STA Blog

Arranged by STA Board Member Gerry Celaya, Elizabeth’s a perfect panellist for the second STA Technicals to Trading Systems Conference in London on the 16^th April 2024 (also available virtually). The theme this year is: ‘From Classic Charting to AI Developments’, which is where this speaker can certainly hold her own. A veteran of the industry, who remembers drawing charts by hand with a pencil on sheets of graph paper and who has, over the years, worked with increasingly fast and sophisticated hardware and software. She admits: ‘’it’s amazing how fast things have moved in the last thirty years.’’

Today, in her own words, ‘’We buy commodities to make things’’ at Mars, the confectionary giant where she has been leading a team for the last sixteen years. She kicks off by questioning what machine learning is, reminding us that what used to be called Deep Learning has existed since the 1950s. Sounding perhaps a little sceptical at the current enthusiasm for Artificial Intelligence, she stresses that it is ‘’a language model and is no good at math. Moving averages, technical analysis and econometrics count as a type of machine learning and are very old indeed.’’ Take a minute to ponder that.

I loved it when she points out that ‘’when trading live commodities (I’m assuming she doesn’t mean still on the hoof) one runs the risk of someone’s fat finger’’, or what’s likely to be a surprise to many, ‘’what do you do when you have a bad actor?’’ The point is: how does one differentiate between outliers, especially when computers work on the Bayesian principle where, if one hasn’t seen it before, it cannot be.

To the risks pointed out by Elizabeth, I’d also add issues like fake news, bots and the authority to edit and take down stories on social media. I’d also be very careful with data vendors; data is valuable, as we all now know, so why are some firms giving it out for free in chart packages?

For more info on the conference, click here.