated cancellations were rejected, costing about 225 million dollars. The lesson is that order entry needs two defenses working together. The first is validation and limits at the point of entry, sanity checks that reject an order for more than exists, a price far off the market, a size beyond a set cap, so an obvious error never reaches the market. The second is a reliable cancel, because humans and algorithms will make mistakes that pass every check, and the ability to pull an order back is the last line of defense, which must actually function when it is needed. So the discipline is to build order entry that questions the impossible before it is sent and that always allows a correction after, because Mizuho showed that either defense alone is not enough: a system that only validates will still be beaten by an error that looks plausible, and a system that only allows cancels will still send the absurd order that a filling market may take before the cancel lands, so both the check going in and the escape hatch coming back must be there and must work.

Recorded loss

$225,000,000

Orientation and Year One are open: anyone can read them without an account. From Year Two onward the lessons are for enrolled learners, because progress through the later years only means anything if it is tracked against a record.

It is free. We do not sell the list and there is nothing to buy at the end of it.