Thread by @patio11

Since frontrunning is again being talked about by technologists, would like to explain the classical version of it, which is almost the opposite of the way that most technologists think it works:

Client Charlie tells Broker Bob to procure him a share of Google, currently trading at $100 a share.

Technologists think "Bob now races Charlie to the market." This is not the classical understanding of frontrunning.

No. Bob buys one share immediately.

Bob then waits "a little while" and sees what happens to Google.

If Google goes up to $100.15, Bob tells Charlie: "Good news, I got you a share at $100.15" and fulfills that share with the share Bob bought at $100.

If Google goes to $99.85, Bob tells Charlie: "Got it at $100."

Broker Bob thereby gets to play short-term speculation with his client's capital at risk and no material chance of loss. The underlying theft is not because Bob was faster, it was because Bob was willing to lie about Bob's intentions after the fact.

From this follows pretty automatically "Front-running is at 2019 at an all time low because computers are really good at keeping accurate paper trails."

Technologists think it is close to an all-time high because high-frequency computers make it easier to race clients / etc.

Note that this frontrunning also structurally requires an informational asymmetry (Charlie can't know what the true price of Google was with sufficient resolution to catch Bob on the lie), which is something which computers/open feeds/etc work to substantially mitigate.

When people propose e.g. "You need to let large positional traders somehow execute their trades without exposing this fact to other market participants, or they will be 'front run'", they are proposing substantially rolling back effective front-running protection mechanisms.