Malt Liquidity 33
A shallow look at the Deep (Blue, not F.V.)
Deep Red
My first knowledge of IBM came from Deep Blue, of course, the infamous chess playing computer which bested Garry Kasparov in a short match (who promptly accused IBM of cheating.) Though my exposure to the stock was essentially predating the “ok boomer” meme (IBM, B stands for boomer), I also became aware of Watson, the Q&A-I bot that bested Ken Jennings on Jeopardy to much fanfare. As I encountered more and more people in the AI/ML software space, I rarely heard Watson mentioned as anything other than vaporware, the Industrial Revolutionist’s idea of an all-knowing Multivac. Presumptive ageist bias aside, a decade later, for the life of me, I still do not know today what Watson does beyond buzzwords. So, of course, it is now a potential SPAC target:
IBM is studying alternatives for [Watson Health] that could include a sale to a private-equity firm or industry player or a merger with a blank-check company, the people said. The unit, which employs artificial intelligence to help hospitals, insurers and drugmakers manage their data, has roughly $1 billion in annual revenue and isn’t currently profitable, the people said.
I don’t mean to make an esoteric point here, but Watson’s failures seem like a classic case of chucking refine-through-error iterative statistics and gobsmacking amounts of data at a complex problem without rigid inputs or a recognizable output.
Kohn says he’s been waiting to see peer-reviewed papers in the medical journals demonstrating that AI can improve patient outcomes and save health systems money. “To date there’s very little in the way of such publications,” he says, “and none of consequence for Watson.”
The crux of the Watson conundrum does not seem to be that it’s bad at NLP or is “mashit learning” (as I call my favorite “import scikitlearn shitty tradebot” papers) but that the value add - the output - wasn’t coherently assessable. Diagnosis does not seem to be a major value add point - if the output of Watson is a better first diagnosis, well, don’t we have follow-up visits? Diagnosis isn’t a binary option, an all or nothing proposal. Don’t doctors have chances to fix incorrect calls? When examining the inner plumbing of the client visits to the insurance billing to drug R&D to every middleman in between responsible for taking a drug just to phase 3 alone (the make-or-break moment), it seems that the intake level diagnosis just doesn’t make a real impact on the cost inefficiencies of the system. Plus, it’s not like they were doing much better of a job anyway.
Watson learned fairly quickly how to scan articles about clinical studies and determine the basic outcomes. But it proved impossible to teach Watson to read the articles the way a doctor would. “The information that physicians extract from an article, that they use to change their care, may not be the major point of the study,” Kris says. Watson’s thinking is based on statistics, so all it can do is gather statistics about main outcomes, explains Kris. “But doctors don’t work that way.”
Why the waxing poetic about Watson? Well, I see a lot of similar errors consistently made in analysis of market data, in that there is a lot done without seeing as to whether it is profitable. A very common practice I see is people downloading a bunch of market data, running some statistics, and immediately back-testing strategies without a real reason as to why that strategy could work, while using flawed methods of assessing a backtest such as Sharpe ratios. Remember - stock market data is incredibly noisy, yet, concurrently, all this data could also contain signal, so you can’t just throw it out. Rather, I recommend taking a green-to-tee approach (to borrow a Tiger euphemism) for backtesting strategies:
Exploratory data analysis is fine, but don’t just number-crunch to try and “find edge”. Correlations exist everywhere in finance - every product is tied together to some degree - so raw numbers won’t provide insight without contextualization as to why that correlation is significant
Once you have found potential edge, find the optimal case for your trade - “where does my edge work best?” This is the “green-to-tee” philosophy, as from there, you find the optimal conditions, find what breaks as you iterate out of the optimal scenario, and slowly widen the thresholds. Start with the simplest version of your idea - the tap-in putt - and then add all the nonsense on top, as every step of math you add increases the potential noise exponentially.
This is why pure black boxes tend to backtest much better than they perform live. When you don’t know what the output is supposed to look like - is Sharpe supposed to be optimized? Outperforming the S&P? High expectancy, low trade frequency? - how could you possibly account for overfitting and prevent it when backtesting? Trying to brute force returns through processing terabytes of data simply cannot cut through that much noise, especially given how reinforced learning works - the noise iteratively gets worse and worse and your trading signals become more and more overfit.
These are just high level guidelines, and to paraphrase the legendary Supa Hot Fire,
I’m not a programmer.
EquiTatis in San Diego
A favorite financialization hypothetical of mine is the re-imagining of consumer debt as consumer equity. In practice, there is the coding bootcamp model where the coursework is free for a cut of your earnings if you are hired within a year of finishing, but why not extend this to student loan debt, and make it student equity instead? (Much like ZIRP, subsidized uncollateralized loans under market rate that aren’t disbursable through bankruptcy cause serious risk-transferal externalities, but that’s a rant for another time). So when I saw this happen with Fernando Tatis Jr’s new contract, I got excited immediately:
Tatís signed a contract with Big League Advance, an unusual investment fund that pays minor-league players money up front in exchange for a share of their future MLB earnings…
Schwimer says the company has invested more than $150 million in nearly 350 players so far, none more prominent than Tatís. Many of them never played in the major leagues at all. Schwimer declined to reveal how much the fund gave to Tatís as a minor-league player or how large of a stake Big League Advance has in his major-league earnings.
However, he said Big League Advance’s average pact with players is for around 8%. In the case of Tatís, that would add up to $27.2 million. Representatives for Tatís declined to comment.
YES! IF! YOU! CAN! TAKE! LOANS! WHY! CAN’T! YOU! SELL! EQUITY!!
I love this value proposition so much, because it completely throws out the rigorous bucketing that is creditworthiness and allows individuals to bet on themselves in more unique ways given a backer who believes in them. And isn’t it comparatively nicer being able to cheer people on rather than botting shoes, Supreme gear, Pokémon cards, or NFT art in the hope that they go up in value? The model obviously isn’t new. Low % chance to hit, high risk, high payouts is the Venture Capital model. I much prefer this offshoot - perhaps we call investing in humans the “Vitality Capital” model, and one day, I shall start my own Human Fund. Money for people, for a stake in their success. Just don’t ask about the offsetting short delta hedging mechanism.
On that note…
Something tells me I'm gonna have a lot to say about Bitcoin ETFs going forward
Facebook continues to be the worst tech company in existence
