For anyone that doesn’t already know this, the world of trading is now split between two types of trader:

  • Humans; and
  • Robots (AKA the algorithmic traders AKA algos).

Robots…trade on new data at almost the speed of light. And “expect to reach those constraints in the next 18 months“. Back in 2010 when the guys at Chi-χ first said such a thing.

Humans, meanwhile, are still waiting for the news to come through on the email.

In the modern era, it seems that “reading things” is the new “postal service”. We’ve innovated ourselves into vestige.

Which seems like bad news. Here’s a history of the algo-high-frequency-traders:

Thanks this website
Thanks this website

My two favourite parts (from 2013):

“A server farm situated in Washington DC can transmit data at the speed of light to New Jersey via superfast microwave transmission service.”


“September 18th 2013: At 2pm, the Federal Reserve shocked the financial markets by announcing not to scale back its level of support to the economy. An estimated $600 million in assets changed hands in the milliseconds before other traders in Chicago could learn of the news.”

That, right there, is the real dystopian science-fiction. Because all that trading took place in the milliseconds before other robot traders in Chicago could learn of the news. Ne’er mind the old human codgers who were still seconds away from even realising that there was news. By order of magnitude, that’s millennia.

So to keep this sci-fi analogy going, from an “algo” perspective, all human trading takes place in very slow motion. And in the time between a human trader placing an order and confirming the trade, the algos have nimbly swept in, had a good sniff around, painted a moustache on the human trader’s face, and left wearing his pants.

In practice:

  1. A human trader places a bid-order (buy-order) at a certain price.
  2. The algos sweep through the market ahead of the price, buying up any ask-orders (sell-orders) that are priced below the bid-price.
  3. The algos then come back and sell to the human trader at his maximum bid price.

Theoretically, the algos take on no risk. They’ve basically stolen the bid-ask spread, all the while confident that they can buy and sell the shares before the buyer and seller realise that they’ve been dealing with a middleman.

But the algos haven’t stopped there. They now trade on all kinds of information, with pre-programmed trading strategies, and with an eye on much larger prizes than just the measly bid-ask spread.

The Good News

I think we’re beginning to realise that the algorithms may be fast, but they’re only as good as their coding. Coding that was designed by software programmers with human minds that (I expect) cannot comprehend the infinite multiplicity of feedback loops that you get when you have hundreds upon thousands of literally lightening-quick programs buying and selling on the back of different parameters in concert.

Sometimes, you get flash crashes that have come and gone before the human traders even realise that anything was happening. The only reason they know it happened is because they see signs of it in the day-trading graph.

And other times, you see the market break. Like this week past.

Here’s a screen cap from my new favourite article on the “August 24-26 market crash” that seems to have already ended (?):

From "Have The Robots Broken The Stock Market?"
From “Have The Robots Broken The Stock Market?”

The take-home message: the algorithms can get incredibly stupid.

In fact, they’re almost human.

For more, read this article from Matt Levine: “Algorithms Had Themselves a Treasury Flash Crash.” I LOL’ed and I won’t apologise for it.

Happy Friday!

Rolling Alpha posts opinions on finance, economics, and the corporate life in general. Follow me on Twitter @RollingAlpha, and on Facebook at