Ask a quant researcher how many factors have been "discovered" and they'll either laugh or cry. As of Harvey, Liu and Zhu's landmark 2016 paper, over 300 factors had been published claiming to predict returns. By now that number is past 400. The problem? Most of them are probably flukes dressed up in academic clothing. This is the factor zoo — and it's messier than it sounds.

The core issue is multiple testing. If you run 300 separate statistical tests at a five percent significance level, you'd expect fifteen false positives by pure chance alone — even if nothing real is going on. Harvey et al. argued the t-statistic hurdle for a new factor should be closer to 3.0, not the traditional 1.96. That single insight invalidates a significant chunk of the published factor literature, and it hits ASX-focused quants especially hard, given our smaller universe of stocks.

CONCEPTA t-stat above 3.0 is the modern minimum bar for believing a factor is real — anything lower deserves deep scepticism.
WARNINGBacktesting 20+ factor combinations on ASX data without correction almost guarantees you'll find something that looks great but means nothing.
KEY IDEAFewer factors, robustly tested, beat a zoo of marginal signals every time — especially in a smaller market like Australia.

Think of it like taste-testing 300 wines blindfolded. Statistically, you'll call a few "exceptional" just by accident. The factor zoo works the same way. Researchers — sometimes unconsciously — mine data until something sticks, then publish. The corrections for this, like the Benjamini-Hochberg procedure or Bonferroni adjustment, are rarely applied. On the ASX, with roughly 2,000 liquid names versus 8,000+ in the US, the data is thinner, so data-mining problems compound fast.

t=1.96 t=3.0 4.2 Value 3.5 Mom. 2.8 Quality 1.9 Zoo* t-statistic *Representative zoo factor — illustrative only

So what actually survives on ASX data? The honest answer is: fewer factors than you'd hope, but enough to work with. Value, momentum and quality clear the 3.0 hurdle in most rigorous out-of-sample studies when tested properly. The practical approach is to start with economically motivated factors — ones with a logical reason to persist — then apply conservative corrections before trusting any signal. For deeper reading on the methodology, multi-factor model fundamentals explain how factors interact in portfolio construction, while the multiple comparisons problem on Wikipedia gives the statistical backbone in plain language. Harvey's original argument is also well contextualised alongside factor investing principles that distinguish structural risk premia from statistical noise.

The factor zoo is real, but it's not a dead end — it's a filter. Most signals die in the wash; a handful don't, and those are worth understanding deeply.

This content is for educational purposes only and does not constitute financial product advice. Past performance is not indicative of future results. Profit Logic Ltd (ACN 688 669 936) accepts no responsibility for errors or omissions in this content or anywhere on this website. Always seek advice from a licensed financial adviser before making investment decisions.