Top tech quant aims for Wall St.

Self-learn­ing ro­bot leader left AQR to sell his al­go­rithms

Baltimore Sun Sunday - - MARYLAND - By Justina Lee

The man dubbed the world’s top quant has a mes­sage for Wall Street firms lav­ish­ing mil­lions on ma­chine-learn­ing pro­grams: Most of you are do­ing it wrong.

Ro­bots could save ac­tive money man­agers from doom but the tech­nol­ogy is strug­gling to live up to the dis­rup­tive prom­ise, ac­cord­ing to Marcos Lopez de Prado.

The Spa­niard spent years at the sto­ried names of smart-money fi­nance in his bid to shake up in­vest­ing with self-learn­ing ro­bots. Now he’s trav­el­ing across quant­land to de­liver a re­al­ity check — sell­ing his al­go­rithms and ex­per­tise to all com­ers af­ter an un­ex­pected split from AQR Cap­i­tal Man­age­ment this sum­mer.

“There is tremen­dous hype and very few peo­ple have a track record,” Lopez de Prado said in a phone in­ter­view. “It’s not help­ful.”

He should know. The ma­chine-learn­ing pioneer stands out in a field flooded with self-styled vi­sion­ar­ies. He wrote the book on ap­ply­ing the tech­nol­ogy to fi­nance and has boasted se­nior roles at both Guggen­heim Part­ners and Ci­tadel.

Last year AQR, the $185 bil­lion gi­ant of fac­tor in­vest­ing, tapped Lopez de Prado as its first head of ma­chine learn­ing. The matchup didn’t last; his de­par­ture was an­nounced af­ter less than a year at the firm. He speaks highly of AQR and calls the split am­i­ca­ble.

“My am­bi­tion has al­ways been to help mod­ern­ize fi­nance and of­fer dis­rup­tive so­lu­tions to in­vestors,” he said. “The best way to do it is with my own setup.”

With pro­ceeds from sell­ing patents to AQR, the Cor­nell Univer­sity pro­fes­sor based in New York now plans to hire a team of about 20 and to bring in out­side ex­perts de­pend­ing on each man­date. The firm has al­ready signed up new clients, in­clud­ing a sov­er­eign fund and sev­eral hedge funds, Lopez de Prado said.

Self-taught al­go­rithms are touted as the next fron­tier of the nerd revo­lu­tion, with the po­ten­tial to scan a broad set of data to di­vine the links be­tween mar­ket forces and se­cu­rity prices. This month alone Lynx As­set Man­age­ment started a new $140 mil­lion au­to­mated strat­egy, join­ing the likes of Man Group, Re­nais­sance Tech­nolo­gies and Two Sigma in us­ing ro­bots de­signed to im­prove over time with­out ex­plicit hu­man in­ter­ven­tion.

The prob­lem is, com­puter-pow­ered strate­gies are strug­gling to live up to the hype, with a Eureka­hedge in­dex of AI hedge funds lag­ging peers in re­cent years. That spells op­por­tu­nity for the likes of Lopez de Prado with his out­fit True Pos­i­tive Tech­nolo­gies — a dig at the er­ro­neous con­clu­sions de­rived from data that are ram­pant among quants, or false pos­i­tives.

His di­ag­no­sis: Fund man­agers are rou­tinely throw­ing data at a ro­bot with­out form­ing a the­ory. If a back­test sug­gests in­vestors should snap up stocks on a given day of ev­ery month and sell them a num­ber of days later, only a joined-up ra­tio­nale for the trade will work in live mar­kets.

In the other ex­treme, econ­o­mist-led quants have a pen­chant to use ma­chines to con­firm pre­ex­ist­ing ideas, he reck­ons.

One way to fix all this is to use the tech­nol­ogy to de­velop core propo­si­tions — the sci­en­tific way. “With­out this the­ory-ML in­ter­play, in­vestors are plac­ing their trust on ei­ther toy mod­els or high-tech horoscopes,” Lopez de Prado wrote in a re­cent paper.

Hail­ing from Gali­cia, an au­ton­o­mous re­gion in north­west Spain, Lopez de Prado earned two doc­tor­ate de­grees from the Univer­si­dad Com­plutense de Madrid, be­fore re­search gigs at Har­vard Univer­sity and Cor­nell.

With two decades of ex­pe­ri­ence chan­nel­ing the power of ma­chine-learn­ing at the in­ter­sec­tion of math and fi­nance, he’s now the go-to guy in this cor­ner of the sys­tem­atic world.

The Journal of Port­fo­lio Man­age­ment rates him Quant of the Year for 2019. He’s the best-read eco­nomics author on re­search net­work SSRN.

Lopez de Prado paints a pic­ture of an AI-pow­ered fu­ture un­friendly to the self­s­tart­ing stock picker. One day, each em­ployee of a fund will be in charge of one part of a re­search process akin to assem­bly lines in a so-called strat­egy fac­tory, ac­cord­ing to the for­mer Berkeley Lab re­searcher.

“It’s not nat­u­ral in fi­nance, right? The his­tory of fi­nance has been that dis­cre­tionary port­fo­lio man­agers used to run the place,” he said. “And these firms fit­ted quants as if they were dis­cre­tionary PMs.”

Ma­chine learn­ing re­mains an ex­pen­sive and fledg­ling in­vest­ment yet to de­liver the kind of gains Lopez de Prado dreams of. But in this view, it’s the nat­u­ral evo­lu­tion for an in­dus­try ev­er­more re­liant on com­put­ers and less on hu­man in­stinct.

His is also an eth­i­cal mis­sion: Smart pro­grams, if de­ployed ju­di­ciously, could one day save Mom and Pop from al­lo­cat­ing to flag­ging strate­gies, spurred on by pseu­do­science and self-in­ter­ested fi­nanciers.

“Even­tu­ally fi­nance should be some­thing rel­a­tively boring, just like go­ing to the doc­tor. You go to the doc­tor, and there’s a pro­to­col for ad­dress­ing a prob­lem,” he said. “That is my hope: even­tu­ally we make fi­nance more sci­en­tific and as a re­sult it be­comes less of a casino and more of a util­ity.”


Marcos Lopez de Prado has al­ready signed on new clients to his forth­com­ing firm, in­clud­ing sev­eral hedge funds.


IBM’s Wat­son com­puter demon­strated the po­ten­tial for AI, win­ning on “Jeop­ardy” and di­ag­nos­ing a dis­ease. How­ever, com­puter-pow­ered strate­gies haven’t lived up to the hype.

Newspapers in English

Newspapers from USA

© PressReader. All rights reserved.