Rogue Trader at UBS

An article on the New York Times’ Business Page this morning deals with one Kweku Adoboli who lost $2.3 billion in so-called rogue trades. The thrust of the article, by James B. Stewart, is to blame the UBS for failing to supervise Adoboli. All right. I managed to learn, but somewhere else, that a rogue trader is “Someone who makes trades that haven’t been authorized”; thanks for that Julia Felsenthal (link). My informant also adds that if the “rogue trader” happens to make profits, and especially very big profits, we usually do not hear about such unauthorized events. To qualify as a rogue trader de facto rather than merely de jure, you have to lose lots of money.

What Stewart’s article omitted, however, were the really interesting features of this story. Adoboli, who is 31, graduated from Nottingham University in 2003 with a degree in Computer Science and Management. He got his job at UBS in 2006, thus around age 26. His initial job at UBS was to assist in programmed trading; specifically he provided tech support for algorithms. Presumably his expertise lay in adapting trading algorithms to trading goals, these latter expressed in mathematical terms. Sometime in or before 2008 he became a trader, but I cannot find any press mention of the date of his transition from trainee to trader.

Now anyone who’s ever been involved with playing with algorithms knows the temptation of trying out this or that variant to see what happens. If a trend is going left and up, will it keep on going up more if you tweak this variable? Now in the usual test environment, no harm is done. But if you have to make a trade to test your brilliant insight, the results may be serious. That’s what seems to have happened here.

This story started looking much more complicated when I got this far. A young, relatively immature, and possibly already successful computer trader—computer emphasized because Adoboli obviously had no genuine trading experience or history—makes mistakes and then compounds them by trying to cover them up. He was the one who confessed—a bit late, to be sure. Add to that now that the very essence of such trading is speed, that computer trades are triggered by split-second changes in indexes—and that it would take equally smart computer algorithms to ride herd (thus “authorizing”) such trades—and now a much richer story is in front of us.

It’s lessons are that there is a problem in (1) appointing inexperienced people to engage in computer trading; how does the supervising senior trader know if a particular small change to an algorithm is good or bad? Has he been trained in computer science too? There is a problem (2) in computer trading as such. There is a problem (3) in taking someone out of college with a computer sci degree and setting him to work advising on algorithms. Programming is an experiential activity. It is a craft you learn by doing. It takes years. Algorithms, alack, alas, are weird, if abstract, but still structures. They have odd dynamisms. They are just like maddening reality. They very often produce the unexpected unless you know the terrain.

James Stewart’s column in the NYT is encouragingly labeled Common Sense. It would have been of some service to me to have been told the above, which is not, repeat not framed anywhere in press reports. Well, amateurism to the rescue.