It was noted at both Mizuho and the exchange by actual humans, all of whom made the decision that they lacked personal authority to overrule the trader. The report by the regulator later was frosting when noting this, AFAIK.

Wow, again! I don't know what their system looks like, but it doesn't seem to me that it should require a lot of authority to bounce the trade back to the trader for a second confirmation -- or maybe send it to a second trader in case the first one has a hangover or something.

I had to stop and think about what actually happens when someone posts an order like that, well outside the current bid/ask. What price(s) does it get filled at? Apparently -- if it works the same in Japan as here -- each bid already in the book would execute at its existing price, despite the fact that the asking price on the new order is far below that. You might think that they would execute at the average of the two prices, but that doesn't seem to be the case, from what I've managed to dig up. An example like this suggests to me that an even better choice would be the geometric mean. But the difference would matter only when someone had screwed up very badly.

Yup, orders that are placed well outside the bid/ask just fill every order in the order book until they are filled. It's commonly called sweeping the market and happens on 1-2 ticks (price levels) around the best bid/ask pretty commonly throughout the day depending on the product. Limit order books are actually really fun things to model and the rules around different exchanges books are quite neat.

The problem with disallowing your trader from ripping through a lot of the levels of an order book is that it can be a risk reducing move and what you intend to do sometimes. This trade is a clear fat finger but there are times when you will want to sweep the book to get hedged.

For instance, let's say your desk just got slammed with a ton of risk on an OTC (over the counter) option trade. You can immediately alleviate a lot of that risk (while paying through the nose) by selling 2000 contracts or 5 price levels of the ES (SP500 future). You can immediately place that order and get it filled and be hedged. If there were multiple points of human intervention required then you might lose a substantial amount of money. 2k contracts on the ES is $25,000 a tick. If word leaks that people are going to need to start hedging big then it could easily move 10 or 20 ticks away from you while waiting for your risk management team to approve your trade as not a fat finger.

Generally it's cheaper to just fire error prone traders. Heh, and anyone that is about to execute a 2k contract option trade generally has their hedge order queued up and ready to send to the market as soon as they hear the other side agree to their price.

Ah, very interesting. I had wondered how option writers managed their risk. I wonder how often this is the cause of the spikes I see on charts.

If word leaks that people are going to need to start hedging big then it could easily move 10 or 20 ticks away from you while waiting for your risk management team to approve your trade as not a fat finger.

That's why I would expect it to be done in software. Yes, I understand that software can be buggy, and hard-and-fast rules sometimes need to be bent, but I would still expect it to be cheaper overall. But I haven't actually worked in the business, so this is just my $.02 :-)

Heh it's options writers or buyers (just options market makers in general). A lot of stuff is electronically traded but there are still some very large orders with huge deltas that are put up on telephone calls.

And yeah it makes sense to have an extra prompt pop up if it's an order over X contracts or Y ticks from the market. And I've seen a lot of systems set up like that.

A lot of times traders will just punch the "OK" box and do their trade though.

That's of course if traders are manually hedging their portfolio/trade. A lot of times they just set their portfolio to auto-hedge based on certain parameters (ie at Z deltas or we've moved C ticks in a time period).

I got some of that trade, too. :)

Fat finger trades like that happen all the time in all markets. People fuck up with probability 1.

Now, automation of trades to look for these errors is a great idea... until you get one that isn't an error (say, an earthquake in Kobe).

... Then it probably got disabled two months later, when some internal bigwig complained that it was getting in the way of their old workflow...

On the NYSE erroneous trades like that can be called back. This just shows you that there aren't "fat finger" trade protections built-in to all exchanges.

These days Japanese exchanges have stricter order validation, e.g. orders with price outside a certain range are rejected outright, and if actual trading takes the stock price to (say) 75% of its opening price, the trading for that symbol is halted for the day.

So the Mizuho blunder couldn't happen again today.

Where do you go to see such offers? I always look at the price displayed by yahoo or google. How do you see these offers and buy from them directly?

This can be done in a variety of ways, but it's unfortunately not usually free. 1) A trading terminal such as Bloomberg 2) get API access to subscribe to the exchange feeds or other data vendor feeds.

Usually, this is out of reach for casual traders. I've seen one exchange publish order book snapshots every 5 minutes on their website, but this data is hardly useful for frequently traded instruments if your strategy is based on the order book dynamic.