Constantly in search for an empirical basis for style and taste. Won’t find it without deep culture narratives. https://t.co/qJ70hQmlqI

— William Goetzmann (@WGoetzmann) May 27, 2017

The Financial Times thinks it is being terribly clever by citing an academic paper about machine learning to suggest the art market is going to be subsumed by trading algorithms. But the journalist who wrote the story suggesting auctions for lucky license plate numbers in Hong Kong is somehow analogous to the art market didn’t bother to check with the dean of art market indices, Yale’s William Goetzmann.

Goetzmann took to Twitter over the weekend to caution the puckish journalist that value in art is far less fixed—and more subject to waxing and waning tastes—than numerical superstitions. In his tweet, above, he warned that “deep cultur[al] narratives” are hard to quantify for the purposes of predicting future price movements:

Machine-learning techniques, such as software programs for deep recurrent neural networks, have already been used to analyse and predict other auction processes. One of my favourite examples is described in a February 2017 paper by Vinci Chow, a lecturer in economics at the Chinese University of Hong Kong: “Predicting Auction Price of Vehicle License Plate with Deep Recurrent Neural Network”.

Auctioneers should read this with a chill down their spine. Chinese societies place a high value on particular “lucky” numbers or symbols, and their governments take advantage of this in selling off licence-plate numbers. The neural network described in the paper learned the patterns in these auctions, and became quite accurate at predicting their outcomes.

All very subjective, of course. Just like art collectors who will prefer a Tang horse to a Tang camel, or a Basquiat painting incorporating a skull image to one without a skull.

Art market ripe for disruption by algorithms (Financial Times)