the dirties are down the cleans are up

On June 2nd I tweeted:

June expiry in USO vol change on the 3% rally… OPEC agreed over the weekend to hike production…but you saw the Ukraine-Russia developments. Competing bullish/bearish effects Vol is lower today if you look at the June surface.

But it’s probably up on the day on @moreproteinbars dashboard.

Why?

(a junior option trader interview question @bennpeifert might ask would be compute the actual vol change)

An eager beaver wanted the answer key:

I wouldn’t share Tina’s answer key even if I had it. And I don’t. But understanding how to answer my “interview” question is definitely a prerequisite to Tina’s process.

And we can discuss that because the answer is available from putting together a few key concepts that I’ve already written about in detail. The concepts are conveniently self-contained within just 2 previous articles.

To try to pack even more learning we’ll add some spice for those of you who want to grab a pen and paper to give it a go — I will pose the question with some data to you and Google’s NotebookLM to see how far you each get.

If you want to just read along through the solution without burning glucose nobody is gonna judge you.

Decomposing the vol change on Monday after the event

Let’s start with the 2 articles that hold the keys:

I passed those to NotebookLM.

Now the question as I posed it to NotebookLM:

Scenario

OPEC nations are having a meeting on the weekend of May 31–June 1.

At the close of May 30, near-dated options have an “event premium” baked into the implied vols (note: these vols are based on a standard 365-day model).

In other words, the quoted vol includes the upcoming event, as opposed to having the event premium removed. If we had the event premium removed, we’d call that the “clean” vol.

On Monday, June 2, the implied vol is actually up, and the USO (oil) ETF is up 4%.

Question:
Net of two opposing forces —

The weekend effect and
The fact that the event premium comes out of the surface —
is the clean vol up or down on Monday morning, and by how much?

We’re focused on the June 13th expiry.

Vol at the close of May 30: 36.2%
Vol on the morning of June 2: 37.3%

Action items:

Can you conceptually set up the solution to the question?
Can you compute the clean vol change (i.e., back out the effect of weekend decay and event premium)?

This is your time to try #1. Include the assumptions you’d need to make. (This would be a good interview question for an option trader.)

I’ll confess, there’s not enough info to solve #2. I posed it to the LLM to see what it would say (all of the LLM output will be shared below).

I’ll handle #2 in the solution but there’s enough here to at least identify the data you need.

“dirty vs clean vols”

Dirty vols are the IVs generated by an option model. Whenever you look at an IV from a vendor or broker you are looking at a dirty vol.

The most common models assume a 365-day year. They assume every day is created equal. There are also 251-day models (these subtract weekends and holidays). A 251-day model assumes time only passes on business days. Regardless of which model you use (as long as you don’t mix and match), if you are comparing vols between assets on the same calendar this assumption cancels out. Not a concern.

“Clean” vols are an attempt to smooth vol changes rather than accept artifacts from the market understanding that vol time doesn’t pass uniformly. Not every day that rolls off a calendar is 1/365th in vol time. And weekends aren’t just zero time.

Vols can be cleaned by adjusting the DTE in a standard model to the DTE in a more accurately specified calendar.

But we can clean vols for events as well.

Clean vols remove or extract events so they can be compared both along an asset’s own term structure AND across assets. If NVDA has high vol because of earnings it doesn’t mean it is expensive compared to QQQ. You want to compare event-normalized or clean vols.

cleaning the USO vol for both the OPEC meeting and the weekend

Let’s focus on what we observed from our dirty 365-day model for the June 13th expiry.

Vol at the close of May 30: 36.2%
Vol on the morning of June 2: 37.3%

In that model how many days to expiration (DTE) are the close of May 30th and the open of June 13th?

Close of May 30th = 14 DTE or 14/365 ~ .03836 of a year

Open on June 2nd = 12 DTE or 12/365 ~ .03288 of a year

Because of weekend theta effect, an artifact of treating weekend days the same as business days, we expect dirty vols to go up on Monday just visually if the clean vol is unchanged. To reiterate from that post, a 365-day model treats every day the same, but the market understands that “vol time passes more slowly when the market is closed” — which means in a properly specified calendar there is relatively less DTE transpiring over the weekend than what you see in a 365-day calendar. On Monday morning the 365-day model thinks there is relatively less DTE compared to a better-specified model so for a given option price the dirty vol (ie 365-day model vol) must adjust higher.

In the dirty model, .00547 of the year elapsed [ie .03836 – .03288]

We’ll use a calendar specification from that weekend theta post where we say that non-business days count as 50% of a business day for vol time purposes.

If we continue to denominate our basic unit, a full trading day, as 1.0 and weekend days or holidays as .5 we get the following tenor:

251 x 1.0 + 114 * .5 = 308 day calendar.

Using this calendar let’s see what the DTE is on the close of May 30th.

Close of May 30th = 10 regular days + 4 weekend DTE which get a weight of .5 or 12. We divide that by the 308-day ruler. 12/308 ~ .03896 of a year

Note there is more time to expiration as a percent of a year than the 365-day model which only had .03836!

To convert the dirty vol to the cleaned 308-day model vol we multiply by the root of the relative DTEs

36.2% x √(.03836/ .03896) = 35.9%

Note that we expect a lower IV in the clean model because for the same option price we had more time ‘til expiry — so the vol must be lower in the the model with more DTE. This is a good double-check on numerator/denominator confusion when converting.

Where are we?

At the close of May 30th, we observe a dirty vol of 36.2%. when we clean it with a 308-day model which assumes variance passes half as fast on the weekend we compute a clean vol of 35.9%

So for the vol to be “unchanged on the morning of June 2nd” we need to observe a dirty vol that translates to a 35.9% clean vol.

That’s a simple algebra problem using the relative DTEs on the morning of June 2.

We already calculated the dirty DTE that morning to be .03288 as a fraction of the year.

What is the clean DTE according to a 308-day model?

10 business days + 2 weekend days which get 50% weight = 11 days divided by 308-day year:

11/308 = .03571 = clean DTE as a fraction of a year

Recall:

Dirty DTE on the June 2nd open is only .03288 of a year

Note that the gap between clean DTE and dirty DTE has grown even wider which is a clue that the dirty vol the model implies must go higher for a given option price.

Here’s the algebra:

Clean IV * √(clean DTE/Dirty DTE) = Dirty IV

⚖️You may have noticed that this is just a re-arranged balancing identity where:

Clean Variance * Clean Days = Dirty Variance * Dirty Days

It might be easier to remember that and conjure the algebra as needed.

Again the question we care about is what does the dirty IV have to be such that we believe the clean vol of 35.9% is unchanged?

35.9% * √(.03571/.03288) = 37.4%

The dirty vol must go up to 37.4% for the clean vol to be unchanged!

The dirty vol went up to 37.3%…pretty close to unchanged clean vol even though any vendor tool will tell you that IV is up.

I suspect Tina’s dashboard said vol was much closer to unchanged than up over 1 click.

…except there’s this giant elephant in the room.

How much vol should come out of the surface after the meeting?

We start with asking how much vol was baked into the surface before the meeting. The technique is described in how an option trader extracts earnings from a vol term structure so I’ll just skip to output.

It looks like there was a small amount of event vol baked in. I’d estimate about a 2.50% straddle for the move which equates to a 1-day 50% annualized vol.

The picture below is the result of changing the move size until you get a relatively smooth vol term structure.

The smooth vol term structure imputes a dirty, event-extracted or base vol of ~ 35%

Let’s recap our expectations recalling that dirty vol closed on Friday at 36.2%

We expected dirty vol to up to 37.4% to keep the clean vol unchanged net of the weekend effect. +1.2% expectation
We expected dirty vol to fall to 35% once the OPEC meeting has passed. -1.2% expectation.

They cancel out!

On balance we expected the dirty to be unchanged on Monday at 36.2% to reflect a clean vol falling from 35.9% to the event-free 35% net of calendar effects.

We observe that the dirty vol actually increased to 37.3% so we can say that oil clean vol was up on the day.

If there was no event baked into the surface, our baseline would not be “clean vol falls on Monday” and we would therefore have said Monday’s vol was unchanged despite the dirty vols optically being higher.

💬Inserting my own comment

It’s not unusual to find this very confusing. Vol cleaning is a necessary step for professional vol trading. It’s the epitome of my measurement not prediction idea where much of the battle is just seeing the present clearly. “Is vol actually up or down?” is a hard question.

To get more grimy — think about earnings season where lots of single-name vols are higher because of upcoming announcements. If you simply tracked implied correlation across the year without normalizing base vols it would appear that they became extremely low 4x a year as stock vols increased relative to the index.

If you don’t normalize for events, then you need to compartmentalize what normal levels of implied correlation are inside and outside of earnings season.

You can start to understand why Tina is quiet about her own process. These processes vary across professional vol shops but what they all have in common is they have a process — you can’t trade high volumes for slivers of edge while being wrong about whether the vol is 25.1 vs 25.4.

The good news is for anyone else using options these differences are just brain damage. If you’re trying to trade for .3 vols of edge you’re already cooked by your execution costs and broker funding rates. Long/short funds don’t try to trade for bps of edge anymore than option users should be trading for a quarter click.

Monday, June 2

Back to the main narrative.

On Monday, June 2nd we saw that the dirty vol opened up 1.2 points which coincidentally matched how much we think the clean vol was up. The ETF itself opened higher 4% — it wasn’t that the OPEC production hikes were bullish (as in they were less than expected) but that was the weekend the Ukraine drones surprised parked Russian bombers.

By the end of that Monday, oil relaxed to being up only 3% on the day and dirty vol for the June 13 expiry dropped to 35%. What clean vol is that by the end of June 2nd?

Dirty DTE = 11/365 ~ .03741

Clean DTE = 10/308 ~.03247

First, look what happened — after June 2nd elapses the DTE flips…there is much less clean DTE than dirty DTE now! It makes sense — when one business day elapses 1/308 rolls off the board in a clean model but only 1/365 elapses in a calendar day model. This happens all week as the dirty days increase relative to the clean days (and therefore the dirty vol declines steadily thru the week to keep the clean vol unchanged before sawtoothing up over the weekend again).

For a dirty vol of 35%, we know the clean vol must be higher because for a given option price we believe there is less DTE.

The algebra:

Clean vol = Dirty Vol * √(Dirty DTE/Clean DTE)

Clean vol = 35% * √( .03741/.03247)

Clean vol = 37.6%

Instead of clean vol falling to a base vol of 35% after the event, we see that it’s actually up to 37.6% up from 35.9% on Friday’s close. An increase of 1.7 points but 2.6 points vs our expectation of clean vol falling from 35.9% to 35% after the meeting. Hence the title of this post…”the dirties are down the cleans are up.”

NotebookLM

I fed NotebookLM both articles and that prompt from earlier.

You can see its work here.

It’s a pretty amazing synthesis tool, auto-generating a study guide, briefing, FAQ, mind map and even a podcast so you can learn on the go (my interest in this was renewed as I’ve been listening to an auto-generated podcast by a professor friend on the topic of business education).

Some screenshots of Notebook LM:

The mind map with just a few toggles open:

That’s a wrap for today.

The “most important” gambling topic and a riddle

Non-self weighting strategy

We watched Ocean’s Eleven with my older son Friday night (we’ve recently established a Friday night ritual where we rotate who picks the movie and who picks the pizza. This past Friday we did a double feature — Dodgeball, which the kids loved and O11 once the little guy went to bed).

Clooney gives a tiny speech in response to Pitt’s skepticism about his motive behind the heist.

Rusty: I need the reason. And don’t say money. Why do this?

Danny: Why not do it?

[Rusty shakes his head]

Danny: ‘Cause yesterday I walked out of the joint after losing four years of my life and you’re cold-decking “Teen Beat” cover boys. [pause] ‘Cause the house always wins. Play long enough, you never change the stakes, the house takes you. Unless, when that perfect hand comes along, you bet big, then you take the house.

I emphasized that part because it’s a catchy encapsulation of what Mason Malmuth writes in Gambling Theory & Other Topics, a book I read as a trainee. The most important principle in gambling is to employ a non-self-weighting strategy. In other words, vary your bet size with the opportunity.

I don’t want to get too hung up on whether this is the “most important” as Malmuth contends (you can certainly make the case for “having an edge” in the first place), but it might be the most underappreciated with respect to how we port it to real life. Varying your bet size in blackjack is well-understood, but Malmuth argues for more obscure examples like the brevity of the Gettysburg Address, a masterful bet on the right words and quantity of words in which Lincoln varied his rhetoric for maximal payoff.

It’s a provocative reminder to be careful where you enable life-decision autopilot. If you need inspiration to find areas of your life where you can vary your metaphorical bets, paste this whole section into an LLM and prompt it to give examples in your real life.

[There’s probably an interesting essay to be written about the tension between the value of habits vs the punchy payoff of straying from them in deliberate ways.]

Money Angle For Masochists

🔘Interview riddle

You press a button that gives you a randomly uniformly distributed number between $0 and $1

Each time you press, you have two choices:

1. Stop and take this amount of money

2. Try again

You can try 2 times total.

What’s the game worth?

Full discussion on Twitter with solutions as well as general solutions to harder versions of the same question.

the investment industry is a placebo

Giant swaths of the investment industry are doing nothing but selling soothing balms. Placebos. It is the vitamin industry at best and a penis-pill pop-up banner when it “democratizes private investment.”

But it could be no other way.

If you can statistically prove your strategy has alpha you also know exactly how to price it to take all the surplus. The market for making alpha is like any market — it equilibrates based on supply and demand. There’s more wealth out there in search of a return than the capacity to absorb which is why pod shop bosses feast.

I’m not knocking this. Their CAGR, ie returns net of vol drag, through insane market environments after fees are perfectly fine. In fact, it seems that the HF market is more efficient compared to the amount of over-earning for beta that went on in the early 2000s between the dot-com meltdown and the GFC. People that made several lifetimes of loot telling just-so stories to allocators who didn’t notice how Moneyball applied to their field.

The GFC disillusionment revealed many stories to be no more than fairy tales. There was an opening for a new story that perfectly complemented the spread of technological capacity and its rider, technical skill.

Evidence-based investing.

With large data sets and faster computers we could solve investing like a physics problem. Engineers aren’t fooled by steak dinners and silver-tongues. The softest stuff they read is Kahneman who holds the why for why their factors work. All of it has the sheen of the scientific method.

Except there’s one lingering inconvenience. It’s an inconvenience that’s obvious to gamblers. I think most investors can feel it in their bones. Not surprising, we are all natural gamblers to some extent (we eat hot dogs and let strangers drive us around).

The inconvenience is the uselessly long feedback loops. Which we are going to discuss. But it’s worth mentioning that the feedback loops double as a defense for asset managers. They get to say “this works over time and if it worked all the time it wouldn’t work”. That’s true but it doesn’t solve my problem — I STILL HAVE NO FEEDBACK — plus the defense IS convenient to the fee collector.

So we’re left with “keep buying my pills because they might work”. Good luck getting a refund if the fish oil doesn’t make you live longer. The whole arrangement is irreducibly uncomfortable. That’s why it’s called the Paradox of Provable Alpha (and why I need to one day finish writing moontowermoney).

All of these thoughts were stirred up again as I listened to Adam Butler on Excess Returns.

An early quote in the interview:

The size of these edges is so small relative to the noise we encounter daily — especially compared to the gyrations of the underlying indices — that it’s very difficult to make high-confidence, informed choices in advance. In other words, it’s hard to know which edges or strategies to allocate to in a portfolio with any certainty that they’ll outperform a random selection of other possible strategies over the next 10, 20, or 30 years of your investment horizon…your skill in selecting strategies in advance based on even very long histories of performance is pretty close to zero.

I should clarify — Adam distinguishes investment or factor style edges from trading or niche forms of investment that rely on some form of information advantages that have built over time. I think Adam would agree with me when I say trading is like any other business but for superficial reasons gets confused with investing.

I agree with his understanding of pod shops:

Pod shops are really looking for people that genuinely have alpha, so I think it’s useful to kind of distinguish between what we might call sort of systematic factor strategies and alpha.

Alpha comes from somebody who has very particular niche insight or information or experience within a fairly narrow domain of the market. So, for example, we have a client who allocates to a municipal bond manager. Now, this manager has a hard cap at about a billion dollars. The team that runs it spun out of what used to be the largest muni market-making desk — worked there for 20–30 years.

What did that give them? Well, it gave them access to knowledge of where all of the flows from muni bonds — all of the issuance from the muni bond sector — are coming from, the different state governments, who the decision-makers are there, how they can get inside information on what type of issuance is coming down the pipe. And then, being at the center of flows in the muni market — which is a very niche segment of the market — right?

I think that’s just one example, but there are many. For example, somebody who worked for 20 years in the electricity markets — and electricity is a very nuanced pricing market with a very small number of key players, and is largely driven by changes in regulations at the state level and the county level. So, having very specialized knowledge of that, from having worked and gained experience inside the sector, gives you a real edge.

Right, so these are the types of strategies and people that the pod shops are looking for right now. These typically tend to be fairly illiquid strategies — right? You can’t have Elliott Management, a $70 billion firm, running just a niche electricity strategy or a niche muni strategy. But the goal is to find hundreds of people who are all running these niche little strategies, that will all require liquidity to take advantage of opportunities at completely different times from one another, and putting them all together in a diversified basket.

Now, I’m sure there are also very scalable strategies in there as well, that maybe are running more liquid equity strategies or option strategies or whatever. What I fundamentally believe — and my insight from knowing people at those shops — is that the majority of the alpha that you can’t get anywhere else at scale comes from the assembly of many different, less liquid, small niche players that are all operating together in an ensemble.

That last sentence harkens right back to the idea of combining multiple strategies into a portfolio that has a higher Sharpe than any of the constituents by letting the zigs neutralize zags to shrink the denominator.

So I’m nodding along then the host, Jack drops a tight line:

I think Corey [Hoffstein] showed in Factor Fimbulwinter that the amount of time we would need to show that is longer than our investing lifetime.

So this does become about faith.

The verb “showed” following the subject “Corey” is a clue that I get to learn something cool today.

So I read the article referenced:

🔗Factor Fimbulwinter (8 min read)

I’m going to jump to the end because I think Corey stages why he takes the approach he does in the article (emphasis mine):

The question we must answer, then, is, “when does statistically significant apply and when does it not?” How can we use it as a justification in one place and completely ignore it in others?

Furthermore, if we are going to rely on hundreds of years of data to establish significance, how can we determine when something is “broken” if the statistical evidence does not support it?

Price-to-book may very well be broken. But that is not the point of this commentary. The point is simply that the same tools we use to establish and defend factors may prevent us from tearing them down.

Corey uses fire to fight fire.

This is where the learning begins. Let’s see what he does.

We ran the following experiment:

Take the full history for the factor and calculate prior estimates for mean annualized return and standard error of the mean.
De-mean the time-series.
Randomly select a 12-month chunk of returns from the time series and use the data to perform a Bayesian update to our mean annualized return.
Repeat step 3 until the annualized return is no longer statistically non-zero at a 99% confidence threshold.

For each factor, we ran this test 10,000 times, creating a distribution that tells us how many years into the future we would have to wait until we were certain, from a statistical perspective, that the factor is no longer significant.

Sixty-seven years.

Ok, I’m going to raise my hand in class.

I didn’t really understand the method.

Awesome I get to learn something new…which means you do too! This is pretty cool.

Luckily, there’s a tireless teacher known as ChatGPT. I wrangled with this professor at office hours until I was able to have it teach me in words I or a middle-schooler can understand.

🧪 How They “Repeat with New 12-Month Chunks” — Using a Coin Flip Example

We’re walking through the logic of how a statistical test updates with each new batch of data — using a coin flip as our analogy.

🎯 The Goal

We want to detect whether a strategy (like a stock factor or a biased coin) has stopped working — i.e., its returns have gone flat.

Think of it like this:

A coin used to land heads 60% of the time.
But now it’s just a fair coin (50/50), we just don’t know it yet.
So we flip it 12 times (like one year of monthly returns), check what it shows, and keep flipping until we’re statistically convinced it’s no longer special.

🧪 The Setup

The coin is now fair (true heads probability = 50%).
We start with a belief: “maybe it’s still biased to 60%.”
We flip the coin in 12-flip chunks, and after each chunk, we update our belief.
This mimics what the researchers did — taking random 12-month samples from a flat return series and asking: “Does this still look like a working strategy?”

🔁 Step-by-Step Walkthrough (One Simulation)

Let’s pretend we have a long list of coin flips (each 1 = heads, 0 = tails), all drawn from a fair coin.

Here’s a fictional sequence:

[1, 0, 1, 1, 0, 0, 1, 0, 1, 0, 0, 1, 1, 0, 1, 0, 1, 1, 0, 0, 1, 0, 0, 1, ...]

We’ll take this in chunks of 12 flips, like 12 months of flat returns.

1️⃣ Year 1 — First 12 flips

[1, 0, 1, 1, 0, 0, 1, 0, 1, 0, 0, 1]

Number of heads: 6
Sample mean: 6 / 12 = 0.50

We compare this to our hypothesis that the coin is biased to 60%:

“Is 50% close enough to 60% that we still believe the coin is special?”

Yes — this result is plausible from a 60% coin, so we keep going.

2️⃣ Year 2 — Next 12 flips

[1, 0, 1, 0, 1, 1, 0, 0, 0, 1, 0, 0]

Heads: 5
Mean: 5 / 12 = 0.4167

Now combine both years:

24 total flips
11 total heads
Cumulative mean: 11 / 24 = 0.458

Still not far enough from 0.60 to be statistically confident the coin isn’t biased.

3️⃣ Year 3 — Another 12 flips

[0, 0, 1, 0, 1, 1, 0, 0, 1, 0, 1, 1]

Heads: 6
Cumulative now: 36 flips, 17 heads
Cumulative mean: 17 / 36 = 0.472

Now let’s check this against our original belief (that the coin is 60%).

We run a z-test:

Expected mean under H₀ (the hypothesis): 0.60
Standard error: sqrt(0.6 * 0.4 / 36) ≈ 0.0816
Z = (0.472 – 0.60) / 0.0816 ≈ -1.57

Since -1.57 is not extreme enough (we need z < -2.58 to reject at 99% confidence), we still can’t say the coin is fair.

🔁 Repeat the Process

Each year:

Add 12 more flips
Update the total number of heads and flips
Recalculate the cumulative mean
Re-test: “Are we now confident the coin isn’t biased?”

Eventually, the sample mean will drift far enough from 0.60 that the test crosses the 99% threshold. At that point, we’d say:

“I’m now 99% confident this coin is no longer special.”

💡 Why This Works

Even if the coin is fair, short sequences can look biased just by chance. You might see 8 heads out of 12 once in a while — that doesn’t mean the coin works.

So the researchers repeat this full process — from scratch — 10,000 times, each with a new random sequence of fair flips.

Then they record:

“How many years did it take before the test figured out the coin was dead?”

On average, the answer was:

67 years.

🧮 What This Means for Investing

If a strategy (or “factor”) stops working but still produces noisy returns — it might take decades before we’re statistically confident it no longer works.

The noise in the short term can mask the truth for a long time.

I enjoyed Corey’s technique because it gives you a sense of proportion between signal and noise. In so many domains where an assertion is made, that proportion is absent. I always think about how a CRO I worked with would reflexively try to put error bars on any metrics presented in a chart. It’s good epistemological hygiene. It automatically triggers awareness of base rates and outside views. It’s not a panacea for truth, but it rules out obvious bullshit —randomness sold to you as signal. That will save you time and money in life. It may not increase your top line but it will save you your bottom line.

Netting risk, “The Hopeless”, John Arnold and more

Friends,

A grab bag of practical finance nerd stuff today.

Behind the paywall, subs get an exclusive complimentary pass to the 3-hour commodity trading seminar I did for QuantInsider. It’s especially useful for equity folks looking over the fence (especially relevant these days with oil in the news.)

🧵A thread on COIN vol since the inclusion in the SP500 (moontower)

Note Vivek’s reply (I won’t doxx who he is but many of you can guess):

✍🏽More Grit, Less Drip (Market Jiujitsu)

Ari watched me and Mark Phillips’ TSLA Covered Call video:

“If you do nothing else, go watch the video. These are two professionals and the dialogue between them is as important as the research itself”

He draws the lessons that we were trying to convey:

Why is this different than doing technical analysis and/or doing backtests out the wazoo until you p-hack a bad solution? They are not optimizing. They are doing a research project that forces them to ask questions and iterate through. In fact, this roughly the education process that Darrin Johnson took to build himself into the trader that he is today and the inspirational figure that he is, too.
General James Mattis has told us “If you haven’t read hundreds of books, you are functionally illiterate, and you will be incompetent, because your personal experiences alone aren’t broad enough to sustain you.” The same goes for the market where there are not (exactly) books telling you what may work or not. In fact, if you want the good stuff, the stuff that works that is not blandly generic (yet still high quality), then you have to roll up your sleeves and go to work. The key part is that you are not doing this research to lead to an answer. You are doing it to lead toward the tools and the expertise that will lead to a winning trading strategy. This is what Spitznagel refers to as the roundabout strategy in the Dao of Capital. You don’t do this for the immediate reward of a working and foolproof money machine. You do it so that you know the questions to ask and what factors and what risks may lie in wait for you. Then you take that information and strive to either go deeper again or work on the strategy.
If you find this sort of thing interesting, that is a good indication that you are cut out for this. On the other hand, if you find that all you want is the “answer” then maybe this is not for you.
There is quite a bit in the video. There is more that I would like to talk about. I’m sure if I listened again, I’d get more out of it. I feel like this sort of thing is more valuable than doing something more statistical. The actions of doing this for one stock are, of course, living in the land of small numbers. But the crafting of this and the looking at the data and the asking of questions sets the stage for doing something that is more statistical. By statistical, I mean coming up with a set of criteria. But to even get to this point, you need a certain amount of experience and I think it is worthwhile to have a mentor — or a co-worker as these two did.

🎙️Why Asset Allocators Love Multi-Strategy Hedge Funds (Odd Lots)

Ronan Cosgrave, a partner at Albourne, gave a masterclass about multi-strat pod shops on Odd Lots. The whole interview is great but here’s a few topics that I want to highlight because are a big deal if you work at a fund but underappreciated by the casual finance observer.

✔️When Diversification Isn’t a Free Lunch: The Fee Structure Problem in Pod Shops

The traditional hedge fund model charges a management fee plus a cut of the gross performance of the portfolio as a whole.

In a pod shop, things get more complicated.

First, the management fee may be structured as a pass-through, meaning it isn’t fixed—it flexes with the underlying costs, which are often generously defined. Second—and more importantly—the performance fees are charged at the level of the individual PM, not the fund. That means the economics of diversification break down.

Diversification is not a free lunch when you’re paying performance fees on the components of the portfolio. It costs you money—real money.

✔️Pod PMs Actually Get Paid By the LP NOT the GP

Each PM in a pod shop runs their own individual P&L. All expenses—Bloomberg terminals, analysts, trading costs—are charged against that PM’s book. If there’s positive net performance, the PM receives a share of those gains from the fund manager. Importantly, investors themselves don’t pay PMs directly—the manager does. But this structure creates a disconnect between fund-level performance and fee drag.

✔️Netting Risk: When Winning PMs Get Paid and You Still Lose

In a traditional HF like where I used to work, there is “netting risk”.

Here’s a simple example:

PM A is up $10
PM B is down $10

At the fund level, you’re flat. But PM A is owed their their 20% cut—$2—from the overall fund. This is a “netting” problem.

The GP loses money even with flat fund performance, so is inclined to not pay the PM’s that are up what they should.

Pod shops can poach talented PMs who made money but didn’t get paid.

In the pod shop model, the economics are different—and risky for LPs as they still pay the winning PMs even if the overall fund is flat!

The netting risk is shifted from the GP to the LP.

But this risk is not without a benefit…the pod shop will be better diversified than the traditional HF. The reason comes from incentives as you will see in the next section.

✔️Quantifying Netting Risk

Albourne has modeled this dynamic. Across simulations, the average cost of netting risk is about 1% per year. That’s 1% of the management fee being paid to PMs who made money, even when the rest of the fund didn’t.

That cost changes behavior. If you’re a traditional manager, you might start to prefer more correlated risk across pods—everyone wins or loses together—because it minimizes internal netting drag. But that undermines diversification.

To reduce the business risk of netting, you reduce dispersion. But with less dispersion, you get lower Sharpe.

“Alt Data Manipulation”

I just thought it worth posting the full excerpt from Matt Levine’s Tuesday Money Stuff:

We talked yesterday about the wait time for delivery from pizzerias near the Pentagon, which arguably predicted Israel’s attack on Iran, and which more generally is arguably correlated with oil prices. The busier those pizzerias are, the busier the Pentagon probably is, which probably means some geopolitical stuff is going down, which probably means oil prices are going up. None of those things is absolutely true. Maybe some unrelated business near the Pentagon needed a lot of pizzas; maybe the Pentagon’s softball championship is that day; maybe the geopolitical stuff will reduce the price of oil. But it would not be shocking if there is some positive correlation.

I, like, one-quarter-jokingly suggested that hedge funds should pay for a direct data feed of Pentagon pizzeria wait times, since that would be a valuable signal to their commodity trading models. Fine.

Three readers independently emailed me with variants on what in retrospect is sort of an obvious question, which is: “Is it market manipulation to order like 200 pizzas to an office near the Pentagon, and then buy calls on oil?” A few points:

Not legal or investing advice!
This assumes that it would work, which in turn assumes that hedge funds are trading on this data. My thesis yesterday was something like “the oil futures market does not move immediately in reaction to Pentagon pizza delivery wait time data, so if you traded on that data, you would be ahead of the market and make a profit.” My readers’ implicit thesis is something like “since it was published in Money Stuff, now canonically the market will move immediately in reaction to Pentagon pizza delivery wait time data, so if you manipulated that data, you would be ahead of the market and make a profit.” But obviously if nobody trades on that data then you’ve just wasted money on pizza.
More generally, “manipulate alternative data that is correlated with some security prices, expecting sophisticated hedge funds to trade on that data, and trade the correlated securities ahead of them” seems like a rich field for study in modern finance. As people seek more obscure sources of information, data sources that are moderately correlated with asset returns rather than leaks of merger news, there are more opportunities for both manipulation and plausible deniability. We have talked a few times about “shadow trading,” which is the related practice of (1) getting inside information about some company and (2) trading some correlated security (rather than the company’s stock, which will obviously get you in trouble). The field of alt-data manipulation is broader, though — if hedge funds are reading tweets, you can write a lot of tweets, etc. — and less obviously illegal. Trading one security with inside information about another security seems bad in some fuzzy but obvious way; ordering too many pizzas to trick people into buying oil is murkier. “Park 100 cars in the parking lot of some retailer announcing earnings next week, and buy calls on the company,” that sort of thing: You were misleading someone, probably (the hedge funds examining satellite images of that parking lot), but why did they think they were entitled to rely on that parking lot for their trading?

🎙️John Arnold on Conversations with Tyler (transcript)

Highlighting these excerpts that have to do with requisite trader skills and his strategy of concentrating on a closed system (and the risk with such a strategy).

Before we get into skills, a reminder from SIG brass:

COWEN: What do you feel is the skill you had that your other traders didn’t have to the same degree?

ARNOLD**:** I’ve always had a difficult time answering this question. I think part of it is, trading is a team sport for sure. I always took the view of to get around smarter people, and listen more than talk.

In terms of traits, here’s what I would offer as my traits:

Number one is this detachment from emotion. There’s a lot of talk about fear and greed driving markets. To the extent that fear and greed change your process, the more you can remove those emotions, I think, the better.

I think there is a component of first principles trading, where first principles of how you look at information. Don’t accept the information as is, but really test all assumptions that go into it.

I think there is a component of being on the perfect point of the confidence spectrum. You have to be confident in order to say the market is wrong, and I’m right, that other people are wrong, because I think efficient market hypothesis is fairly true. But if you’re overconfident, you’ll blow up quickly.

[Kris: can’t agree more. See If You Wait For All The Info You’ll Be Too Late]

There is this notion of being quantitative enough to build the long-term models, but being quick with numbers in order to jump on the trades as they happen.

There’s this aspect of, I think, a chip on my shoulder. Really having this passion for it. You have to have the love of it. That this is the most important thing. I ate, breathed, and slept it. I would be thinking about it first thing in the shower in the morning. I would be dreaming about it. After work, I’d go out with people in the industry and talk about it. It was that real devotion to the markets.

I think there was a timing component, that I always had great timing in my career. Then, certainly, luck’s a part of it.

…

Part of my career and part of the success, I think, was that the business plan I developed was to be an inch wide and a mile deep in this. It was, find this niche and try to be best in the world at it. Don’t expand the focus. It was North American Gas and Power. At some point, LNG started to become relevant and put a small team in Europe, but mostly for information flow for the North American Gas and Power group. There were numerous opportunities to get into oil, to get into metals, or agriculture. Or start trading energy equities, for instance. Every time I considered it, but, stick with the niche and just focus here.

I think the upside was, if we were successful with that, if that plan worked, and we were best in the world, it was going to be enormously profitable. The downside is that the intellectual curiosity starts to sag.

COWEN: In your niche, do you think the skills needed to be a commodities trader, in particular, are different from other kinds of trading? Or it’s just the same?

ARNOLD: I think it’s pretty similar. One of the great things about the natural gas industry for a long time — and it’s still largely true now — is it was a closed system. You could figure it out. It also had this forcing mechanism twice a year. The fundamentals had to align with price more or less twice a year — at the end of the injection season and end of the withdrawal season of gas. So, where price could deviate away from fundamentals for a time period, it had to come back at a certain time. It was a system that was conducive to being modeled. Apply smart trading on top of that, and it created a lot of opportunity.

That’s a perfect preamble to the commodity seminar…

This link will give access to the zoom webinar recording for free:

[link for paid subs in this post]

Overview

The variable that balances the buy/rent equation

You’ve solved one equation with one unknown a million times. For example:

$20 - 2 * $8.99 = X

where:

X = how much change you are owed after handing over an Andrew Jackson for 2 hot dogs at Wrigley Field.

In finance, this uneventful operation is dressed up with the word “implied”. Fix all the observable inputs to an option price and back out an “implied volatility”.

We imply lots of values. The probability of TSLA expiring below $250 by December 2025, “breakeven inflation”, or as my fundamental investor friend likes to refer to value stocks — “low implied forward ROIC” companies.

If you can effectively “normalize” the important inputs (not easy) so that there is only 1 implied value that acts as a statement about market expectations you are arbitrage-pilled.

My tone suggests that there’s something wrong with you, and to be clear there is, but I’m also a fan of doing this. It’s the heart of a replication mindset that is useful for isolating bets on the exact claim you want to make.

[Replication mindset is the bridge of asses that separates just-so storytellers from rigor. And even that bridge is fastened with some loose bolts, like the nuance between real-world vs risk-neutral probabilities.]

This is an obvious preamble to sharing that — I’m buying a house. I’m going to discuss the financial aspect of the decision in a moment, but I want to confront a few points of context and curiosity first.

I’ve been renting an old but groovy house for the past 5 years since selling my old one during the pandemic.
My in-laws (sis and bro) live next door including my niece & nephew and my MIL lives with us. The commune compound life is recommended if you love your in-laws! But we all rent so this was always gonna be temporary. The house we bought is in toy walkie-talkie distance so given one of us is moving it is a best-case scenario.
This is the 5th home I’ve bought in my life — 4 of them have closed within a week of my birthday (and I can’t remember when I closed on the first one so maybe that was too). Weird coincidence.
We have bid on a home probably every quarter for the past 2 years. Our agent’s effective hourly rate makes sense after what we put her through. She is heavily involved in a lot of the research and bloodhounding because every case has some weird hair on it. She sold our old house for us, sold one of my best friend’s homes the same month she sold ours, found us this rental which wasn’t even available to rent but through an old friend who was keeping it vacant, and even bought a home for a Moontower reader who asked me for a good agent when they moved to our area! I’m nonplussed by the average realtor but the real pros stand out. If you need one in the Lamorinda area, I know a gal 😉
Qualitative reasons to buy:
1. My eldest is going into 7th grade. We want to give him his own room.
2. We want to be able to customize our space (we are going to build an ADU at the new place).
3. A renter always lives with the sword of “we need the house back” hanging over their head.

All this said, how do we frame the finances and how does that relate back to inverting an equation to find what’s implied?

We start with this tweet:

My cost-of-living is going up with this purchase. We have accepted this in light of the qualitative benefits outlined above net of the lower-stress “renter” status. I don’t think it should have to be said but buying vs renting is mostly not a financial decision. The finances are a constraint in light of your broader goals but shouldn’t be the primary driver.

[Personal thinking: whether we rent or own, we ask ourselves “At what level do we want to consume housing at?”. Even if we can afford more, we try to be ruthless about what is a must-have vs nice-to-have and not let the nice-to-have creep out like a wolf spider hatching. Neither of us wants to find ourselves servicing interest payments to some mimetic trend. The cost is not just denominated in dollars but in utils of resilience and optionality which are key to peace of mind and lower stress.]

You should use a buy-rent calculator to understand the financial trade-offs.

NY Times Calculator: This is an OG calculator that does a great job of identifying relevant variables and seeing the IRRs over time
Khe Hy’s calculator: Incorporates qualitative aspects to generate how you should lean. A large aspect of the value of this calculator is also in identifying the levers.

Like real estate investors (of which I’m not) who use sanity-check math like a ratio of monthly rent to home price to get a blunt cap rate, I like to tinker with calculators and basic assumptions to find a handy compression of “what does the decision to buy cost me financially relative to renting?”

A lot of this motivation was also to frame the financial decision in terms of economic cost as opposed to just focusing on cash flows which can obscure reality just as cash flow statements don’t equal income statements.

I’ll show my math assumptions below so you can see how your local assumptions would stack up.

I also assume no mortgage as I want to see unlevered math. (In our area, the home inflation is less than current mortgage rates by a lot).

“Excess cost to own” roughly reduces to the spread of your home inflation rate vs after-tax opportunity cost of your cash.

Since I’m assuming purchase in cash, suggesting that the funds you will be using are being taken from the conservative portion of your net worth portfolio (ie bills, intermediate fixed income), it is reasonable to benchmark the risk to low-risk investments. Likewise, if you have a mortgage (ie leverage) you are fine to benchmark to after-tax risky returns.

[Note: Capital gains exclusion is $500k for primary residence, so for high value homes need to discount the home appreciation accrual by a tax penalty.]

Cost/Benefit calc and what it implies

As a high tax-bracket CA resident, these are my assumptions:

Cost to own

Taxes: 1.25%

Insurance .40%

Brokerage amortized over a decade .50%

Maintenance (including things like amortizing the cost of say 1/2 a roof over 10 years): 1.5%

After-tax opportunity cost of cash: 2.5%

Total: 6.15%

Benefit to own

Not paying market rent: 4.8%

Therefore, if you buy it costs you 1.35% per year or $13,500 per $1mm of house.

But you can re-frame that as:

The implied inflation break-even is 1.35%

If your home value appreciates by 1.35% per year, even though your cash flow is negative versus the counterfactual of renting, there is zero additional economic cost to owning. (If the home appreciates by more than the economic cost shifts to renting.)

The pros of this lens is that you can use off-the-top-of-your-head numbers to quantify the annual cost of owning vs the bet you are implicitly making — what home inflation rate am I underwriting to be financially indifferent.

Again, it’s a sanity check, not a reason to pull a trigger.

It’s also useful because it identifies important local variables:

Taxes, insurance, brokerage, maintenance

As I stepped through these numbers I estimated the rent (ie 4.8%) based upon comps I’ve seen anecdotally (rental rates are hard to find and sparse).

Our rent is closer to 3% of the price of a comparable home which means the opportunity cost to buy is higher — 3.15% implied inflation breakeven. If I’m comfortable underwriting 2% inflation then I’m “willing to incinerate 1.15%” per year” for the psychic benefit of owning the house.

A word on risk

All of this these considerations exist in the context of normalcy. If AI wipes out white-collar jobs then there are lots of expensive houses around here. If not, the East Bay bull case is San Mateo and San Mateo is Palo Alto and so on. The starting cost to build in town is about $625 per sq ft. Given that homes trade in the $800-$1,000 ft and land is not free, most existing home sales trade under replacement value and building margins are thin. Demand determines the upside, but replacement costs buffer the downside.

California

When I was in NY a couple weeks ago my friends who live in Bergen County, NJ whose eldest is heading into senior year of HS, told me how common it is in their area to see a “Congrats grad” right next to a “For Sale” stake. As soon as they have an empty nest they’re 86’ing the big house with the fat NJ property tax that reasseses with market value.

You own the house for 20 years while your kids grow up, the next owners remodel the kitchen and repeat the cycle.

Meanwhile in CA, Prop 13, born the same year as I was — 1978 — means couples age in place in a 4,000 sq ft house thus crushing housing turnover. Prop 13 is a call option on inflation and why I’ve argued that CA home prices are not as high relatively as optics would have you believe once you adjust for property taxes. This isn’t just academic. I grew up in NJ and witnessed natural experiements as some of my immigrant family settled in CA (or elsewhere). NJ with both high property taxes and income taxes is utterly wealth destroying. My mother’s house, net of property taxes, has performed about as well as inflation. If you threw a dart to pick an equivalent house in the Bay Area or LA in 1980 it’s a million if not multi-million dollar asset today.

But it’s CA that’s broken. In the lingo of economists, the “excess burden” or distorting behavior of taxes, seems much larger if you impose low property taxes and high income taxes once you reach the limits of the frontier — as opposed to places like TX which have high property taxes and no income taxes.

How does this manifest in the cost/benefit math above?

The taxes as a percent falls over time in CA so the benefit shows up strongly in a multi-period model. Our landlord’s property taxes are probably 25 bps or so. If your home is reassessed, as it is in NJ, that cost scales with your property value.

This is from my intro to On Georgism:

I love living in CA despite its fiscal framework. CA is what you’d get if you told Wario to design public finance. Let me get this straight… a young worker with a good job will pay nearly 50% in combined Federal and State taxes, while an absentee landlord living in Orange County has plenty of time (the app he uses to collect the rent is built by the young worker, by the way, saving the landlord the indignity of paperwork and trips to the local Wells Fargo branch) to scream at “lazy bums” to pick themselves up by their bootstraps, all without a shred of self-awareness about his grandparents’ fortunate decision to buy a regular house in Newport that’s now worth 8 figures and is protected by Prop 13 from high carrying costs?

My mother bought a house in NJ for $70k in 1982. It has appreciated 5x for about a 4% CAGR. She paid 2% property taxes reassessed every year. In real terms, she lost money. But she had a roof over her head.

An equivalent house in CA cost about the same in 1982. I know because she and her father tried to move us to the Bay Area 40+ years ago. They couldn’t find jobs out west, so I was raised in NJ. You could have thrown $70k at nearly any house in the Bay Area in 1982, and it’s worth seven figures today. And the property taxes can be safely rounded to zero because of Prop 13.

In CA, the reward for getting lucky once was to get to stay lucky. In other words a landed gentry. Meanwhile, family formation in Bay Area suburbs is limited to millennials or zoomers who will inherit their parents’ homes and that sweet stepped-up basis. The ladder is officially pulled up. The rentier is on top while the worker falls behind on the treadmill down below.

CA’s fiscal dysfunction likely has many causes. But the output is plain to see — it’s a state that gives preference to building wealth through capital appreciation instead of labor and the real estate market has internalized that logic. But real estate, in particular, land should be considered a reserved word to use a coding analogy. Thinking land is just another form of capital, like a computer or factory is a subtle but profound error.

To understand why, we will journey back to the late 1800s to meet the economist and philosopher, Henry George.

As an (almost) homeowner once again I feel like I’ve joined the landed CA gentry once again. At least Zak gets his own room.

two vol trader interview questions

Thursday’s post the dirties are down the cleans are up took the form of an extended interview question. If you are high-volume professional vol trader, the topic of vol time is fundamental but I don’t see much written about it. I hope my posts on it fill the gap.

For non-pros it probably best serves as a bicycle for the mind or a seed of inspiration but I wouldn’t stress over it. I suspect it does since tweets like this are popular even though I’m pretty sure the engagement on them isn’t coming from a bunch of practitioners in the middle of the trading day:

Most IVs you encounter use a 365-day model. To convert to a 251-day model (or any other tenor model) you multiply by the square root of the DTE ratio.

https://x.com/KrisAbdelmessih/thread/1921978763140628749

In the spirit of Thursday’s post and the tweet, I’ll pose 2 “interview-style” questions that can be answered in seconds. They require making a reasonable assumption. I’ll give the questions here, then I’ll post an assumption as a hint after the questions for those who need help forming one. The answers are at the end of the post. (Ignore cost of carry — also if you asked about that you’re way ahead of the game).

You do not need any calculators to answer these (just mental arithmetic).

❓#1: Volatility

It’s the close on Wed. Options expiring next Tuesday and next Friday have the same dirty vol (ie the same vol in your off-the-shelf 365d model). Does one of them have a higher clean vol? Explain. List any assumptions.

❓#2: Price of a straddle

It’s Friday close. Next Friday’s ATM straddle is $5. What price is the Friday ATM straddle expiring in 2 weeks to be the same clean vol? You do not need option calculator.

An assumption you could use to help:

We’ll use a calendar specification that states non-business days count as 50% of a business day for vacation time purposes.

If we continue to denominate our basic unit, a full trading day, as 1.0 and weekend days or holidays as .5 we get the following tenor:

251 x 1.0 + 114 * .5 = 308 day calendar.

Answers to the “interview” questions

#1: Tuesday has a higher clean vol than Friday.

Relative to a “dirty” year where each day is treated as equal vol, a “clean” year in which variances passes more slowly over a weekend, the Tuesday expiry has less time to expiry than Friday’s ratio of clean to dirty DTE.

If the model implies the same dirty vol for Tuesday and Friday, we can infer Tuesday must have the higher clean vol bc it has relatively less vol time vis a vis a dirty model.

#2: The second Friday straddle must be $7.07

The approximation for an ATF straddle is .8*S*σ*√t

Since there is no cost of carry we can assume ATF straddle = ATM straddle which is what the question asks about.

We don’t know S or σ but the question asserts the same dirty vol for both Fridays.

We know there’s twice as much time to expiry for the 2nd Friday and we know the earlier Friday straddle is $5 so the second Friday straddle can be computed as

$5 * √2 since there’s 2x as much time til expiry.

So the second Friday straddle = $7.07

💡This is one of those useful trader math ideas — for a given vol the price of the straddles only varies by square root of the ratio of DTE. One of those mental arithmetic things I found myself using constantly especially with short-dated options where you’re like “if the 1-week straddle is X, the 2 week is…”. This is also a clue to the degree to which I ditch the idea of “volatility” altogether in near-dated options and “think in straddles” and move sizes. This intuition is habitual but you can also see why it has theoretical support — vega p/l is a less of an influence on short-dated options. Results mostly come down to “how much did this thing move vs how it was priced”.

Returning to the question.

The $7.07 straddle is based on the same dirty vol.

But does that translate to the same clean vol the same as the first Friday straddle?

Yes — the ratio of dirty to clean DTE is the same for both expiries!

how I sold cotton at an all-time high

In the post If they ban short-selling derivatives become the underlying, I concluded with

If futures are trading at full carry and you think such a ban is possible it’s an asymmetric bet to put on conversions (ie short futures, long stock or for options short combos long stock).

Broadly speaking, the derivatives market becomes the underlying market. When the underlying market becomes encumbered do to technical frictions, derivatives are no longer just “derivatives”. The arbitrage mechanism is severed. The derivatives market becomes the home for price discovery.

I’ve seen this happen throughout my career. Just a few examples in addition to hard or impossible to borrow/short situations:

HYG becoming a liquid referendum on illiquid high-yield bond market
Cotton option synthetics continuing to trade even when the underlying futures are locked limit up (my trading claim to fame is selling the all-time nominal high in cotton in late 2011…it was an option synthetic around $2.25)
When a stock is halted, the premium or discount to any ETFs holding that stock computed using the halted stock’s “last” price before the halt implies the price the stock will open when it resumes trading. These situations were common during the dot-com heyday when stocks would be halted intra-day on pending news announcements.

Today, we are going to cover #2. Along the way I expect several bonus “oh that’s how that works” moments. And call my shot…you will enjoy this one.

When cotton options become the underlying

Cotton futures have peculiar price limit rules. The daily price limit depends on future’s price on the prior trading day.

If a cotton future is 70 cents, its daily price limit is 3 cents. If 2 or more of the first 5 delivery months close limit bid, then the limit is expanded by 1 penny. And this effect can chain every day with the limit capping at 7 cents regardless of the futures price.

Pricing a cotton straddle that cannot necessarily be delta-hedged (ie replicated) is good, clean fun. But it’s not the subject for today.

[Aside: When I traded cotton the rules were meaningfully different but I’m a bit foggy on them and can’t find the exact old rules online. The gist of it was there was a circuit breaker once you hit the limit that halted the market for 10 minutes perhaps. When the market re-opened the market was allowed an additional limit. If it hit that, you were double limit up. If 2 contracts went double limit up the market closed for the day. I once went to work for 20 minutes. The penny expansion rules were the same.]

Important background context

1️⃣I used ChatGPT to refresh me on the cotton planting cycle.

How does this affect futures trading?

The old crop contracts have delivery months before August. Notably: March, May, July.
The new crop contract , December reflects the next year’s harvest.

A squeeze in 2011 old crop acutely affected Dec 2010 thru July 2011.

2️⃣ In general, cotton futures, like most commodity futures are correlated to each other. In addition, nearer-dated futures are more volatile than further-dated futures. Even if you never traded commodity futures you’ve seen this in VIX futures. The front future can get to 70, the 6-month future has never done that.

The academic name for this “futures get more volatile as expiry approaches” is the Samuelson effect. The post Seasonal Volatility goes hard on this stuff if you care. But the point I want to get to is that cotton futures have a diminishing beta as you go out in time relative to the front month mostly because the volatility ratio drops off. Recall that beta is volatility ratio * correlation. However, when you cross a crop year, the correlation can also significantly drop off. In fact it can invert!

This is logical. If there is a shortage of cotton this year, farmers are incentivized to plant more acres next year increasing supply. Today’s bull market literally sows the seeds of tomorrow’s bear market.

This leads to interesting behaviors in future spreads. For example, if month 1 (M1) goes up 10 cents in a week, M2 might go up 6 cents. The spread expanded by 4 cents over the week. Generally speaking, spreads themselves are positively correlated with M1. (You can think of them as having their own delta to M1).

But when you get into these squeeze cases, the spreads can really blow out as their delta becomes dominated by the volatility of M1. If M1 is $1.00 and M2 is $.80 the spread is 20 cents. If M1 is squeezed up 25 cents in a week because there’s a shortage of deliverable supply into that expiry for logistical reasons it’s possible M2 barely moves. Suppose it doesn’t. The spread therefore has blown out from 20 cents to 45 cents, which is so wide it is more than half the price of M2 itself!

If you are long an option time spread you just got carried out. You are long vol on something that didn’t move and short the thing that roofed and because it can only go 25 cents in a week by hitting limits you just got gapped.

[When I joined Parallax, I worked with the CRO to come up with all kinds of limits for not just vol exposures at the ticker level but gross exposures in individual months because we understood that the problem in commodities is not just correlations that go to 1, but -1!]

3️⃣Here’s a crazy wrinkle — when the futures are limit up, the options still trade! The futures pit would be dead as the futures brokers would run over to the option pit to quote synthetics from the option market-makers.

The option market-makers were effectively making markets in the underlying. As you can imagine, liquidity disappeared as the adverse selection problem dominates. Markets that are 2% wide with 5 lots on the bid and offer.

But this is where I thought it would be fun to point out that there was a way to price that limit-up front month future.

Suppose M1 is March 2025 and it’s $1.30 limit bid and you need to make a market.

You can imply its fair value with 2 pieces of information.

Where a non-limit future is trading, say Dec 2025. Let’s say it’s market is $1.12-$1.13
Where the March/Dec futures spread is trading. Let’s say it’s $.20-$.21 (remember in commodities spreads are quoted as near month minus deferred month, the opposite of how equity roll conventions).

So “on legs” the March future is $1.32 bid, offered at $1.34.

If you offered the March future synthetically at $1.35 and got lifted you could buy the Dec future at $1.13 AND the Mar/Dec spread for $.21.

You outlaid $1.34, the Dec future cancels with the Dec leg of the spread and you are left long 1 March futures which you synthetically sold in the options market at $1.35.

The series of transactions has left you with a conversion on your books — short call, long put on the same strike which assures you sell a future at $1.35 at expiry that will wash with the future you bought at $1.34, netting you a 1 cent profit.

So when a future went limit up, you need to know where all the non-limit futures were trading as well as the matrix of spreads from those futures back to the ones that were limit to triangulate the tightest market in the limit future. You could quote synthetics around that value.

This is a screenshot of the actual spreadsheet I had on a tablet computer that I held in the pit:

If you can zoom in you’ll see the RTD links to my X-Trader software, ie TT

as well as the “beep” alert which was kinda hard to hear in the pit especially because I had a headset on to talk to our upstairs trader.

Oh, there’s another little Easter egg…see BAL? That was the cotton ETF.

I’ve talked a lot about how I used to trade USO vs CL and UNG vs NG but I also traded BAL vs cotton futures…this one was neat because there were only a few people in the markets who knew in real-time where the locked limit cotton future was trading AND also jumped thru the hoops to be able to trade SEC products like ETFs. But as I explained in Battle Scars As A Call Option — The UNG Experience I had set all that up when I was at Prime so we were uniquely capable of this.

Zoom in a bit more and you can see that we were also looking at JO (coffee ETF) vs KC (arabica futures).

So going back to this post’s baity title…I sold cotton futures synthetically at ~$2.25 which is a higher price than it ever settled IIRC. In fact, I don’t think the futures ever traded that high, it only happened via options. It wasn’t some genius move — I was hedging gamma at levels that seemed reasonable based on where the futures were on legs.

Hedging gamma?

I might as well explain how I traded cotton. I went to the cotton pits specifically because the risk manager at Prime mentioned that there was a crazy squeeze happening in the cotton markets. Apparently, a large grower with cotton in the ground, who did the sensible thing — hedge the crop with short futures — was caught in a liquidity crunch. As the futures rallied in late 2010 based on the short supply, the large hedges were marking against the grower.

This was in the wake of the GFC so credit was tight. The grower couldn’t borrow to fund the margin despite having the crops. Classic example of the market smelling blood and “running in” a large exposed party.

I went to the pit to participate in these reindeer games. The option market was wild. Vol roofed. Call skew was jacked to record highs. Limit up every day.

So what did I do?

My trading partner will give me tons of credit for basically doing the thing many of the other options traders wouldn’t…I bought as many calls as my risk could let me. High vol, big riskies to the call…gobble gobble. But I did it because it was only positional trade that made sense (you could make some money market-making every day but this seemed like a chance to score on a position trade).

My reasoning:

There was no sensible path where we just grind higher. This thing gets to the uncle point as sloppy as the limits allow then crashes. We don’t know where the uncle point is but it’s up from here. By buying expensive calls with big fat deltas, I could sell lots of futures. Hard deltas. Those option deltas are soft…they will melt away as vol declines and time passes but the position I want is to be directionally short but have my upside protected.

And sure enough it played out pretty much that way, straight line up, straight line down (with the limits as speed bumps all along the way).

We played until maybe March expiry. The episode was over and I had no intention of being a cotton trader for the long haul.

Luckily, coffee was just getting busy and its pit was about 30 feet away. So I sauntered over there…for the next year.

Story time

By late 2011 I was talking to Parallax and once it was clear I was moving to SF, I started “closing only” trades with Prime. Winding down an option book is expensive. You can work out of your positions but it takes months depending on how far out your longest-dated inventory resides. Since you aren’t adding any trades with edge, it’s basically just hedging and losing money. The other choice you have is to pay another trader say a couple cents a contract to take over your book. Either way there’s an exit tax.

Once my book got very small, I was just playing Gears of War until 4am with my west coast friends every night, waking up at noon, and dabbling on my first blog, shoxland, named after my NYMEX badge — SHOX. There’s still a lot of people out there that only know me by that name and even many who know my real name who still call me that. (My XBOX live gamertag is “shox da monkey” a nickname a trader named Keith gave me in the office.)

Anyway, all the pit-hopping experience made me familiar enough with these products that when they got interesting they became part of my options book at Parallax. When they were dull, I ignored them.

Ok, ok thanks for indulging story time. It’s legit nice to write this down. For some readers out there this will be a trip down memory lane.