AI and Stock Picking

There have been a few articles recently written about AI and the attack on hedge funds / funds management. Computers have been making in-roads for sometime, particularly in quantitative and algorithmic trading. My interest is in whether AI will start infiltrating classic, bottom-up fundamental analysis. 

One argument against the invasion of AI is the breadth of experience and information sources required: meeting company management, suppliers, competitors, ASX announcements, store visits - the list goes on. This argument seems reasonable. However, the opposite side of the coin is just how much any of this actually matters. It is an increasingly well-known fact that most fund managers fail to beat the index over time (particularly post fees). If you think about it, the index itself is a fairly basic algorithm (selects stocks based on metrics such as size and free float) that fends off most fund managers - most humans have already lost to machines. 

There are chunks of fundamental analysis that are manual and repetitive and therefore susceptible to AI performing the task. For example, a common hunting ground for the fundamental investor is to go through the ASX announcements, looking at company updates, insider buying/selling etc. I imagine a robot could trawl through these announcements and compile a shortlist and dossier of stocks meeting a checklist of investment criteria. A robot could identify a company buying back stock in decent volume, go back through years of announcements to compile annual and half-year data to identify the trend in financials. The robot could then probably search the web for products sales, compare with competitors and even check out the CEO's personal background. Google is the big obvious beneficiary of this. Just imagine the power of how much Google can understand about a company. Google can identify key word flow and website traffic and probably deduce the impact on companies. And all this can be delivered with zero emotion or bias. 

So not only do fund managers struggle at the best of times, it is hard to imagine how AI won't deliver a decent kick in the ribs in the coming years to mainstream managers. If there is an edge in fundamental analysis, AI could probably deliver it cheaper than humans. 

That being said, I think human managers will have roles in speciality areas. Start-up or young  turn-around companies are hard to quantitatively analyse (there isn't much data) and the quality of the management team are of higher importance - whether fund managers or a machine are best place to determine whether managers or winners or losers or not is another issue. Very illiquid, small companies may just be too small for AI to bother with. And there will be cases of joining the dots where AI may never catch up. 

If AI skims the cream from a lot of traditional finance professionals, they may be forced into the remaining niche areas such as venture capital and small-caps thereby overcrowding those trades and invalidating them anyway. 

Some individuals are particularly adept at buying in panic and selling in boom, however if AI increasingly displaces humans from markets, then irrational behaviour may also decrease reducing the ability to arbitrage fear and greed. 

It's really hard not to see how the landscape won't continue to change and get harder for humans. 

Kristian 

Disclosure: own Alphabet / Google (via 401k)