Much has been made of the hacking of the AP Twitter account and the subsequent tweet that sent the markets in the tail spin for a brief period of time.
There has been criticism of algorithms and traders that used this tweet (and generally of social data) without apparently realising it was a fake or that there weren’t ‘circuit breakers in place’ or without having more risk procedures.
But let’s consider a few things:
1) The tweet came from the Associated Press. The Twitter account is a verified account, it is a respected company, publishes good content and has a large number of respectable followers. No matter how your algorithm assess social value, this account scores highly – it should be listened to.
2) Hacks are rare – for all the talk, the assumption has to be that the content you are receiving from a respected source is valid because 99.999% of the time this is the case. This is another reason to listen to this source.
3) How do you know it’s not real – someone has to report it first and why not AP? After a week of not too dissimilar events, one does have to at least take it seriously.
4) Social data is only one input of many different inputs to an algorithm. It is difficult to imagine any algorithms only using social data. So the drop itself and the aggressiveness is more related to the nature of high frequency trading taking advantage of price change momentum, even if the initial momentum was caused by the tweet.
5) The algorithms worked well – the algorithms that used social and price momentum should have done pretty well out of this. News came out, markets went down, momentum increased, price went down further, turned around and headed back up again rapidly when the sentiment and related message volume changed again. It was those algorithms or humans who were too slow to realise what was happening that would have got burned.
So the problem wasn’t the account and it wasn’t even the tweet, it was the hack. Focus on how Twitter’s authentication process, but marvel at the technical superiority of algorithms using social data.