How High Frequency Trading Works, Trading Speed, and the Flash Crash

Speed in the world of trading is everything.  In the beginning, floor traders would be huddled together shouting orders inside the major exchanges.  On average, these orders would take about 11 seconds to execute a trade.  Nowadays, 50 to 70% are now executed by an algorithm at a speed rate and scale beyond our comprehension. Do yourself a favor and take 20 minutes to watch the video above. Then stumble into our trading courses page to arm yourself with strategies that work for regular, retail trader like yourself.

At first, you’ll visit a small New Jersey trading firm.  Trading time, at the most precise level, is often measured in microseconds (millionths of a second).  By finding the number of microseconds since midnight of the previous day, you have a highly precise scale to occur when price movements occur.  Using this scale, a high frequency trader might do 1000 trades each minute in quick bursts.  Depending on the high frequency trading algorithm, a number of buy and sell orders will be blasted out to see if there’s any response.  When another computer bites on one, the original computer will cancel the ones that didn’t “stick.”  A perfect example of algorithmic trading occurred recently with Kraft (KRFT).  An algorithm bought a bunch of Kraft and effectively jammed the other robots.  The algorithm was then able to sell at much higher prices to other algorithms, netting a gain of over a half million dollars in a matter of seconds.

“There’s always an advantage to getting information faster than the other guy.”  Take the case of news events.  If you’re aware of positive news regarding a company that hasn’t yet hit the wire, you can make a lot of money on the buy low, sell high principle.  With electronic markets, this “get there first” idea created a race of who can get their orders to the exchange first.  Yes, your physical distance to the exchange matters when you’re trading robotically.  The speed of light is about a foot per nanosecond (one billionth of a second).  If you can get in and out of trades very fast, millisecond after millisecond, you can game the system in your favor.  Serious algorithmic traders began buying up real estate next to the exchange for the quickest connection possible.  Eventually, regulators stepped in and allowed servers (for executing trades) to be hosted inside the mothership (exchange).  To make it fair, every single computer in the NYSE server room in the NYSE has EXACTLY the same network cable length.  Consider the amount of vertically stacked computers in this room that is the size of three football fields.  This ensures competitive fairness to all the computers in the room placing algorithmic trades.  These computers are owned or rented by banks, hedge funds, brokers, and financial institutions.

Is the race for the fastest connection over at this point?  No.  There are 13 regulated exchanges and roughly 50 non-public dark pools.  You can imagine the amount of complexity created by those who need to use trading data from one exchange for strategy purposes in another exchange.  For example, say Chicago is known for its commodities (basic goods like oil, soy beans, corn, etc.). Let’s say the price of oil goes up in Chicago.  You can pretty much bet a company like Exxon will also go up in New York, although the time between the two is not instantaneous.  This actually created a race to lay the straightest fiber line from Chicago to New York.  Some reports say a company spent eight figures to lay a line with a 13ms connection through a mountain.  Right now, a product is underway to create a series of towers to beam signals between the two exchanges, as light travels faster through air than fiber optics cable.  For the big high frequency trading players, this arms race is a huge drain on resources.  If you don’t adopt a new connection technology and your competitors do, you’ll miss out.  This is actually great for the little guy traders like us – trading has become cheaper.  In 1992, it would have been $100 to trade 1000 shares.  Now it will cost around 10 bucks.

On May 6, 2010 at 2:42 in the afternoon, the flash crash occurred.  At 2:25 p.m., an emergency “circuit breaker” stopped the market after it fell 1000 points.  In two and a half minutes, several hundred billion dollars vanished.  When the market came back on five seconds later, the markets bounced right back up.  Two and a half minutes later, the market was 600 points higher.  Still, 2.5 years later, we still don’t know exactly what happened.  The official U.S. regulators investigated and said it was the result of one trader in New York’s overly aggressive algorithm going haywire.  This claim is still disputed among experts.  The scary part is that hundreds of billions of dollars can be lost in the blink of an eye, and years later, we still don’t know exactly what happened.  Within the last decade, we’ve entered an era where complexity, speed and volume is no longer human readable – we don’t know until it’s too late.

What do you think about speed, high frequency trading, and the accountability of our technology?
Check out more of our videos here.

1 Comment

  • Babak says:

    Very Interesting video, Makes you to be more appreciative having the stop loss in place. I do remember the day, I was at work, just closed my position on the day trading and I was riding with a co worker to get lunch, and I heard that on the radio!

Leave a Reply

All fields are required. Your email address will not be published.