Friends,
In this issue:
- streetlight effects
- Investment Beginnings Class 5 was a delight
- thoughts on the current market cycle and options
I’ll be short up here with one rec since the investing sections are a little longer. Just have extra exhaust since last week’s graduation issue skipped finance stuff.
everything is a nail, or at least it ought to be | 5 min read
This is a short book review by Dan Davies of The Irrational Decision by statistician Ben Recht.
Recht argues that if a hammer is genuinely your best tool, it’s actually smart to look for nail-shaped problems or reshape problems them so they’re more hammer-friendly.
In Recht’s case, the hammer is mathematical optimization (especially linear optimization). The book traces how, over the past century, people haven’t just used optimization to solve problems, they’ve restructured problems to fit optimization’s strengths. He gives the example of the feedback loop between chip design and optimization algorithms where better chips enable better optimization, which enables better chip design, repeat.
If you recall, in my post slashing away parts of their humanity, Davies plugs C.Thi Nguyen’s book The Score (which I bought as I enjoyed Nguyen’s first book about games). Davies’ latches on to the manner in which metrics ultimately create “streetlight” effects by optimizing not necessarily for what really matters but for what’s easy to measure.
Well, as you might expect from any self-respecting James C. Scott fan, Davies praises Recht for resisting the profitable route of hyping this trajectory into AI singularity cheerleading. Instead, Recht asks: what do we lose when we reshape problems to make them algorithmically manageable? Once you’ve defined what to measure and set a success metric, an optimizer will always confess an answer. But all the real judgment happens upstream, in choosing what counts as success in the first place. That same process that advances chips is a convenient way to persuade overly left-brain, ahem, “data-driven” decision-makers.
Recht’s book ends with a thoughtful chapter on how we should make decisions, but Davies also flags an argument that knocks you a bit off balance. He makes the case that randomized controlled trials are better understood as a regulatory tool than a scientific one. I found that a bit haunting. It’s a dark thought that also casts a shadow on our accepted standards of rigorous epistemology.
Recht, Nguyen, and Davies are all quantitatively minded scholars converging on nuanced skepticism of data in what I’d describe as a next-gen version of How To Lie With Statistics. In the first gen, inferences were distortions of what the data told us (i.e. sleights of hand like substituting causation for correlation), but is the new deception to do inference honestly but use proxy measures that, to use a trading term, have a lot of basis risk to the thing we actually care about. And then the bigger question is how much of that is laziness vs trying to manipulate the terms of discourse.
Money Angle
This week I taught Class 5 of the Investment Beginnings series I’ve been doing with the 12+ year olds.
It’s the last class in the series before we do “labs” in July. During lab, we’ll convene when the market’s open and I will give each kid individual attention as they execute an investment. I want to make sure they know how to read the screen, navigate their broker site, see the confirmation of the execution etc.
This last class was special. I’ve been posting all the materials online and there are families following along remotely. One dad sent me an app that consolidated and vibecoded the slides and games. He and his son worked on the project together:
https://investment-class.vercel.app/
And this next part blew people’s minds in the class, not to mention my own. A mom brought her son from Miami because he’s been obsessed with the class and wanted to be here in person with the other kids. I’m speechless. Supermom and superkid.
We took them to dinner with my family and brought along my son’s good buddy so our visiting friend would know a few people before stepping into the class. I can’t gush enough about how nice this all was.
When Class 5 ended, a lot of parents came to talk to me and said all this kind stuff and gave me totally unnecessary, generous gifts (I would have done the same so I get it but also just feels like too much). The most important thing is how all these kids’ gears are turning. It feels like a no-brainer to really clean this up (I learned a lot from doing it and know how I’d mod it in the future) and turn it into something. Maybe a well-produced YT thing, but I’m stretched pretty thin. We’ll see, I guess. Famous last words.
Anyway, here’s the outline of class 5 and link to all the materials from the classes.
Class 5 — Making Trades & Reading Markets
- Different kinds of auctions and how markets are continuous auctions
- The order book: bids, asks, spread, and what “depth” actually looks like
- Price discovery as consensus — the price aggregates what everyone knows
- Market hours, plus what pre-market and after-hours really are (and why beginners should avoid them)
- Public vs. private markets, with real estate as the bridge example
- How an IPO turns a private company into a publicly traded one
- Why baskets exist: the easy button for diversification (callback to Class 4)
- Three kinds of baskets — index, themed/sector, manager-picked
- ETF vs. mutual fund: same idea, different checkout (auction all day vs. one daily NAV)
- Index construction math: cap-weighted vs. equal-weighted, with four real stocks
- Why SPY and RSP — the same 500 names — can produce materially different returns
- 🔨 Homework: talk to parents about a brokerage account, ahead of the July lab where students place their first real trades
We spent much of the class doing a mock trading game:
How the game worked
There are 16 kids.
- Each gets 2 cards — that’s private info
- There are 3 “stocks”: Hearts, Spades, and Red
- At the end of the game, each stock settles to the sum of the cards held collectively in that category across all 32 dealt cards
- Card values run 1 to 13 (ace to king)
Kids bid, offer, and trade with each other based on what they think final settlement will be. They log transactions on index cards they carry.
Every few minutes, news hits — I reveal some of the remaining 20 cards. These are cards that will NOT contribute to the value of the 3 stocks.
The Teaching Moments
Basic valuation
- What’s the maximum value of each of the 3 stocks? (Also a fun way to teach someone to quickly compute the sum 1 to N.)
- What is the fair value of the stocks at the start of the game, when no common information has been revealed?
Information and private signals
- What is the fair value of Hearts if you’re holding the 9 of Hearts?
- Ask the kids: what’s a good hand to be dealt, and why? (A very simple exploration of what “information” actually is.)
- After news is revealed, how do you update fair value? Walk through the exact math.
Reading flow
- Your fair value is always subject to adjustment based on flow. What is Alice’s bid generally saying about Spades? What is Mike’s offer suggest about Red?
- At the end, computing P/L is a big exercise — marking trades to settlement.
We didn’t go into crazy depth on any of these. Just getting a basic understanding easily takes a group of 16 kids an hour, and even then some are lost. Totally expected. Honestly, many adults are too.
It’s super interesting to see who gets it very quickly though.
The origin of the game
This was the first trading game I remember doing as a trainee at SIG. All the new hires in NYC played while the trainees who had been around for 3-9 months traded options on these “stocks.” Their hedge orders would get sent into our trainee market!
Money Angle for Masochists
Not to deter any stubborn bears, but just understand your history. In 1999, the Nasdaq returned 86%.
If we ignore the small 3.2% down year in 1994, that run looks even crazier and capped with an insane blow-off top.
The 1999 blow-off top is pretty interesting from a how-do-I-reconcile-option-pricing-with-real-world lens.
I’ve written a bunch on how volatility measures are sensitive to sampling periods.
See:
Volatility scaling can feel unintuitive.
If an asset’s annual standard deviation (ie volatility) is 16%, then its daily standard dev is ~1%
Well, from that, it’s clear that you can have say a 10 sigma move in a day, but not in a year. That alone seems to point to a weakness in how we scale volatility through time.
But this is mostly resolved by measuring distances in lognormal space correctly. When you do that, you find that +86% is MUCH closer than -86%.
Just screenshot the formula in the post and treat Gemini like a calculator:
This explains why calls that are 50% OTM calls are worth more than puts that are 50% OTM (even if the spot and forward were the same).
To consolidate knowledge, it’s why collars can look so attractive in high vol stocks:
“Stock pickers market”
I’m old enough to remember when investment managers complained that the Fed drove the market, everything was correlated, and there was little reward for discerning between companies.
Well, we’re in the opposite world.
This is showing put skew falling and call skew rising in QQQ:
This is front-and-center to the options market:
There’s a record disconnect unfolding in the trading pits right now
A chart from the article shows the spread between the weighted avg stock vol in the SPX vs the index vol. It’s another proxy for cross-correlation as the index vol is dampened relative to the stock vols because the stocks are doing a great job diversifying each other.
A stretched relationship can get more stretched. But as it stretches, there is a mathematical reason why the spread would revert. Think of the limit. If 1 company achieved singularity and ate all the other companies, its weight would increase relatively as each dollar it made was a dollar less for the others until it was the index. This is NOT a tradable idea. That is a make-believe world. I’m only being pedantic to help you move the pieces around in your brain to help you see how they fit together.
But the price action of anything related to AI ripping vs everything else is a giant singularity trade, and from that context, the low correlation makes superficial narrative sense for now…a handful of companies are expected to eat the rest.
But I’ll pose this one…if instead this handful of companies are becoming the COGS of all other companies, wouldn’t that look more like the stories we’ve heard that I had Claude reconstruct by asking it to describe the circular revenue phenomena in bubbles:
Company A buys ads on Company B’s site. Company B uses that revenue to buy servers/software from Company C. Company C buys ads on Company A’s site. Everyone books revenue, everyone’s growth numbers look great, valuations rip higher — but no net new money is entering the system from outside customers. It’s just the same dollars chasing themselves around a closed loop, with each pass inflating reported revenue.
The poster children were the late-90s telecom and dot-com names. Global Crossing and Qwest got nailed for swapping fiber capacity with each other and booking both sides as revenue (”capacity swaps”). AOL was accused of round-tripping ad deals. A lot of dot-coms were essentially selling ads to each other, with VC money funding the ad budgets — so the “revenue” was really just recycled venture capital.
The concern isn’t fake deals today, but the circularity possibility means index vol will have its revenge. Good luck with timing though.
SpaceX
It’s interesting to hear Mitchell step through the numbers of how much day 1 shares need to be absorbed and how unprecedented this is. Recall in PTJ’s interview on Invest Like The Best:
2000 was the easiest bear market I’ve ever seen in my whole life. It’s got so many similarities to right now, in the sense that the bear market of 2001 and 2002 were a consequence of all the IPOs in ’99 and 2000. And then as they unlocked, you just had this never-ending cascade of selling, that’s a great way of putting it. And we’re getting ready – I want to say that the contemplated IPOs for next year are going to be five or six percent of market cap. So why are we where we are right now? Because we’ve been retiring two or three percent of market cap, probably a little less than 2% of market cap, every year without fail for the past 10 years [through buybacks.]
And so now all of a sudden, you’re gonna completely reverse that math. And so, I don’t think it necessarily happens instantaneously with the IPOs, but then there’ll be the unlocks. So you can see a situation where, okay, maybe we go through some kind of rolling top. And then 18 months from now – six months, we’ll have to look at the unlock schedule. But you’re going to want to watch those because that’ll just be adding equity supply. And you’ve already gonna be diminishing the buybacks because of all the commitment to capex from the hyper-scalers. They’re already gonna be eating into their cash flow.
The current set-up feels very strange. It seems too easy to think it’s going to be a local top right? But it’s a hard idea to resist. It feels like it’s a negative for gross returns and then under the hood, possibly chaotic for sector flows (like to absorb the IPO do people rebalance out of what went up the most recently? That feels like a pretty natural idea).
You have pockets of extreme bullishness manifesting in options with cheap put skew and risk reversals and left-for-dead implied correlations but the bearish factors are also common knowledge:
- IPO issuance
- Midterms (assuming Dems are the bear choice)
polymarket on 5/29/26 - elevated bond yields
In other words, current prices are NET of everyone knowing about these factors. By the way, this is always the problem with markets. If you see something on the horizon and think it’s bearish, whose to say the current market wouldn’t just be higher if that thing you’re latching on to is observable by others.
What’s the more contrarian position right now, to be bullish or bearish?
The option point spreads have shifted in a way that suggest bullishness is consensus.
Moontower agent
The moontower agent has been in the wild for a month now, and it’s been an awesome companion to help you reason through trading questions. It’s connected to the data we buy and process, and it’s tuned and under ongoing reinforcement learning for trading contexts.
It’s powered by AI harnesses, of course, so the output is non-deterministic, so it’s helpful to report back on what it does well and poorly so we can keep course-correcting.
Thursday webinar
Been a lot of recent chatter about volatility funds as a strategy and as a business. There’s a lot of misconceptions about them. Which makes sense, it’s an opaque world.
I’ll host a live webinar on Thursday for paid subs to share some thoughts on the topic.
Details to follow this week.
This Week In The Options Trench
Stay groovy
☮️
Moontower Weekly Recap