from https://www.livewiremarkets.com/wires/is-eps-the-best-factor-for-measuring-investment-successI wasn’t surprised when last Sunday I read about a quant team in UBS spruiking the forecast outperformance of its machine learning model.
Here’s the story in the AFR:
"Five years on, and equipped with more data and more powerful computers, investment bank UBS’ quantitative research team reckon they’ve finally cracked it. They’ve built a model that shows the machines can finally out-forecast the humans, including in Australia. The model comes up with 12-month forecasts and the best bit, according to the quant analysts, is that it is completely unbiased. Its numbers are typically lower than consensus.""The regression model is always trying to pick a company’s earnings over the coming 12 months. It uses prior year earnings as an anchor starting point (about 75 per cent of the end value), and uses a range of traditional quantitative signals and metrics (cash flow growth, for example) and macro factors. It’s all written in Python computer code."The model is smart enough to work out how the company’s earnings have responded to macro factors like bond yields, currency and oil pricing, consider current levels and use that to formulate the earnings forecast. It works on a company level but can also be aggregated up to a sector or market level..."
Source: Australian Financial Review
Yet, if earnings as measures by EPS don’t really matter, neither do accurate forecasts — whether human or AI-generated.
In investing, picking your battles matters just as much as how well you wage them. Battling others to more accurate EPS estimates may be a battle not worth starting.
