Getting AI To Fall In Line

Barron's had a short profile on the recently listed iShares Systematic Alternatives Active ETF (IALT) which is similar compare to the longer tenured Blackrock Global Equity Market Neutral Fund (BDMIX), same managers running a slightly different strategy.

IALT is running three strategies, one is market neutral, the second is a "dynamic" macro strategy and the third is long only. The three will be combined such that the volatility will be in the 7-9% range. 

You can see the comparison. The idea with Portfolio 4 is that HFGM leverages up so I wanted to get a little closer to apples to apples with IALT. Portfolio 5 is a build it yourself version (maybe) but the volatility is way too high at 14% unless...

...you go further back, subbing in QGMIX for the newer HFGM. Portfolio 4 tries to tamp down the  volatility to be closer to BDMIX (cousin of IALT).

The performance of BDMIX is suspicious, trading flat and then turning up so markedly. Copilot said it was not a change in strategy, the fund started to do better because the various factors and markets it tracks began to do better so it is a function of patience paying off. 

The blend of momentum, macro and market neutral from Portfolios 2 and 3 in the longer backtest looks pretty good over the longer period but it differentiates from VBAIX by quite a bit. The extent to which that is a good thing versus a challenging thing is a toss up. In 2019 VBAIX outperformed by 850 basis points while in 2021 it outperformed by 700 basis points. That gets made up for and then some along the way but a point we've made before is that differentiation can be difficult emotionally which is worth remembering before building something that gets too far away from plain vanilla. 

Copilot likes the blend of those three funds, it notes the biggest risk would be some sort of event that simultaneously hurts risk assets like SPMO, "cross asset macro signals" impacting QGMIX and what it calls deal flow for ARBIX but when I said that ARBIX is more of a relative value strategy, Copilot softened up on that point. 

It said the biggest thing lacking from the portfolio was convexity which to me sounds like client personal holding BTAL. It didn't love BTAL, ranking it third best after managed futures and funds like SWAN which is mostly treasuries with long call options. SWAN should go down less than equities but get a kicker from the calls when equities rise. BTAL helps with SPMO but wouldn't protect against things going wrong with QGMIX and/or ARBIX.

Copilot then laid out ten macro strategies to stress test the portfolio but I am not confident it did it correctly. It included examples from 2017, 2018/19 and 2022 that could be bad for the portfolio but it did well in all of them. I pushed back on that and it rationalized that it was more of what could have gone wrong but didn't. 

This post has a couple of instances where AI gave an output that didn't quite seem to be correct. I've been asked questions about how to use AI and part of my opinion is that it is important to question outputs. It will go back and rethink answers which makes for more productive work. 

SPMO, QGMIX and ARBIX are all in my ownership universe. 

The information, analysis and opinions expressed herein reflect our judgment and opinions as of the date of writing and are subject to change at any time without notice. They are not intended to constitute legal, tax, securities or investment advice or a recommended course of action in any given situation.