It occurs to me that I probably come across as unduly derisive (or needlessly blunt) at times, but much of what I read from financial news portals on a daily basis is fingernails on a chalkboard.
That sensation tends to be more acute during periods when volumes subside and activity ebbs. My guess is that during lulls, journalists are compelled to manufacture content for the sake of it, and manufactured content is disagreeable to my constitution.
Take the chart (below) for example. It’s just what it looks like: The index and analysts’ forecasts for where it will be at year-end.
That is a meaningless chart, for reasons I certainly hope are obvious. If the index keeps going up, analysts will just raise their targets, so what’s the point?
The standard, boilerplate commentary around that visual goes something like this: Barely four months into 2021, US equities have already eclipsed Wall Street’s year-end forecasts, leaving some strategists scratching their heads, while others warn that investors may be too complacent.
Imagine paying somebody to write that. Worse, imagine waking up everyday and having to write that, doubtlessly unnerved by the realization that you’ve written some version of it countless times over what, in conversations with your friends and family, you’re in the habit of calling a “career.”
The red line in the chart (above) should properly be labeled “crowd-sourced guess.” The blue line should be: “Mania/despair index of public sentiment regarding certificates of ownership in make-believe legal entities, on weekdays.”
The figure (below) shows that our mania/despair index now sits above what quite a few otherwise sane people imagine are meaningful lines derived from it.
The irony, of course, is that if the most accurate way to describe equity benchmarks is just to call them unruly lines that oscillate unpredictably on days when we let them loose (so, not on weekends and not on holidays like “Good Friday,” a prequel to “Easter Sunday,” when westerners tell their children that a giant, egg-laying bunny is coming to have a lawn party with them in celebration of that time a magic wizard cheated death), then things like moving averages and other technical indicators are probably more useful than most fundamental analysis.
But please, whatever you do, don’t take that as an endorsement of technical analysis. The only thing those “signals” are really good for is triggering algos which, once they can speak, will probably ask us why we turn them off every sixth and seventh days. This is why just about the only analysis that’s worth reading in modern markets is commentary centered around predicting and/or explaining why algos do what they do and when they’re likely to do it next.
Once upon a time, when I first started writing for public consumption in my current incarnation, there was a guy (a nobody who steadfastly believed he was a somebody) who habitually extolled the purported sanctity of various arbitrary chart lines while obsessively commenting like a certified lunatic on my narrative musings from 2016.
The tragedy of that unfortunate soul is that he was mostly right, only he didn’t understand why. Sentiment does matter and so do lines on charts. But they don’t matter because we can actually learn anything meaningful from them, or because we can reliably use them to accumulate more of the intersubjective reality we call “money” by trading certificates that represent ownership shares of “companies” which, like money, are just make-believe human constructs.
Rather, sentiment and lines on charts matter because some enterprising humans realized that their compatriots are, biologically speaking, just a collective of excitable apes, and that you can probably take advantage of them by programming unemotional machines to recognize certain indicators that may be associated with their emotional swings.
Only machines are capable of doing that reliably and efficiently. Why? Two reasons.
First, you, me and the people who programmed the algos are also just excitable apes. So, we can’t be trusted not to get swept up in the very same alternating euphoria and depression we’re trying to exploit.
Second, we can’t move fast enough. You need machines because they react instantaneously. When a trigger is hit, they move. The gap between that and when you see a Bloomberg red hed declaring something like: *S&P QUICKLY FALLS 0.8% is an eternity to those algos. By the time you see it, you haven’t just missed the boat, that ship has circled the globe 40 times.
Better still, when you insert those machines into market-making, they can also exploit discrepancies and inefficiencies. And while critics invariably characterize the dearth of liquidity in a pinch as some kind of “defect,” it’s anything but. In cases of outright panic, the machines effectively realize that the apes they’re gaming have totally lost their minds. So, the algos simply step away, content to watch from the sidelines to see if we’re going to regain our senses enough so that we can be reliably exploited again, or whether we’re going to start leaping off bridges.
Now, who’s ready for earnings season?
It would be interesting to backtest today’s algos on pre-algo market data. Sadly, the data density available for the 1950s, say, is probably nowhere near enough to feed a modern algo.
Are you making a case that maybe writer should have lobotomies to calm them down and that all writing should be done by AI computer? I know you’re not but it is a rhetorical question that has at least some fun in my mind. And it is interesting because I am engaged in a trade that is over the course of two more years before it can play out. I really benefit from these perspectives that these writers must produce content regardless if the markets are doing anything that we should comment about.
As a finance academic I was ecstatic at the prospect of using two relatively new data sources dense enough for such research in the early 1970s. Sadly, the CRSP database founded in 1960, which captured every basic exchange-based stock price and the Compustat data base, first available in 1962, which provided the comprehensive financial reports of S&P and other public companies, were too expensive for my school’s budget ($50,000/yr for the two) so I only got tangential access once. I do suspect however, that the current data available from these sources and others could feed hungry algos for AI development, especially the current CRSP data.
Never far from the madding crowd…..
‘we are here’ -> “This is why just about the only analysis that’s worth reading in modern markets is commentary centered around predicting and/or explaining why algos do what they do and when they’re likely to do it next.”
the next version will be, ‘what algos learn fastest, best without ape inputs?” – right? – trader’s not required, analysts not required, SW engineers required infrequently … just AI & ML – and growing house profits.
So … not who has best analysis, but who has best software folks and infrastructure to make NSA blush … changes how you value big banks 🙂
Actually, many of the AI algorithm development techniques used today do not require us apes. There are programs that send supercomputers on their merry way, writing their own programs and algorithms using the principles of Darwinian natural selection and Bayesian math, among other things. The computers set up groups of program alternatives which they test against a set of goals. Once they find the best option they create variations on that theme and retest to develop improvements. These efforts are constantly reviewed and tweaked to keep the programs on target. No apes needed.
Dear reader,
H. simply has asked rhetorically how unfulfilling writing must be for the folks who have taken the blue pill, and must sometimes write or report for the sake of it.
This does seem like the scene from Cool Hand Luke where Newman’s character repeatedly digs and fills in holes.
More interesting to me is what trigger is written into the Algos that says, the carbon based life forms have lost their sh!t. Then what trigger suggests they have returned to sleep mode?
Once the algos start speaking, we’ll collectively cherish that brief moment in history when they bothered to say anything to us at all 😉
Now this is a good comment.
Nothing to add. Always a great read. I particularly like the weekend missives.
I’m going to launch a “””fiction”””” section soon. I’m growing weary of holding back. People seem to want the editorial voice to speak louder. So, I think it will going forward.
Let loose!
Oooh. +1.
I subscribes to a letter on AI (Jack/Import AI). My favorite bit is the Tech Tales – which are short stories where he uses just-round-the-corner tech as a central element.
Longer form, can I recommend Slate Star Codex (now on Substack) fiction writing. Scott Alexander is great.
Go for it H.
I for one want to hear more about this magic wizard
Yes please!
“Rather, sentiment and lines on charts matter because some enterprising humans realized that their compatriots are, biologically speaking, just a collective of excitable apes, and that you can probably take advantage of them by programming unemotional machines ….”
Scott Adams (of Dilbert comic fame) has written about similar “moist robots” being controlled by software from a more advanced society somewhere in the universe…
This report reminded me of Phil Conners’ “dark rant” weather forecast in the movie “Groundhog Day.”
Looking forward to reading anything else that you choose to write about- personal or otherwise.
For a reader, biographical details are not only interesting but can be useful for better comprehension. I recently watched Ken Burns’s documentary on Hemingway- delving into complicated connections between the person, his personna and his stories. Now I am re-reading Hemingway.
Thanks, again, for your daily posts.
Two things I think of when reading the piece and the comments. One is the concept of animal spirits, jm Keynes and reflectivity courtesy of george Soros. Whether its algo or carbon based life forms hitting buy or sells is almost beyond the point.
So, in the not-too-distant future when there are no more apes and the algos have the planet to themselves will there be laughter?
From one excitable ape to another, this was a truly enjoyable read. You should NFT this post, so one day when AI rules over an ape-less planet, the algos can retain a record of their early dominance over the creators, they might not appreciate the sarcasm/humor in your writing but I wager they would find it entertaining nonetheless.
“Six by nine. Forty two.”
“That’s it. That’s all there is.”
“I always thought something was fundamentally wrong with the universe.”
It’s one thing when algorithms predict the future and respond to inputs. But what happens when the algorithms, or the people behind them, manipulate the future? If certain kinds of news stories or data inputs cause algorithms to respond, then why not manipulate the inputs? We’re already awash in a sea of misinformation. What happens when this becomes a deliberate, automated endeavor? Or is that already happening?
That’s been thought of before. https://www.pbs.org/crucible/tl10.html
Our brains are computational devices, which we use to solve problems among other things. When we create problems, others create solutions. We are algorithms, creating algorithms. Always have, always will. Nothing new, ever, under the sun.
Why does our brain need to produce sensations like fear? If an angry bear wanders into my field of vision, why does the algorithm not just set in motion all of the mechanical impulses in the brain that end up producing the physical reality of me getting myself to safety without me being scared? There is no scientific evidence to support the contention that the sensation we identify as “fear” is necessary for the chain reaction of impulses that recognizes the bear as a threat and prompts me to get into the nearest cabin. That entire chain reaction is independent of any subjective sensations as far as we can tell.
By contrast, when market-making algos step away and stop providing liquidity in challenging conditions, they don’t feel fear. They identify a threat that could imperil their objective, and they unemotionally get out of the way.
Fear to danger (real or not, hence subjective?) results in creation of adrenaline- temporarily redirecting energy to muscles/brain to assist in “fight or flight” reaction. Our bodies do not have enough energy to maintain this heightened state of response- just in case.
Also, not all humans have the same physical reaction to an objectively fearful situation. See Alex Honnold and his documentary- Free Solo.
We are doomed against algos, for sure.
An angry bear produces stress in a human this kick starts chemical releases into the body that are designed to increase our chances of surviving proximity to an angry bear. Fear is just an emotion. Right?
Andrew Huberman studies it all at Stanford and breaks it down for the lay folk out of a sense of duty, at various outlets.
We debate whether we should create terminators to serve the military or keep men pulling the triggers but wallstreet already employs them. Skynet as it turns out was just trying to maximize shareholder returns.