Knight's Algo Gets Off the Chain
When I saw the news yesterday that Knight Capital was directing customers to "trade elsewhere", my initial reaction was, "Jesus Christ, not another one stealing customer funds!?!" So I was almost relieved to find out that it wasn't theft but the rather more mundane occurrence of a high-frequency trading algorithm that got off its chain and started tearing shit up. It's a big mess for Knight, and in the half hour the algo ran amok before animal control put it down it appears to have affected more than 140 stocks.
The one that cracked me up was Wells Fargo preferred (NYSE:JWF). The stock normally trades 39,000 shares a day, but for some reason Knight's computers had a hard-on for it, because it clocked over 4.2 million shares traded in the first half hour and got pummeled 10%. Oops.
It's pretty easy for the anti-social cynic in me to see the humor in something like this, but in reality it isn't funny. These kinds of occurrences, which are happening more and more frequently, really undermine investor confidence in the markets. That, in turn, erodes liquidity, which is the main argument I always hear in favor of high-frequency trading.
Knight's shares dropped more than 20% after traders saw extreme volume spikes in a number of stocks, including preferred shares of Wells Fargo (JWF) and semiconductor company Spansion (CODE). Both stocks, which see roughly 100,000 trade per day, had changed hands more than 4 million times by late morning.
Anyway, here's Charlie Gasparino's take on what happened. Early in the morning he put forward the idea that it could have just been a "fat finger" fuck-up. Now he's hinting that Knight could have been hacked, which is a much much bigger deal:
And for those of you interested in the intricacies of what actually happened on an almost trade-by-trade basis, here's some excellent analysis:
What should be done about this? I see a couple of major potential problems here:
- Either a rogue algo went batshit and wreaked a bunch of havoc on the broader market, or
- Somebody hacked Knight and turned the algo loose
Neither scenario is good for the market.
My understanding was simply that an algorithm started going gorillas, not necessarily foul play of any kid, but I have no clue. We had problems with RadioShack in the morning thanks to KCG though, really funny story.
Honestly Eddie, I'm not convinced that these occurrences undermine the confidence of average people in the market. Frankly, most of the guys here on the desk didn't even know it happened until I mentioned it in the Afternoon. I only saw it because I happened to have one of my watch lists up and I noticed one or two stocks go bonkers briefly but then come back. That to me speaks volumes, though, that everything came back to order pretty quickly and strongly. The 'flash crash' undermines it because everyone saw the 600 points down and freaked out. I remember I was at the beach at the time and opened my phone and about dropped it. Then again that might not have been because of the markets but. My point is, I think these are tremendous issues and they need to be addressed but I'd wager the average person has no idea what happened or even that it happened.
Appears they lost 440million from this mess. As a note, KCG off 50% pre-market with losses "severely" affecting their capital position.
Good God! I had no idea the hit would be that large.
Even though Johnny Lunchbucket (and even a bunch of institutions) may have missed this, I think it speaks to the overall fragility that HFT inflicts on the market. Okay, so nobody knew this happened this time. What about the next time when an algo goes apeshit and trips circuit breakers and crashes the market?
I think we need to get on top of this shit pronto.
I agree 100% with you that it needs to be addressed and quickly. As easy as it is to simply let the machines take over and do their thing, you see the limitations on days like yesterday. One 30 to 40 minute span will probably sink an entire market making firm, where they absorbed all the losses from out of position trades. They also said that no customer accounts were affected. Either way, it seems to me that this puts the argument against HFT in the driver's seat going forward. This is why I highly doubt traders will ever be fully pushed out. Machines just can't deal with screw ups.
Frankly, it looks like Mario Draghi does just as good a job at whipsawing the markets as algo's do.
In principle, I see nothing wrong with algo-driven HFT. If your algo is wrong, you lose money (which Knight did). If it is right, you improve market liquidity. A risk reward trade-off, like any other business.
When these algorithms go off the rails, they can decrease market efficiency, making life worse for the rest of market participants. I'm not sure this alone warrants additional regulation- a lot of jobs can create negative externalities is mismanaged. Like if a newspaper publishes an inaccurate story.
To give an opinion on the regulation of HFT, we need to form on opinion on broader market regulation. Do we want it to be a strictly regulated public good, or take a free market approach? I am inclined to choose the latter- even if that means accepting some short term volatility. HFT is still a nascent field, and errors are inevitable. To stunt its development through excess regulation could make US exchanges less competitive.
I'll say I'm out of my depth here. I'm not a trader; somebody who saw a market undergo a massive change in regulation would know better.
From what I've read the algos kinda forgot the "buy low sell high" rule and started doing the reverse, and now Knight is on the hook for the customer bids they put through above market price.
all trades were cancelled, lucky night for knight
No, they weren't all canceled. Only trades that were outside of a 30% band were cancelled, and that was only on about 6 securities. Other than that, they are on the hook for the balance of the trades. Looks like they are asking JPM for financing help because it crushed their capital base so much.
They should ask JPM to take over their trading operations.
They are speaking with JPM this morning concerning an addition to their capital base that has so far been decimated.
The fact that the NYSE cancelled trades is more damaging to market integrity than Knight's Algos Gone Wild.
The fact that JWF could trade 100x it's average daily volume on strictly bid hitting and be down ONLY 10% is an indication that there is healthy competition out there to provide liquidity.
In other stocks with broken trades, how were people to know that it was just a glitch? Only in retrospect are trades "clearly erroneous." Foreign nations nationalize utilities, companies commit fraud, go bankrupt, etc. These types of events ALWAYS hit the stock price before the newswires. During those panic moments is when liquidity providers willing to take actual risk truly provide utility to the marketplace - and they should get paid for accepting that risk.
What if EJ, a Chinese real estate company that had broken trades yesterday, legitimately had problems? A guy in my office bought heavily into the dip and then sold his position. 6 hours later, he found out his buys were all broken, but his sells all stood. This left him short and turned his PnL from hugely positive to negative. It's a bunch of bullshit.
If an order is sent to market and filled, it should stand - no matter what.
No, this was almost certainly a programming mistake.
In all likelihood, none of the managers know what happened, they are combing through the code to see who committed what to CVS and how much testing was done.
They may never figure out exactly what happened or who broke it.
Needless to say, a bunch of Knight's developers probably have brown pants right now. Most bugs are unintentional, meaning nobody knows who is responsible, and in analytics and algo trading, bugs tend to be pretty high profile and tend to cost hundreds of millions of dollars rather than merely crash your Excel.
Charlie Gasparino has never spent a single day in his life programming or working in technology at a large corporate outfit. At every major bank, there are a few "earthquakes" every year where one part of the firm causes millions, even $10s of millions in damage. Almost always due to a bug.
This earthquake simply happens to be a 9.0 magnitude one that is apparently costing Knight hundreds of millions in damage. 2% chance it was hacking; 98% chance it was simply a bug.
Anyone have any thoughts as to where the shares go if JPM provides DIP? They are at $3.50 now.
Knight's CEO was on bloomberg http://www.bloomberg.com/video/joyce-we-have-work-to-do-and-we-re-doing…
I had a large position in KRO that benefited from this mess yesterday. KRO had a massive short interest prior to this so I wonder if some of the bigger shorts used the confusion as a covering opportunity.
More likely they were shitting themselves.
. Just noticed the Joyce video was already linked...
Eddie, I had a different reaction. I saw that there were issues in 140 stocks being traded before I saw the message about Knight telling people to trade elsewhere. I put the two together and realized that one could be part of the other. The other thing is, Eddie, I don't remember Knight being held to the same restrictions as guys like MF Global and PFG. I'm not surprised you didn't bring up the recollection of the Oil Algo gone rogue from 2010. As IP said, there could have been an unintentional bug in the system. From a programing point of view, no code is 100% bug free. It's just that the end user doesn't experience 90+% of them and when a bug happens, it may not always be repeatable.
My guess is that the guys at Knight are doing a line by line review of the algo in question. This surprises me because Knight has been one of the guys on the forefront of Algo trading and they have been pretty good about keeping their shit straight (or at least not to the point of Public Failure).
I think this will be viewed as a One Off and a warning of what could go wrong. Then again, wasn't that was the Flash Crash and every other Algo screwup supposed to be? For Knight, this is a One-Off. For the industry, this is just a reminder of the issues of HFT and Algo trading and what happens when they go wrong.
Frieds,
Yeah, it looks like some kind of logic error. Obviously the program executed exactly what it was programmed to execute.
And I didn't mean to imply that Knight is anything like MF Global or PFG. In fact, I had that thought about missing customer funds before I even read that it was Knight. They're obviously a top notch shop (or maybe I should say "were" if this whole ordeal tanks them).
Any info on KCG bonuses this year? Especially for quants?
Stock price at $3.02 is a BUY BABY BUY, just like Citi stocks in 2008. If they tank, only $3 per share loss. If they move past this, the share price will go right back up the first quarter they post a profit.
Anyone else see this?
The monetary loss isn't the only damaging thing. People have lost confidence in Knight. Fidelity and a bunch of others have said they aren't routing to Knight. Hard to make money as an agency brokerage with no flow coming your way.
I must say - I'm quite disappointed with knight. Goodness. What sort of knobhead application doesn't have an inbuilt 'panic button' based on liquidity?
Common sense - a security drops a certain amount, and a soft alert gets triggered, and manual intervention is required.
This is sloppy. Sloppy design.
I don't think they were hacked in the traditional manner, they may however have fired the wrong person... or they may have two algos competing with each other resulting in unexplainable behavior.
Coupled with the cesspool that is UBS, I think they're toast. I predict some market instability as well.
Spoonfork, I think that is a short term thing. I really do. Look at what happened and realize that it can happen anywhere there is algo trading being done. Algo trading has an inherent risk of something going wrong (The oil algo that raised the price of Oil by 5% in under a minute comes to mind with that one), but it's something that can be overcame. The problem here is that they are so far ahead of the game in terms of Algo trading and doing what they do, but like every computer that uses the same data feed, we don't know whether it was their own algo getting gamed by another internal algo, whether it was their algo versus another external algo or whether it was a real bug in the system. We don't know what entirely caused the crash, and until we do, I'll reserve the right to continue doing business with Knight if I see fit.
Some are suggesting this was (at least partially) intentional...
http://www.nanex.net/aqck2/3522.html
[quote=panchofabricio]Some are suggesting this was (at least partially) intentional...
http://www.nanex.net/aqck2/3522.html[/quote]
"Painting the tape with a fire hose."
Definitely gonna be using that one.
[quote=panchofabricio]Some are suggesting this was (at least partially) intentional...
http://www.nanex.net/aqck2/3522.html[/quote]
First off, awesome find. Second, if that's the case and this was intentional (to test an algo the day new provisions were instated), then what bothers me is what was unintentional. That said, This is Knight doing what they know how to do. I chalk it up to a little bad timing, but its definitely something I'll continue to follow. And +1 SB to you for bringing up Nanex.
I think this is something that needs to be investigated, understood and hopefully corrected so it does not happens again.
Now, I know how investors overreact and the stock of KCG may be oversold. I don't know because I don't follow it. But I do like the idea on opening a position like UFO said. But if they tank, it is not only $3, is the whole investment.
So my question for you guys who might have a better insight, is it a good opportunity? Are they toast for good? Should you wait a couple of days and if so, for what?
Thanks for any recommendations.
Anyone else think the stock will continue to tank today?
Hope you didn't short it...haha
So any of you guys long/short this one?
Probably best to stay away. Whoever knows what's really going on will be making a lot of money off of you otherwise.
Were I a day trader, I'd have made 30+/- cents on the dollar yesterday. I've been focused on the corporate hierarchy until now, but I finally "get" why trading is such a rush. Wow.
30 cents per dollar investment. One day. Wow.
Hope anyone who went long on Friday got out. The company just raised $400 million to stay in business with a preferred offering convertible at $1.50 a share:
http://www.cnbc.com/id/48516238
A fucking dollar fifty.
Buy again! BOOOOYA SKEE DADY
Seriously, this is fun.
Sint cumque illum et dolor aut incidunt. Iure eos neque nulla dolor vero.
Qui doloremque ipsum dolor. Culpa nobis quibusdam officiis quidem quaerat est. Quia sint labore alias veniam voluptatem. Corporis facere veritatis voluptas quia. Repudiandae omnis exercitationem delectus sed architecto.
Sequi et qui laboriosam cupiditate quo consequatur dicta. Ullam iste est tempora et totam aliquam. Qui odio laboriosam ea eligendi sit. Non laborum possimus quas fuga minus velit.
Nesciunt aut deserunt doloremque. Ipsam est omnis quia quis autem ea. Et eligendi culpa ex pariatur enim quos. Corrupti provident officia eum quis maiores corrupti in. Error voluptate fuga officia at assumenda beatae. Illum et quos voluptatem qui est.
See All Comments - 100% Free
WSO depends on everyone being able to pitch in when they know something. Unlock with your email and get bonus: 6 financial modeling lessons free ($199 value)
or Unlock with your social account...
Modi atque corporis nobis suscipit omnis officia porro voluptatem. Aut maiores voluptatem nulla dolorum et. Et ut ut dignissimos est eum recusandae distinctio. Sapiente est consectetur id ipsa repellendus. Architecto voluptatem dolor beatae ipsam eligendi.