Knight Capital and its algorithmic trading snafu

Knight Capital

Though not particularly well known to individual–and even some professional–investors, Knight Capital is a very large market-making and trading broker in the US equity market.

algorithmic trading

Algorithmic traders, or “algos,” are typically IT-savvy arbitrageurs.  Like any other arb firms, their business is finding and exploiting differences in the pricing of identical, or very similar, instruments.  Algos differ from traditional arbitrageurs in that they use computer programs to do their searching for them.  That way they can cover more ground than humans, potentially trading more quickly and spotting more opportunities. Computers also execute their trades.

On Wednesday morning, Knight Capital was running for the first time an algorithmic trading program it had apparently developed itself.  The story isn’t 100% clear, but it sounds to me as if Knight was hoping to create an algo service that could be used by individual investors.  In any event, Knight’s computers started churning out buy orders for about 150 stocks at the opening bell.  But the quantities being asked for were huge–much larger than Knight had intended.  And some of the stocks in the bundle were, well, weird.

One of the more offbeat selections was Wizzard Software (WSE).

The issue had closed on July 31st at a stock price of $3,50, on volume of 15,067 shares.  Wizzard provides home health care staff in the West, resells podcasting services from ATT and Verizon, and, yes, it apparently also develops corporate software.  In 2011 WZE had total revenue of about $6.5 million, and lost money.

From the chart I looked at, Knight’s initial order for WZE seems to have been for an astoundingly large 150,000 shares.  That’s a bit less than 2% of the company.  It’s also at least two weeks’ total trading volume. (My guess is that it would take several months to accumulate that amount, if you wanted to do so without moving the price much.  And then, of course, absent a sharp reversal of WZE’s fortunes, you’d have much greater difficulty getting back out.)

It reportedly took Knight almost an hour to figure out that something had gone wrong with its software.  Rival market makers were much quicker off the mark and were providing boatloads of stock to Knight at ever-rising prices.  When the music finally stopped, WZE was close to $12.  WZE was one of six stocks where erroneous trades were cancelled b market officials.

But that left around 146 issues where the Knight orders weren’t simply torn up.  The firm accidentally owned massive (for it) amounts of t=stock it didn’t want.  Once it realized what was going on, Knight cancelled any remaining buy orders and began dumping out the stock it had just acquired.  The company estimates it lost $440 million Wednesday because of the software glitch!!! (To be clear, I think Knight made the correct decision in selling immediately.  The gaffe was too big and too public for it to hope it might trade out of its positions slowly and quietly.)

press comment misguided

Most of the press stories about this incident have revolved around the idea that computerized trading is undermining the confidence of traditional long-only investors, especially individuals, in the integrity of the stock market and the desirability of holding equities for the long term.  I think the stories are  crazy.

For one thing, the S&P 500 only rose about five points in early trading on Wednesday–and then went sideways for most of the day.  If you weren’t a day trader, it may well be that the first you heard of the Knight Capital fiasco was on the news Wednesday night.  Or it might have been the paper on Thursday morning.

The real story?

Consider what has happened to Knight because of its foray into algo trading.

–Its stock has lost about three-quarters of its market value in just the past two trading days.

–The Financial Times reports that major clients have shifted orders to other market makers–Vanguard, e-Trade and TD Ameritrade among them.  Brokers did this initially at Knight’s request.  Clients are remaining mum for now.  Certainly, no one I’m aware of is saying the crisis is over and they’ve gone back to business-as-usual with Knight.  The silence on this score suggests clients think Knight may be badly enough wounded that counterparty risk is a concern.

–According to the FT, Knight has hired an investment banker to help it consider its options, including a merger or sale of the firm.

In other words, it’s conceivable that the management that built the company may soon no longer be in control of it.

I can’t imagine this snafu makes anyone more eager to get involved in algorithmic trading.  Quite the opposite.  The Knight experience may become the cautionary tale that prevents the spread of algo trading away from specialists and into the mainstream of equity trading.

trading errors: what to do if you push the wrong button

trading errors happen

Trading errors–buying when you mean to sell, or buying/selling much larger amounts than you intend–happen.  Not often, but they do.  (For completeness sake, I could have included buying/selling much smaller amounts of stock than you mean to, but that’s not painful to fix.)  Their possibility is the main reason your broker records all the buy and sell instructions he receives.

discovering an error

In my experience, discovering a trading error comes almost immediately on having made it.  So potential damage done before you find out is less an issue than figuring how to trade back out of the error you’ve made.

professionals have more safeguards

Professional investors have safeguards that individuals don’t.  In the US, SEC-regulated managers always employ in-house traders as intermediaries between the portfolio manager and broker.  (The idea is to prevent brokers from bribing portfolio managers.  Of course, the measure doesn’t prevent wrongdoing, just shifts its focus.)  Entering an order that’s 100x normal size will certainly elicit a confirming phone call from the trading room.

foreign securities

Trading in foreign securities is a particular sore spot.  It’s easy to lose a decimal point when the local currency isn’t something you’ve dealt with since childhood.  I’ve also had foreign brokers observe to me that some clients insist on demonstrating their language “skills” by delivering orders in the local tongue–only to botch the numbers badly.  This is a particular problem for Westerners in Japanese, where the number 10,000 is “ichi man” = one ten thousand and not “ju sen” = ten one thousands.  All the Asian brokers I knew had a policy of confirming all orders in English.

what to do

I made a trading error the other day–my first in over twenty years.

As I’ve mentioned in other posts, I own two Japanese social networking stocks, DeNA and Gree.  I have what’s for me a large position in DeNA, a company I’ve followed for years.  I only began to pay attention to Gree recently, when the company made a complaint to the Japanese Fair Trade Commission that DeNA was pressuring third-party mobile game developers not to offer their output to Gree.  To me, that means Gree is a serious rival, so I bought a small amount while I started to research it.

A couple of weeks ago I decided to double my position size.

I use Fidelity’s online international stock service.  It’s very inexpensive, fast and easy to use.  There’s a slight, but noticeable, lag between the prices Fidelity shows and the real-time quotes in the local markets, but I’ve never found that a problem.

I entered the number of shares I wanted, clicked the “trade” button, and clicked again to confirm the order (which was now denominated in yen).

The order was filled almost instantly.  But when I looked on my positions page, I had bought 10x what I wanted.  (I’m not sure how this happened.  My best guess is the following:  as you enter an order online, a drop-down box appears beneath the number of shares box in the order form.  It contains possible order sizes for the number you’re in the midst of typing.  The possibilities include 10x the amount you’ve already entered.  Somehow, I must have scrolled down to the larger amount.)


…actually, worse than “Whoops!”.  Suddenly not only did I have a huge position in Gree, but the Japanese social networking tail of my portfolio was now humongous enough to wag the entire portfolio dog.

I had two choices.  I could either enter a sell order immediately for the entire extra amount and take whatever price the market would give me.  Or I could try to trade out of the position over a period of time, hopefully at higher prices.

One of my first bosses used to say that a situation like this is a choice between fast death and slow death (she was a rather pessimistic person), and fast death is always preferable.  I think that’s true, that the second alternative is a form of denial and will lead to no good.  But I’d frame the issue a bit differently.  I’d say instead that Murphy’s Law is in force with this trade:  if you sell right away, the stock will go up as soon as you’re done; if you hold on, the stock will plunge.  –Actually, as I write this, I realize my thoughts are the same as my old boss’s, only prettied up a little.

So I sold 90% of the extra stock.  I gained about a yen a share on the sale, which came close to covering commissions, but I lost about 1.5% of the principal on currency conversion.  And, of course, Gree went up by about 15% in the following week.  …oh, well.


The main points, though, if you make a trade error, are:

–consider the risk you’ve introduced to your portfolio as a whole.  If it’s unacceptably large–meaning, say, you would be tempted to hide bad results from your spouse–reverse the trade immediately.

–consider that your judgment at the moment is suspect–after all, you created the mess you’re in–so err on the side of caution.

–act according to behavioral rules that have worked for you in the past, assuming you have enough investing experience to know what your tendencies are.

–the most conservative course of action is to reverse the unwanted trade at once.