Losing Your Company in 45 Minutes

Knight Capital Group is one of the large trader with $1.4 Billion of revenue last year. Per the profile here:


Knight Capital Group, Inc. is a global financial services firm, which provides access to the capital markets across multiple asset classes to a broad network of clients, including buy and sell side firms, dealers, brokers, institutions, and corporations. It operates business through four segments: Market Making, Institutional Sales and Trading, Electronic Execution Services and Corporate and Other

So Knight is not a ‘fly by night’ ;-)

Yesterday August 1, they decided to turn loose an automated trading platform. ( I don’t know if this was an ‘upgrade’ to an existing platform or a new venture). It “had issues”. How much of an issue? They did 20 weeks of trading in 45 minutes (so perhaps someone set a time parameter wrong…) In 45 minutes of automated trading it lost about $440 Million. In comparison, it made about 1/4 of that in each of the last 4 years. So 4 years of total earnings, flushed in 45 minutes.

Net Income


It caused a large number of stocks to plunge, and another group to shoot up, just from the artificially high volume being created by this one trading program “gone wild”. The volume spikes and price swings caused the Talking Heads on the financial programs to comment at the time that something was up. In the days since, the news has been about the nature of the fault and the likely result.

Today it has the CEO using those words like ‘strategic options’ and ‘seeking financing’ and ‘exploring options’. All mean “get the hell out if you are not already”. Citybank has issued direction to NOT route orders through them. Others will be doing the same. They are most likely “toast”. I would expect a “buy out” by someone who just wants the customers, assets, and some of the staff. (Clearly not the automated trading guys…)

So how would you know to “get out”? First off, behind any and all positions you ought to have a “stop loss order” that automatically executes if things are just “crazy wrong”. For active trading, using a 5% stop loss is common. For investors, a 10$ band. If you are watching things (daily or weekly) you can adjust these down if you see a ‘dip’ where you don’t want to sell, but want to add to a position. They are there to get you out when you are NOT watching and adjusting them…

But when you ARE watching, how do you know to exit a stock if the “news flow” isn’t saying anything?

Lets look at the 10 Day chart. In this case, with 15 minute price bars.

KCG 10 day 15 minute 2 Aug 2012

KCG 10 day 15 minute 2 Aug 2012

I’ve added to this chart a “volume by price” indicator up top. Each lateral red bar from the left margin is an indication of the total volume trading at that price in the entire 10 days.

Notice that price was relatively flat for 8 days. The first three ‘volume by price’ bars are fairly small (compared to the rest) and cover all that volume over all those days. Then, yesterday, the fall started. Initially it will have been part of the general trading disruption it was causing with the bogus machine trades. Then the news will have spread (via rumors at first) to the trading desks that KCG was the cause. That’s when you see the volume really pick up as the price drops. Some folks dumping long positions, some “stop loss” orders being hit (eventually enough to swamp the ‘buy if touched’ orders as they are being withdrawn by some parties). As the news goes public, the volume picks up even more as the price plunges.

The next day, once even more folks have ‘heard the news’ and know how bad it is, the stock opens “gap down” hard. Then the fall continues… At this point, down 62% today.

So when was the “right time” to get out? Frankly, ANY time is the right time. The sooner the better. A 10% (or even a 20%) stop loss order would have been a very welcome thing in retrospect. ALWAYS take an early loss. ( “The first loss is the best loss” is an old market aphorism.)

Would our ‘indicators’ have helped? Yes. Notice that RSI had a ‘near 80’ peak with a ‘lower high’ on Monday ( 4 days ago) and was saying “be out soon” even on a fast swing trade basis. MACD (though hard to see due to how large it becomes in the last two days) was ‘red on top’ and slope headed down three days ago (Tuesday) so saying ‘get out now’. Anytime Wednesday the MACD was
below zero and headed down hard. Shouting “Get Out Now”. The Black ADX line was above 25 (so strong trend) and had DMI- (the red line) cross over to on top last Tuesday, confirming the “be out” call.

Realize none of these are predicting the disaster that happened, they are just saying “time to be out” based on the preceding price action. Once the plunge started, they would have turned highly negative even if they had been saying “buy” before the event. Basically, they would have been telling you “take the first loss now.”

But what if you are not a swing trader and only looked at the daily charts? (Remember that you can click on the charts to get a much cleaner more readable version):

KCG 6 Month Daily 2 Aug 2012

KCG 6 Month Daily 2 Aug 2012

You would have been out last March… That’s when MACD crosses below zero and is headed down fairly hard, or even Jan / Feb when MACD first goes to “Red On Top” and RSI is dropping from a high point. DMI/ADX are slower to react on this scale, so we look more at ‘inflections’ and expect crossovers to be confirming not predictive. Even here DMI+ (blue line) has inflected to a downward slope in February and ADX is dropping to ‘low trend’ at 15 saying “the run up is ended – shift to swing trades.” Later it goes to DMI- (red line) on top saying “Short or be out is the bias”. The last favorable ‘swing trade’ to the long side was in April. After that the trades are most productive to the downside as price is clearly below the SMA (Simple Moving Average) stack, MACD is below zero, and DMI is “red on top”. Odds are that trading on this time scale you would already have been short this stock and have made a bundle…

Here is a longer view, a 2 year daily chart:

KCG 2 Year Daily 2 Aug 2012

KCG 2 Year Daily 2 Aug 2012

Looking at an even longer time scale (long term investors, checking positions weekly):

KCG 10 Year Weekly 2 Aug 2012

KCG 10 Year Weekly 2 Aug 2012

On this time scale, you would have exited any position back in 2006/7 or 2009 (depending on your activity and when entered or reentered) as that was when RSI had ‘lower highs’ while price cut through the SMA stack and MACD went to ‘red on top and falling’. There were a couple of swing trade opportunities (as folks could move to faster charts for them) but sitting out is the style at the typical person using this time scale. DMI- has red on top saying “just stay out” most of the time since then. Even the DMI+ ‘blue on top’ of early this year (with MACD ‘blue on top’ and above zero) was in the context of an ADX line below 25, so saying “faster swing trade – move to daily interval chart”. So investors would stay out.

Does any of this mean it would save you from a fault like this?


Mostly it is the 10 day chart that would save you (if a company that was doing great had such a fault). The longer time scale charts were not predictive of this kind of fault. However… Companies that are “doing great” are usually doing it for a reason. Companies that are “having issues” are usually “having issues” in many areas. So simply being out of a company that was “having issues” avoids proximity risks.

In Conclusion

Yes, it takes time. Yes it’s a PITA to do charts all day. I get tired of it. When I’m ‘taking time off’ I’ll put stop loss orders in place behind large positions. Or I’ll just move to cash.

The simple fact is that, due to companies like KCG doing “high frequency algorithmic trading” it just isn’t safe for the small guy to do “buy and hope” investing. In this case it killed them. Sometimes it kills the rest of us. (Anyone remember the “flash crash” where HFT drove the whole market down 900 points in hours?) Even being in things like “index funds” doesn’t escape that kind of thing.

So what can you do? Well, part of it is to use “index funds”. Holding SPY or even XLF (financial fund) would have avoided this “company specific risk” while getting more stable and predictable movements (better behaved with the indicators) and less ability for a single trader to ‘move the stock’. Then having “stop loss orders” in place and paying attention to your positions at least once a day (about 2 hours into market open is a good time, or at least an hour before the close).

It is this kind of ‘action’ that has caused me to move from “value investor” to “hedge fund” in my style. The removal of the “uptick rule” for shorting lets companies do the “long / short” trades that are most often used for this kind of HFT more easily. (No worries about shorting whatever you want to short). Moving to decimalization lets folks do trades on a single penny of range (instead of 1/4 or 1/8 dollar spreads) so increases frequency. In short, most of the changes of the last decade have encouraged this style of trading (which is no surprise since these kinds of companies where asking for those changes…)

The other thing you can do is to diversify across more market segments. Not just US stocks. Add in some bonds, metals, other commodities, REITS. Then even a HFT in SPY will not jiggle your whole boat. At this time, about 75%+ of ALL trades are computer generated. It’s just crazy. The stock market has largely just become an automated casino with dueling computer programs and some investors doing “duck and cover” every so often when one of them blows up.

Oh well… As they say: “Sure the game is rigged, but it’s the only game in town!”…

But by using some charts, paying attention every day (or two) to your positions, and using some stop loss orders, it’s survivable. Sometimes you can even make a bit of money…

Subscribe to feed

About E.M.Smith

A technical managerial sort interested in things from Stonehenge to computer science. My present "hot buttons' are the mythology of Climate Change and ancient metrology; but things change...
This entry was posted in Economics - Trading - and Money and tagged , , , . Bookmark the permalink.

18 Responses to Losing Your Company in 45 Minutes

  1. BobN says:

    EM – A very good analysis. When I first started trading everyone seemed to be a value investor and I was told over and over, don’t play the market, play the stock. Hang in there for the long term. That was a pure crock in my opinion. I held on way too long and was slow to get out, when I did I was taking big loses. I finally got smart, used stop loses to get out, no exceptions. Once I started pulling the trigger fast I started making money. Long term is now 6 months for me as business conditions change too fast. I’m a value investor until I see a change. If you play by the old rules, you can lose years of profit while your stock is down, its best to get out quick and use that money to trade and make money, not set dormant in an account waiting to come back so you can get out.

    Instead of buying stock that I really like I buy leaps typically a year but better if its 2 year leaps. Use the leaps in place of stock and its a lot cheaper. Also, write covered calls against your leaps. Typically you can pocket money monthly by selling out of the money calls, this money usually pays for the leaps over the coarse of a year or two, so you end up with free leaps at the end. To do it right you need to study the trends to properly pick the right calls to sell.

    Take losses fast and don’t ride it down. People get attached to stock and have a hard time selling. This is the biggest mistake many traders make.

  2. KevinM says:

    Something stinks. I’d like to talk to an insider about the other things that were going on before the subject event. They did not run their program on simulated buys and sells at the ticker prices first? They did not have alarms on size of losses or even number of trades?

    I wish I trusted the FCC to find the real problem and expose it.

  3. E.M.Smith says:


    The news flow ( right or wrong ) is that they turned it on, looked good, so went away ( i.e. nobody was watching it as it ran) and only after market impacts caused them to notice something was up, went back and looked at what was happening.

    Reading between the lines (not a lot of tech talk on it…) it sounds like they had a 20 week ( or longer) trade plan that they loaded in, and it got executed in 45 minutes. That can be a “fat finger” on the model load. ( i.e. someone put a “1” in the minute field instead of the day field… it’s hard to make things foolproof as fools are so creative…) or it’s possible that they DID do a test, but not wanting to wait 20 weeks, set the time step up… then forgot to set it back down on the rollout to production… The other possible I can think of is that there simply was no time step loaded in. They intended to do it in steps, but someone just didn’t enter that parameter, so it just ran open loop.

    Per “people in the loop”: I’ve lost so many arguments with upper management about the need to keep a person in the loop to avoid catastrophic loss that I have no doubt there was a deliberate decision NOT to have people in the trading loop…

    Ever have a $40,000,000.00 asset and need to justify a near minimum wage guard – just one – for watching the place during shutdown / off hours times? “You mean you want them to sit around and do nothing?”… The buzz phrase for a long time was “dark site”. See “lights out” or “dark data center” here:


    I’ve been on site at a major manufacturer as the only person at a very large data center for the entire weekend. ( It paid well with a lot of over time ;-) During lunch / sleep break the place was abandoned. Depending entirely on automated monitoring.

    The general expectation these days is “turn it on and walk away”… lights out as you leave…

    One of those “trendy sounds good on a projected slide” things that is really dumb when things go wrong…

    At our Cray site, we had 24 x 7 operators even though the night shift didn’t do much other than swap backup tapes. Oddly, they were partly justified based on another bureaucratic FUBAR. The City Code demanded sprinklers in the computer room. Never mind our incredibly expensive and highly effective HALON system (which we had to test with a gas dump for the Fire Marshal…) One of the sprinklers had to be where it would dump water on the Cray. (Despite our protests that dumping water on a device being fed from a 750 kVA power supply being a ‘bit dumb’…) So one of our operators Major Standing Orders was: IF a fire alarm goes off, make sure the Halon wins the race condition… or hit the “dump suppression” button if there isn’t any fire ( it had some seconds / minutes of delay so a false alarm would not dump several $Thousand of HALON for nothing…)

    So yes, BY DESIGN, the system was made to “Destroy the Cray in order to save it” by a water dump… mandated by local government. But at least it let me have a 24 x 7 operations staff ;-)

    So I’d not be too sure that “rampant structural stupidity” isn’t part of the mix any time Management meets Data Center Operations… (For a very long time I.T. at many companies reported to the Finance Guy, who typically didn’t care about much other than cutting costs…. and “why don’t we just use Microsoft for everything”… It explains a lot. Government regulations covers a lot of the rest…)

  4. P.G. Sharrow says:

    So far the last I’ve heard is that they installed a new program, turned it on and went to lunch, and it executed its’ commands. By the time somebody realized there was a problem, the orders were being executed and the company was broke!
    Today the companies brokers were shopping their resumes around town.
    I was watching the market open this morning and the floor traders were running around covering their butts and making sure they were not hurt by this debacle.
    To error is human, to really screw up you need a computer.!
    Can’t say I’m sorry to see computer traders cutting themselves on their very sharp trading tools. pg

  5. Another Ian says:


    Did they hire some of the IT people that helped the Queensland Health payroll problem?

  6. Dave says:

    Sci-Fi always portrays AI destroying the human race with weapons. Will be ironic if it is with finance. HFT should not be allowed in my opinion. Just because it can be done does not mean it should. Human beings should never be take out of the “loop”, when it comes to pulling the trigger. Be it with weapons or financial trades.

  7. Pascvaks says:

    “2012: A Head Space Odyssey”
    This really would make a great movie or mini-series. Unfortunately, the LA ‘Follywood” Mob hasn’t been up to anything remarkabley good in decades. Everything today is cut and paste dialogue tid bits, gay-this-and-that inuendo, long perspective scenic shots of Manhatten or the Grand Canyon, pathetic music, and wanna-be writers, actors, directors, and producers, who have no idea what life is like or what they’re trying to say. On second thought, trash my bright idea, maybe some guy at a blog will write something classic someday and some Brits will do the rest.

  8. tckev says:

    There appears to be nothing to stop a rogue nation (group of nations?) from building up a large stock portfolios, trading them normally using automatic computerized trades to build up their bona fides. Then at a moment of market crisis, that gives these rogues a political (or financial) advantage, they launch a coordinated chaotic flash trading that causes a significant loss of confidence in the US (and/or Western) stock markets.
    Are there any rules within the stock market to prevent such a scenario, or is it ‘every man for himself’?

  9. adolfogiurfa says:

    A virus?……………………..

  10. Pascvaks says:

    The Next War: Not with Horses, not with Muskets, not with Canon, not with Ironclads, not with Battleships, not with Flattops and Airplanes, not with Jets and Superbombers and Stealth and Nucs, not with UAV’s and Robots, the future is, and will be, fought with Software! (Kind’a poetic isn’t it? I always felt rather Soft myself. And the buttons I wore were too dang BIG too;-))

  11. P.G. Sharrow says:

    It appears that the newly installed algorithm was created to take advantage of loopholes in the recently adopted trading rules to prevent such problems. Seems in the last few minuets of trading and the first few of the next day they found a way to gain advantage and profit with computer executed shotgun trades. Too bad they used a computer to execute judgement. When it comes to judgement computers are even dumber then bureaucrats. ;-) pg

  12. Pascvaks says:

    Oh how I wish that NBC would lose the broadcast contract to the Olympics in the next 45 minutes… what did we do to deserve these idiots, their interviews and programming are pathetic, and the corpse they have on the anchor desk needed to be buried 128 years ago. What did we do? Why do we have to watch ABC for eternity? They act like they’re announcing for the Soviets, the Chinese, and Iran. Don’t they have any red-blooded, Mom and Apple Pie Americans. Am I dead? Is this Hell?

  13. Pascvaks says:

    “Why do we have to watch ABC for eternity?”
    No! No! I meant NBC!
    Sorry, bad freudian slip-up there; was thinking of the ‘old answer’ to the current problem as I typed (I guess;-) and slipped them into the tirade about the NBC bozos running the show now. Should read: “Why do we have to watch NBC for eternity?”
    Doubt that ABC today could come up with another host like the late great James Kenneth McManus (September 24, 1921 – June 7, 2008), better known by his professional name of “Jim McKay”. What a guy! Disney owns ABC now, right? They’d probably have a bunch of cartoon characters doing the show or hire the nits at NBC to ‘carry on’; nothing stays long on planet Earth. Especially the good stuff. (We really do need another Little Longer Ice Age to clean up the place and remind us again why we’re on this rock in the middle of nowhere — there I go again, mixing eternity with a Wall St, I guess I really am repeating myself on a recurring theme. If I get too bad, bang you’re fist on the table or stomp the floor, used to work when records got stuck in phonographs*.)

    * – The phonograph, record player, or gramophone (from the Greek: γράμμα,
    gramma, “letter” and φωνή, phōnē, “voice”), is a device introduced in 1877…

  14. Hal says:

    It looks like the programmer who set up the San Diego fireworks on 4th of July got a new job at KCG.


  15. dp says:

    How do you sell when the price is falling on light volume? Nobody’s buying and nobody will when the plummet trend is seen. Not that you can’t try.

  16. E.M.Smith says:

    @Another Ian:

    I think there are plenty available locally, but perhaps they used some imports …


    Well, it would be hard to make a rule that forbid only the bad HFTers. Take a company like Morgan Stanley. A LOT of trade volume. Somehow you would need an audit process to sort out the very large client volume from the very large proprietary volume; then divide that into trading desk vs people driven… Probably worth trying, even if it is hard to do.


    Watching modern computers screw up isn’t very dramatic. They don’t have the blinkey lights of the old ones ;-)

    Yes, software is ‘a big issue’ now…

    Disney owns ABC, while GE owns NBC. Pretty clear why the NBC coverage looks like something from a long fossilized bureaucracy now, eh? ;-)


    There are ‘circuit breakers’ for prevention of whole market moves ( like the Flash Crash that caused the rules to be adopted) but $440 Million is just so puny as to not even register… For a $10 Billion Investment Fund, it could be a simple ‘modest’ position / buy…


    Nope. It is known to be a human / programming error. No virus.

    @P.G. Sharrow:

    At the open and close there are known market inefficiencies. For example. at the open the “Market Maker” will take a stock both up and down in price to find where the volume is happening. Knowing that, you could try pushing that behaviour both ways and doing some ‘buy near the high sell near the low’ on top of that. At the close, whatever was dominant has a fade most of the time (as day traders exit their positions).

    @R. de Haan:

    HFT and computer driven trades are about 75% of volume on some days. Gonna be hard to convince brokers to walk away from that…


    I think they are cousins ;-)

    Used to be that folks paid decent money for good admins and programmers. Then they got the H1B visa bug and hired a flood of cheap folks from India. (No slam on them, some of them are pretty good). But it drove the prices down so much the best folks found other things to do… So, for example, were I running either of those two ‘shows’, the problem just would not have happened. Just not in my ‘style’ to allow for the potential.

    But folks don’t want to pay for my level of care any more. From the mantra of “Fast. Good. Cheap. Pick any TWO.” They are picking fast and cheap. Oh Well…

    (No, not ‘bragging on me’. While I’m pretty good, the complaint is more about how much effort they will allow for ‘non-productive’ things that prevent problems rather than tries to fix them after the fact. I’ve had a lot of times that companies have chosen “fast and risky but cheaper” even with full disclosure of the risks. ( “In their grill” with the risks sometimes…) )


    Well, since most of the volume is HFT, and most of what’s left is Institutional, for “just plain folks” you don’t even get into the noise level. There is a “Small Order” system that bypasses pretty much all of it. Any small lot / round lot orders get an automatic fill via computer based on a the “bid” programmed in by the market maker.

  17. tckev says:

    I found this interesting piece about ‘Flash Crash’ on the BBC.

Comments are closed.