A web-reading bot made millions on the options market
slate.com
slate.com
Robbing who? This is about as far from robbing as I can think of.
1. Someone offers to sell a bunch of options, because he thinks he'll make more money from this than if he didn't sell those options.
2. Someone buys these options because he think he'll make more money than if the didn't buy them.
3. Time reveals the guy who bought the options was right, and he makes money.
This is simple speculation, and has nothing to do with robbing.
If you'd ask me those derivative products are a convenient alternative to casino style gambling and mostly add instability to the market.
I would love to hear an informed opinion, though. What do these products solve, why are they essential? Also, what would happen if we would tax them more heavily, perhaps incrementally over a period of time?
For example, suppose someone has a long term stock portfolio -- a classic fundamentals investor with a decades-long time horizon, the farthest thing from a casino style gambler you can find in the stock market. Maybe it's a pension fund or an institutional endowment. They very much do not want to take highly risky short term bets. They are only interested in safe, slow, long term returns.
But markets can only provide that in aggregate. In the short term, markets are very unstable, and individual stocks go to 0 if the company folds. The long-term investor thus constructs a portfolio of many small investments rather than a few big ones.
The portfolio manager has reason to believe that XYZ, one of the portfolio companies, is about to go bankrupt. This is not certain, but the stock has already begun to drop. If the stock drops to 0, the portfolio stands to lose a lot of money (despite continuing to exist overall because of diversification.)
The portfolio manager can buy put options to hedge XYZ. These give the holder the right to sell XYZ stock at a certain price, regardless of its current market price.
In this case, they function as a form of insurance. If the rumors are false, XYZ stock will bounce back up and the put expires worthless. If the rumors are true, XYZ drops to 0, but the portfolio manager can still sell his XYZ stock to the speculator who sold him the put option at a pre-determined price.
This is no different from buying fire or flood insurance on your house -- if the value of your house goes to 0 through disaster, you can in essence sell that worthless house to the insurance company for a predetermined price and buy another house. Options are the stock market version of that.
If you tax them more heavily, that will make them more expensive to use, which will discourage people from using them and arguably increase instability while decreasing growth in the stock market as everyone will be forced to carry the full value of any losses themselves, with less possibility for insurance.
However, I still don't see the necessity of these instruments.
For example, wouldn't it be more natural for a pension fund to just spread the risks by buying different stocks?
Also, I'm not convinced that derivatives would decrease instability. Like I said, I would guess they do the opposite.
Perhaps there is some theory that can show this (something akin to e.g. the fact that passive systems are always stable, for some definition of "passive"). Or perhaps there exist empirical simulation-models that can shed a light on this?
Options also allow you to bet on more specific stock movements. Maybe there's a merger rumor and you think the stock will either go up (merger goes through) or down (merger fails). You couldn't make that bet with a static position on the underlying, but you could buy a put and a call (straddle).
That's like saying, "Wouldn't it be more natural for a homeowner to just spread the risk of fire by buying multiple houses?"
Yes, in theory you could buy 2 houses and just move into the 2nd house if the first one burns down... but it's much more cost efficient to buy one house and get insurance on it. You'd have to buy 2 inferior houses instead of putting all of your money into 1 nicer house and paying a small amount for insurance.
There are lots of theories and studies on the topic of stability, though it is very hard to draw firm conclusions as the economic and financial conditions at any given moment are unique, making repeatable experiments difficult or impossible.
Another complicating factor is that you can make lots of money by discovering a way to make market systems more stable that nobody else knows about, so much of the research on the topic is secret.
There are many examples of failed derivatives instruments -- that is, derivatives instruments that once existed and no longer exist today because of lack of interest in using them. They performed an insufficiently useful economic role, so nobody traded them. Today, for example, there is an ongoing effort to list real estate futures, but there is little marketplace interest and they will probably be delisted within the next few years.
The derivatives that survived survive not from mere inertia, but because people (not some abstract "system," but real people) find them economically beneficial in some way.
I wouldn't say this is necessarily a bad thing myself just pointing out this is definitely not the spirit of what options are for.
I don't understand the mystery here. It would be trivial to write a bot that could interpret a headline and make a trade on it in under a second, and that's without any special high-speed links. With API documentation for a brokerage house, I could write such a bot in probably half a day. Given the astronomical sums of money involved, I have a hard time believing that this isn't occurring on a daily basis.
This is indeed occurring on a daily basis. A number of funds trade automatically based on fully automated analysis of the news wire and other sources.
I believe most use in house analytics but there are 3rd party solutions:
I think people outside the industry don't appreciate the depth and breath of automatic trading that is going on. Understandable given that it's in the interest of insiders to stay out of the spotlight, and the public mostly only sees the HFT headlines.
It may not even have had to discern a new product announcement or launch. It may have just looked at twitter volume mentioning Tesla which doubled that day, and did sentiment analysis (much easier, which showed the communication that day was very positive), and purchased a bit of stock based on that. Who knows.
Maybe the analytics provider wants to focus on doing one thing and (hopefully) doing it well?
So if WSJ has a headline that X is in talks to buy Y, that's pretty much that. Also you could eliminate most of your risk by placing a limit order based upon the last trade prior to the news coming out. The market will either a) confirm that you are one of the first people to understand what has happened by filling your order at a price that doesn't reflect the news, locking in profits, or b) your limit order won't be filled because the price already is already above your limit, in which case you have assumed no risk.
[1] https://www.google.com/#q=site:wsj.com+%22in+talks+to+buy%22
Is there a timestamped archive of DowJones Newswire articles? I found it difficult to find archived news articles the last time I was looking for them.
Come to think about it, such a system, with an API, would be really nice to have. Something that allows anyone to create a trading bot and see how it performs. Any one what to make one? :)
1) Virtual money. So you can try out your bot and see how it'd do without risking money. (with $10k on say the Forex market which trades about 5 trillion per day, it doesn't matter if your trades are virtual)
2) Backlogging. So you can write a bot that'll analyse prices, volume, the orderbook etc of any moment in the past 6 months, which allows you to immediately simulate your bot's success in that market over a long period of time. (of course this is if you want to write a bot that analyses the exchange data, as opposed to things like news, for which you'll need to find your own database going back 6 months which isn't too hard btw, but a lot of data to process quickly).
I bet since writing this article the opportunities has pretty much slipped. Now anyone living close to the exchanges will do this.
[0]: http://www.washingtonpost.com/blogs/wonkblog/wp/2013/09/24/t...
Another issue is whether you know this is fact, but the public only knows it as rumor.
Of course, you'll want to consult a lawyer about all this.
Definitely no. Just watch the news and see that the stock price of any company changes after pretty much any news.
Also, if I understood the article correctly, this trade involved to transactions: first, purchasing the options, and then, purchasing actual stock using those options. And (once again, correct me if I'm wrong), the options themselves were purchased before the information was made public, so this transaction may have been the direct result of insider information.
No, according to the article, the options were purchased one second after the information first became public in some news wire headline. The fact that it was also 19 second before a journalist's tweet was apparently irrelevant, since there was that headline before the tweet.
Also, I wonder if it even has to be all that complicated as some kind of computer program monitors keywords. For all we know, it's just a guy sitting in his living room with with 2 Chrome windows opened, continuously refreshing the Dow Jones Newswire, with the trade all setup and ready to go in the other window.
Even with a clear understanding of the situation, the reaction from the market was mild compared to the swings of the global tides. When it worked, it worked well. But you had to have the wind in your back to get decent moves.
If I were in the industry with low cost access, and I were on the US market, I think things might have been different. But in my experience, you don't get rich searching for gold, you get rich selling the buckets and the shovels.
"“It would be impossible for me to do. By the time you could read the news, process it, and press the ‘buy everything’ button, it would take too long."
I'm picturing the context switch between reading a tweet or news article and some rocket ship of a program which gives you enormous power but takes some time to place a (possibly complex) order.
What if it's just turking with a hotkey? Watch feed, see headline, Ctrl-u.
This feels a lot more plausible to me than the idea that someone made sentiment analysis not suck.
If you can respond faster than everybody else, you don't need sentiment analysis. Just bet on volatility increasing (aka. buy options on both sides) whenever a news story appears about a company.
See every time a company announces earnings in line with expectations for example. "x company made 1 billion in profit" could fail to move the stock at all.
Betting that news is market moving requires a decent amount of analysis of the news and the credibility of its source.
A lot of people here are saying "I wouldn't bet $2.4M on that false positive rate", but that's not how traders think. You only have to be right more often than you're wrong (or alternatively, very right to offset being wrong a lot more often) - "betting" is exactly what they do for a living. It's pretty much the perfect application for statistical machine learning - users never see how bad your algorithms are, and so it doesn't matter if the quality is worse than a human as long as the speed is better.
Keyword analysis by itself is almost completely useless.
Also, an algorithm doesn't have to be perfect, it merely has to be right more often than it's wrong. So what if you get a few false positives and a couple of your trades blow up? That's why you're managing a portfolio and not dumping your entire assets into a single trade.
You don't need to be right every time in order to make a profit in the long run.
New WSJ Headline: Intel in Talks to Buy Altera
ALTR Stock: Trading within 30-day range
{ALTR 30-day stock graph}
March 2015 Call Options available
..and a live-updating, clickable GUI displaying, in a linear manner (i.e. low risk, small bet at one end; high risk, large bet at the other), the call option market depth at each strike price, annotated with the implied volatility and the % delta to the current stock price.
The trader reads the info, decides how much he wants to bet, then clicks the appropriate spot on the GUI to execute the trade.
More likely, the human prepares a list of scenarios of the form of "If a news story with these keywords appears, it will affect the valuation of this company in this way, perhaps plugging these numbers from the article into some model." News story hits the wire, the program reads it and applies some simple text mining, it compares the current market prices with the expected valuation, and it sends an order to the exchange to execute the trade.
Way back in 2006, I worked at a financial software startup where one of our products (the only one that made money, actually) did exactly this. The company is defunct now - it had a bit too little focus for a small shop of a dozen or so people, and its other products weren't all that succesful - but that was one of the things that actually worked.
If you're trading manually on time-sensitive information these days, you are an idiot. Back in 2006 it was something like 80% of trades on the market are automated bots; the only people who still put a human in the loop are the rubes.
I expect they'd be getting the newswire in real time via Bloomberg or Reuters, instead of off the Internet.
Is it possible to also bet directly on whether some piece of gossip is true or not?
Generally, there are two types of financial news events. First is scheduled, for example unemployment data - here, there are APIs to get the number and place trades, and as a human it is impossible to compete. Similarly, for FOMC statements, there is an API feed which provides objective answers to certain questions about the statement, e.g. "Did any Fed members vote for a rate increase?" Again, computers dominate. The other type of news events are surprises - unscheduled events that people are unprepared for. I'm certainly no NLP expert, but I do watch financial news feeds every day, and I can't imagine it being remotely easy to write a program to filter out the false positives. I've certainly watched a lot of market reactions to headlines, and I can tell that a lot of headlines that people would assume would be easy to write algorithms to trade on, produce market reactions that look far more "human" than the instantaneous reactions to unemployment data or embargoed Fed statements.
It doesn't matter that it is a bot. The controller of the bot happens to be, in this instance, directional order flow: unlike the overwhelming majority of market volume, he actually has an edge.
The options market maker is, unusually for market makers, forced to transact with any comer. They've got a license to print money and that is the price of the license. They get to pick their pricing.
In the old days, when people actually talked, parties would be cagey about whether they were buying or selling, to make sure the market maker didn't use that against them. The conversation went like this:
"Make me a market [quote a buy price and a sell price] for 100 contracts of May Foo CALLs at $15."
"0.10 by 0.15."
"Buy a hundred."
"Done."
"Make me a market."
"0.10 by 0.15."
"Buy a hundred."
"Done. #'(%# you man what's your angle."
"Make me a market."
"0.11 by 0.16"
"Buy a hundred."
"Done, you mother(#%)0&. You want WEIGHT? I hate carrying this. Next one is 0.20 by 0.25"
"Buy a hundred."
"DONE. You through with me yet?"
"Make me a market for 1,000 contracts."
"YOU MOTHER()#%0# #O%'#(&')$#$ ((#) '%)(#%). 0.30 x 0.35."
"Buying 1,000."
"DONE YOU #%)0#0%)."
And the reason the marketmaker just hated that interaction is because they now have what is politely referred to as "inventory risk." Market makers by definition don't want exposure to the market, they just want to harvest non-directional order flow, where buys and sells roughly cancel each other out, capturing the spread each time and making out like bandits. (In the case where the market maker's second conversation was someone selling the same option rather than buying, they just made $500 for about a minute total of work. A "seat" on the CBOE, which is the license part of the license-to-make-money, costs millions. The lucrative ability to collect a little toll from marker participants is why.)
But once in a while, you get stuck with inventory risk. And, if you're stuck with inventory risk by being short in grossly out-of-the-money call options close to expiry, the vast majority of the time that's great news! They expire worthless.
But they don't have to expire worthless. There's literally thousands of contracts which expire every month worthless, and (statistically speaking) a handful which go from being worth pennies in the morning of expiry day to being worth dollars in the afternoon.
You probably shouldn't feel badly for the options market maker here, except in the general sense that you should feel badly for people who underperform at their jobs. He's suppose to place orders that he'll be happy when the fills come in. (i.e. When his offer to sell a call option is matched with someone who inexplicably wants to buy them at a penny a contract.) He got the fills.
His problem is that a counterparty was smarter than he was and put in buy orders in response to late-breaking information before he could cancel his sell orders. Which is unfortunate, from his perspective, but avoiding this is literally his only job.
This is, incidentally, the whole premise of Flash Boys: people with lots of stock to shop are the guy getting sworn at in the above conversation with a marketmaker. They would prefer that marketmakers always buy 100k shares from them at a go without experiencing any price slippage. Market makers do not like this, but can't do much about it. HFTs are much better at making markets than humans are, and are capable of managing the risks of large block trades better. Some people who preferred getting an exceptionally good deal from human market makers would prefer dealing with them instead of algorithms which can actually, you know, do math many times a second for an entire trading day.
The other side of this can be pretty sweet. You get lots of flow from sales people, who are basically ripping off clients 3-7% on structures. If you have a variety of stocks you can also do some dispersion trading against the index.
One thing to note is you aren't betting on the options expiring worthless. It mostly pure vol spread, just dump everything in a book, hedge the deltas, win if it realises different than implied. Maybe back book some out-of-the-money options so you don't get killed on bleed (Taleb's book Dynamic Hedging explains this).
For those who don't speak the lingo of options, this means that he'd be hedging his exposure to directional stock price moves (by buying stock as he sold the calls) and also hedging his exposure to the size of stock price moves (volatility) by buying some cheap volatility, say at a different strike or for a different expiry, to cover the expensive volatility that he's just sold.
But like I said, I like the vignette :)
What would probably happen is tweaking their risk exposure: be less willing to play large resting bids/asks for large positions close to the NBBO, and instead either resize them or layer them over several price points.
My feeling with regards to the fairness of this is that one should not offer to sell 4,000 PUTS for e.g. Zynga at $1 in May 2015 unless one is prepared to accept delivery of 400k shares of Zynga in return for $400k. Some options market participant is signaling their willingness to do that, or something close to it. They are theoretically responsible adults who know the risks. If that trade blows up in their face, and they take to the Internet to decry the unfairness of it all, I will not be maximally sympathetic. (Full disclosure: I was probably their counterparty a time or two buying the option on "Zynga blows up in next short-increment-of-time" and did poorly.)
I'm sure this trader has programmers working on this problem and putting in some type of limiting such that he doesn't sell off 3000 contracts in one go for a very cheap option close to expiration. 3000 contracts is nothing in SPY options but for a single name stock (not an ETF) it can be considered a sizable trade compared to the open interest.
So the title of the whole article is totally misleading. It has nothing to do with Tweets & Twitter.
Maybe the same bot (script) has done dozens other trades (false positives) at a loss before.
But it is pathetic for a serious businessperson to feel entitled to their future business success. That's the kind of pride that comes before a fall.
Let's say that I want to experiment with some language processing and bet some dollars on parsing news like these. I have decent general programming skills, and have experimented with parsing stuff, but never touched trading.
Where do I even start?
It's been happening forever and of course it's now going to get automated. And the electronic market makers will use the same techniques to cancel orders so they don't get run over. The only people worse for wear will be the human market makers and the human point and click traders who can no longer compete.
That's who seems likely to be doing these trades to me.
http://www.newscientist.com/blogs/onepercent/2012/08/robot-t...
It is related to what people think that other people think is related to the stock value.
It is because of my experience with that hedge fund that I won't take quantitative investment work anymore. I came to regard it as unethical.
I would not be in any way surprised if it were my former clients that pulled this off. There were some smart people there.
Suppose instead that the hedge fund had lost money. It happens every day. Would you suddenly start to pity the unsophisticated hedge fund managers being exploited by all those other vicious investors?
However I do not wish to contribute to making their misfortune even worse than it would otherwise be.
The owner of the hedge fund had a model that was built on decades of CBOT commodities futures prices. His model would buy and sell a basket of about 1000 many times per day. There was some error in his model - at times he would lose money - but over the long run he was able to earn far, far more money than any other fund I've ever heard of.