Ok - first, disclaimer: I'm up here in the woods of New Hampshire. I don't have Bloomberg, I can't see how much stock printed at each of the crazy prices, and I was out buying plants at the time the crash happened, notified by a text from the Big Show that read simply: "HOLY FUCK." BUT, I do have a lot of experience with market structure, index arb, electronic execution and the like, and I am still in close contact with several senior people in these roles at major firms - I'll get to what they have told me in a minute.
Now, here's what I do know:
1) people are talking about a fat finger error in the e-mini futures. That makes no sense to me. If there's an error, I think it HAS to be in the "cash" basket - in other words, stocks, not futures. That's the only way you can get stocks like ACN to trade down to pennies. YES - pennies. Futures errors wouldn't do that. Complicating issues, though, SIZE traded in the e-mini SPX futures as this was happening - $15B or so in notional... So that makes it look like futures were involved also! Or perhaps that was everyone else in the market panicking to sell the most liquid asset as they saw the stock prints.
2) another initial thought I had was related to a discussion I had with a friend a mere two days ago about "peer factors." Peer factors are, dumbed down, the way to describe how a quant model will model the fact that everyone has the same inputs in their model. If all the models are correlated, then, when they are triggered, things can go bad in a hurry. HOWEVER, a buddy of mine who runs quant trading at a major firm told me that he has several of these models running currently and that they were fine today - it wasn't a Terminator-esque case of the machines becoming self aware - at least not in the peer factor quant models.
3) The rumor was that Citi had a major execution error, yet another buddy of mine told me he had a lengthy call with them, where they denied it.
So, how can this happen? Well - some execution systems will allow you to type in the size of the basket. The rumor is that some Joey Baggadonuts meant to type in 15M (for million) but typed 15B (as in... BILLION) instead. Now, I can't believe that guys can still do this. It boggles my mind that there aren't risk controls in place. When I was trading, we took great lengths to make sure that you couldn't do this - even if you had fat fingers, if you dropped a can of soda on your keyboard, if a ferret ran across your computer, or if you suffered from some sort of dementia and just went crazy - you couldn't send an order like that without being 100% sure that you were doing it, if you could even do it at all. When we moved to a new execution system from a third party, we mandated that they install confirmation screens on every order entry window. "The devil is in the details," my boss always used to say...
Each trader had execution limits - that meant that I couldn't send a $1B basket no matter how badly I wanted to. Pure and simple - my systems login wouldn't send it. Now, on certain days, I'd need to execute billion dollar baskets, so we'd have to change the risk parameters - but in general, there were caps on the size of each basket, the total volume for the day, and extra warnings if you tried to send an order that was even 25% of your max quantity - you'd have to click through confirmation boxes of the type "ARE YOU SURE YOU WANT TO DO THIS?"
Anyway, what will happen in the aftermath of today's frenzy? One contact told me that they did some quick math and figured that in S&P500 stocks, $31Billion of notional traded below the 1100 level in the SPX. That is a lot of stock. A different former colleague told me "I think people figured it out and bought it back ahead of the offender. I also think the person with the mistake had time to cover." I'm also told that all trades more than 30% off of stable last sales after 2:30pm will be busted. THIS IS WHAT I'M TOLD - THIS IS NOT A GUARANTEE. In other words, if you scooped ACN at 10c, you're probably getting that trade busted.
we'll await further developments...
-KD
22 comments:
What do you mean by trade getting busted?
student - they make it so that those trades never existed - if you "bought" the stock there, they cancel it. again, you should call your broker if you're involved in these - i can't be sure what will and what won't get busted.
Holy Fuck!
I'll second that.
From an options pit at the CME...
Slow and controlled selling until about S&P 1110....
Thats when all hell broke loose. I heard it may have started in some Nasdaq stocks that were getting pounded with market orders, but I saw with my own eyes the e-minis get SWEPT from 1110 down to 1065 or so in about 30 seconds. I thought the world was ending. A literal vacuum.
e-minis normally .25 tick wide were 7 or 8 dollars wide. Not a ton of volume traded on the lows and we rallied 40 handles back on air. The whole thing took ten minutes.
Freaky.
more thoughts later...
KD,
thanks for the inside type info. I cannot add much except that the idea that stocks have been going up on low volume ramp jobs and algo buying crap was called tin foil hat stuff, but this drop is explained away as just some glitch in the matrix. You cant have it both ways it seems to me?! How come rhere is never an error when stoicks go up? Would they really tell us if it was indeed a 1987 type mess with the programs? The wheels, while not off, are getting loose.
could this have been a massive tree shake? there are always some clever guys looking to make a mark by spotting some flaws in the system?
According to Reuters, trades 60% off 2:40 print are canceled.
Curious to see how that will affect the day net net. Total disgrace.
Eamonn, GYC - well, yes - i think the market SHOULD be going down - but not like this - this just doesn't make sense. you had a broad array of stocks and etfs trade down to PENNIES. i've never seen anything like it.
i had another theory - perhaps everyone in the market who has been riding this bull had STOP LOSS orders in - that means that if, say, PG is trading at $60, you put in an order to sell it if it hits $50... but guess what - if everyone does that, and all the stops get triggered, it just releases a massive sell wave. that's why you shouldn't use stop orders (unless they have limits - so that when PG hits $50, your order gets put in to sell it at $50, not at market)
I thought of that when I read somewhere "my buddy got stopped out of (some ETF) at 11c"...
which gets to Eamonn's question - could someone deliberately shake the tree to trigger stops. sure - but you'd have to be suicidal to try to manipulate the market like that - all this stuff is electronic, and there's a record of all of it - and if someone was trying to do just that, it's blatant market manipulation, illegal, and crazy.
But, But, But, I thought the HFT guys provided all the liquidity we all need? As for the electronic record, I will hold my breathe for the investigation.
OT,
If you read The Automatic Earth, site the writer Stoneleigh will be in Mass (northhampton) May 21. Let me know if you want to go and listen to a seminar, maybe we could meet up?
GYC - i don't read that blog...
and don't fall into the ignorati trap. the claim was never that HFT provided "all the liquidity we need" the claim and the truth is that it provides liquidity, which is good. ANY liquidity is better than any lack of liquidity.
OK, I just go by what I read by the defenders, maybe they are full of chit.....which is my point. Sarcasm and all.
The neo-Luddites over at ZH are calling for the end of HFT... ugh..
Let me offer this analogy.
My father is a pilot. He flew in Vietnam and then as a corporate pilot flying hotshots in Lear Jets, and now Falcons and G5's.
The aviation industry has seen, like all industries, enormous technological innovation. From an era where a pilot could literally "feel" the wind, to now, whereby you enter a few coordinates and the GPS does the rest.
Occasionally, the computers make an error, a chip goes on the fritz, or a wire gets crossed, and the pilot who had become a touched outdated and unnecessary quickly becomes very needed. The fact that this occurs doesn't render all the technology evil and useless, but rather it reaffirms the need for a hybrid system, and the watchful eye of a seasoned pilot.
Trading is, and will always be, an incomplete information game, and thus - a human one.
Today's market action reminds us that technology is going to have its problems. And today's excessive market volatility shouldn't be taken lightly. That being said - the technology revolution has (in my eyes) improved the world, but of course isn't cost free in any sense.
Today's market action had all of the characteristics of a typical crash, albeit it happened across 15 minutes and not 2 days. But the song remains the same. I am certain all segments of our society enjoy these growing pains with respect to technology.
Looking forward to an insightful KD post rebutting the lunacy that is about to be brought forth regarding the evils of computerization in the trading world.
yeah -Scharfy - i'm not sure i want to get into that debate again... here's the thing: the price action today is SCREAMING either 1) error or 2) algos gone wild. now, i'm a little surprised that an algo guy i talked to said that his models didn't have a problem today...and the WSJ already wrote a post about how some other HFT guys turned their computers off, because they certainly don't want to try to navigate that chaos.
the real question is, though, why cancel the "bad" trades? if you're going to allow that much volume to trade, you have to hold the guys responsible... i don't know how you get around that. it's BAILOUTS all over again - in another form. And yes - i believe that these bailouts are designed to help mom and pop retail - who got plugged on stop-loss orders. what we SHOULD do is educate people not to use those orders also, especially if we're going to let them off the hook.
seems like it was "real" selling in the sense that many different entities were participating in the sell off. I'm not sure what the trigger was, but when the market is weak - is doesn't take much.
Given that volatility is (by definition) the rate of return across a given time period - I think freaky speed of the whole sell-off frightened the hell out of people. The speed was awesome.
Wild stuff...
regarding HFT...
One of the great things about working in New York City is lunch. Tons of choices, very convenient, etc. But what if .. somebody learned to build super fast robots, that would come to your desk with whatever you want. This would be slightly better, and a lot of existing lunch places would go off and do something else. Then, one day, there is a 10 sigma snow storm. The snow in the street is too much for the robot wheels. No lunch that day. Hunger instead.
Think about how 12 months of artificial low volatility (like we've just seen) causes all the guys with good intraday reversion strategies to leave the market. A small minority that is faster and more levered can dominate.
Could a human trader sit on his hands during a crash? Sure. But the risk of a lot of quant and other balance sheet intensive strategies would be better recognized. More people might actually study securities, and crashes would be less necessary. Without HFT, what we would lose in efficiency (definitely don't mean market efficiency here) would be gained in stability.
Plus I think those machines get information when they "provide liquidity" which they use to front run flows.
Total BS to insist it must have been an error. There were market orders and no bids. That's the real market. If you bot SP at 800 a year ago, why not sell it at 1050 amid the new info of a panic?
What happened to earnings prospects since Feb when SP was last at 1050 is of same magnitude to what happened this afternoon.
I am not a Luddite and I don't comment at ZH.
oddlot - i'm not so sure about an error... in any case, the price action is absurd, and cannot simply be chalked up to "the market deserved to go lower since the rally was a load of crap"... I FIRMLY believe that statement, but it doesn't explain stocks trading down to PENNIES... I have another post imminent...
Hi Kid,
not for nothing, but the thing abt confirmation screen is that when mkts are moving fast and you need to key in a trade fast, you tend to ignore what the confirmation screen is telling you and just click through blindly.
I do agree with you that it's unlikely to be an error trade as the trade size is too large and would exceed normal trading limits (for an individual, not firm-wide or anything)
Why 60% as a cutoff? Why any cutoff at all in the complete absence of evidence of errors? This is all very mysterious to me. How the heck can liquidity be zero during a 900-point drop? Too too weird.
yeah anon - I totally understand and agree with that... but i'd still be surprised that firms don't have trading limits to prevent this kind of thing. It's not hard to put in a line of code that physically prevents someone from executing an order of a certain size.
I was long 5 AAPL 200 puts for $.10 basis as part of a trade where I was short a different AAPL put. It mitigates the margin requirement. When things went to hell I put in a STC for $5.00 when the spread was $1.00 by $5.15 and it executed. If they bust this trade I will be beyond pissed. I keep this position open for the possibility of this very exact circumstance occurring. How does a mistake in PG make AAPL trade down to $195 for 10 seconds? Many many stocks did not print these absurd lows that existed for mere seconds. I don't get the mistake argument.
Ha! Check this out!
http://www.nyse.com/events/1273053647643.html
The Procter & Gamble Company Celebrates Mother’s Day at the NYSE
AYFKM?
5/7/2010 -- no freaking way
daniel - keep me posted on if they cancel your options trades. i think it's insane if they do, but i'd love to know...
side note: not advocating the mistake argument, but no one is saying it's just a mistake in PG...
Post a Comment