Automated trading

Where algorithmic trading names the logic, automated trading names the hands off execution of it. The range is wide. A trailing stop managed by the platform is automation of a single instruction. A strategy left running unattended automates the whole sequence, from the condition that triggers an entry to the size it is sent in and the exit that closes it.

What separates it from manual trading is that the failure modes become operational rather than analytical. The software has to be running, connected and receiving prices, which is why unattended strategies are commonly hosted on servers rather than on a laptop. A dropped connection, a stale quotation, a duplicated instruction or a market condition the rule was never tested against will all produce behaviour the author did not write. And every cost that was an assumption in a historical test, spread, requotes and slippage, becomes a measured quantity in a live market, which is the usual source of the gap between a tested result and a live one.

Two cautions attach. Practitioners disagree about whether automation removes trading psychology or relocates it, since a person still decides when a rule is switched off and that decision usually arrives after a run of losses. And systems sold with published historical returns are a recurring subject of regulatory warnings in many jurisdictions, for a simple structural reason: a curve fitted to history costs nothing to produce and cannot be distinguished, from the outside, from one that was not.