Drops of Process
Vol #128: May 11th, 2022
Welcome to all the new readers this week. Thanks for coming along, and the best way to show your support is a share.
For old readers, sorry I missed you last week. I’ve been heads down on a new project that I’ll be opening up next week. Stay tuned.
Cheval Blanc makes some of the best wine in the world.
Their blend of Cabernet Franc and Merlot, with just a dash of Cabernet Sauvignon drives consumers and collectors mad. Perfectly manicured vineyards are designated by simple white markers with their name in red paint. Even that designation is a rarity for the area - Petrus is across the road and doesn’t even have their name on the building.
If you watched the classic wine film Sideways1, you know this winery as the producer of Miles’ (Paul Giammati) grail bottle - from the 1961 vintage. The winking irony of the movie is that he spends the whole trip blasting merlot, yet it’s 43% of his coveted juice.
Wine tasting is hard. I’ve watched sommeliers who just finished telling you what the weather in Pessac Leognan was like in August of 1986, go on to confuse prime examples of left and right bank Bordeaux (Cabernet Sauvignon vs. Merlot). Don’t worry too much if your palate prefers the stuff with a kangaroo on the bottle. That’s good too.
Sommeliers go through a fairly structured process to identify wine. It’s not so much an “a-ha” moment (unless you are very familiar with the wine), as the result of a step by step breakdown of the flavors, and pattern matching with known profiles. By checking off scents and chewing tannin structures, you narrow the range of possibilities until you can make an educated guess.
As all magicians know, there’s a lot that goes on behind the scenes. While the swirl of the glass opens up the wine and long pensive inhales savor the tertiary notes, they also serve to provide the mental space required for analysis. Nailing a region and grape, or even chateau and vintage is a feat learned through much training and structured observation.
Watching a brilliant trader manage their book has the same effect. As each decision comes down the pike, bids are placed, vols are nudged and rates tweaked. Why two clicks and not three? Why 500 and not 1000? Why did you pass on that ticket?
Throughout the day stocks zig and zag, theta decays, and edge is collected. If an experienced market maker is cold blooded enough to let you see their intraday pnl, many days it just looks like a cash register slowly ringing up revenue. Each of those decisions seem so arbitrary, where are the commas coming from? (It’s not all roses in the garden of structural alpha, other days it’s a bloodbath.)
As much as any craft, trading requires process. To make decisions under uncertainty - like a blind tasting - you need a framework to systematically evaluate the facts and juxtapose them against the zeitgeist. Science with a dash of art.
Process oriented folk start every day the same way. My first job as an intern on the floor was learning which sheets each of our market makers wanted. People traded 25 names then, not 1200, so they could reasonably get a stack of print outs for each of their positions. Our proprietary database was DOS like software that would generate greek summaries in many different templates. YKW got FATs, YKS preferred BRIEFs, and YKJ wanted STRIPEs.
These position reports were delivered each morning about thirty minutes before the bell. Rumor had it a clerk in San Francisco once got fired for stapling the opposite corner and handing it to someone on the wrong side of gap risk. Don’t upset creatures of habit. And in their defense, the bell rings at 6:30AM on the P-Coast.
When you sell liquidity as a service, a fundamental part of the job is evaluating how much you want to provide at what price. For the electronic side of that, it’s a matter of configuring settings like display size or which NBBOs you want to join. Experience with a product or one like it is why the seasoned vet sets a value of 10 instead of 20.
Electronic trades are about positioning yourself nimbly and reacting to the activity. For voice trades, there’s more of a negotiation that takes place, which lets the liquidity provider sniff and swirl the order. What looks like a snap decision is the result of a high octane evaluation of all the points of light surrounding the trade.
You must think about what the fair price is for the trade (given that the customer has an order), and also what the right premium to charge is for warehousing that inventory. Fair value is a function of the implied volatility and rates; where have those been trending and which way is this order going?
Relative value is an important metric to lean on. Ideally before the quote comes in, you’ve checked your time spreads and lined up butterflies. You’ll know that April has a kink in the term structure because of earnings. Tidying these loose ends is the disciplined duty for dull moments.
Size is important too, how does this package compare to overall open interest and historical activity in the name? Assuming I want to get on the ticket, what percent of it can I afford to be? Is it better to be all or part of it? The answer depends partly on who the broker is.
Once you have a trade on, managing the book obeys its own process. Delta hedging is the primary maintenance operation for market makers. Anything that moves in time or space is going to change your delta, so there’s always something to adjust. Even with a small portfolio, there are so many other decisions to focus on, that the best delta management is sticking to tried and true rules of thumb.
Having a process where you hedge deltas at noon and the close frees up the other 95% of your trading day from having to worry about that. A system that hedges your backspreads at 1x gamma and your front spreads at 4x gamma (kidding2) means no mental energy is spent second guessing whether to hold on for one more day.
If the process is so important, why can’t a computer do this?
It can and it does, pretty well. There are a lot of tasks that have been automated by scripts and algorithms; everything from display size on quotes to bots that respond to sales traders. But like the conversation about AI, the real edge comes from knowing the system so well, that you can see where it breaks.
You can’t break the system unless you’re deeply familiar with it. That’s the parable of the jade master. By seeing thousands of reps, you’re able to hit the fat pitch when it comes at game time. For risk takers, the only thing that prepares you for uncertainty is having stomached a lot of uncertainty.
It’s easy to think something is special. A lot of the time it’s not. Coincidences are just realized probabilities. In a room of 75 people there’s a 99.9% chance of two people having the same birthday. Not every upside call buyer has inside information.
One of my great managerial feats after I moved off the floor full time, was building a system that allocated opportunity algorithmically. As a firm we wanted to balance our exposure for risk management and tragedy of the commons reasons, but individual portfolio managers all wanted to quote the best names. This was a highly politicized process, that became even more politicized once you tried changing it.
I negotiated benchmarks for a set of indicators that would allow our footprint to rise and fall with activity. But to spice things up, we created a wheel that introduced an element of randomness to every marginal quoter add. Market makers could bid on which names they wanted, where the winner wasn’t the highest bidder, but the result of a random spin with your bid size representing a proportional chance of the total overall bids.
Even for people who trade tens of thousands of options contacts per day, the randomness was maddening. There was interesting game theory to setting your bids. You could groan about overpaying for a dud, or rejoice in winning the biotech IPO with just one point. Whenever there was a long shot win, I had a standing joke with one of the particularly observant participants that “one percent happens”.
When the wheel spins for 20 names a day, one percent should happen every week. If you didn’t have a plainly coded algorithm doing Vanna White’s job, it would be easy to believe in conspiracy theories. Some people still did.
A variant of this would be what trader educators and risk managers say to young bucks first strapping into their monitors; “process not results.” Ultimately results pay the bills, but they do have a certain amount of noise in them. If you have a solid process, then you don’t have to worry about each specific iteration. One percent will happen one percent of the time, so just make sure you sell it for two percent and don’t sweat the outcome.
Process also gets you partial credit. If the freshly minted badge is doing everything right but takes a gap hit on his second week, a red month doesn’t kick him to the curb. It’s better to have those cash register days, but if you’re sweeping the floor and making tea, eventually they will happen.
Trading highlights all the little edge cases that exist in our world. If there’s a loophole or unexpected outcome, finance will find a way. Having a system to make decisions gives you confidence about the ones that fit the mold, and partial credit or defensible logic for those that don’t.
There’s a bistro in Paris run by a former “best sommelier in the world.” They strongly encourage a blind pairing with the meal, and it’s a fun parlor game for the dining table.
As the first glasses come out and I pull a tentative sip, the chorus echoes: KIMMERIDGIAN!
This was a softball lobbed over home plate. The characteristic soil that is rich in oyster shells from the Jurassic age sings loudly through all that is Chablis. Harmonized with light citrus and exhibiting the tinge of green at the edge, I was already banking points.
Chablis is of course made with chardonnay. The wine I was tasting however, was not. Unable to even get the grape correct, my tail was fully between my legs. From out of left field, this was a pinot noir grape, also grown in the Yonne valley and irregularly used to produce a white wine. The terroir was coming through so strongly that I never even considered it would not be classic Chablis.
At the end of the day, I gave myself partial credit. My process was sound. I would make that trade again.
The post’s title derives from a contemporary wine show, “Drops of God” on Apple TV. Highly recommended. Hit me with a reply if you’re watching too.
A frontspread means you’re short gamma, and a backspread means you’re long gamma; the terminology comes from how the PnL graph looks. The delicate balance of hedging either position is between doing too much and too little. If you don’t hedge a backspread and stock drifts back, you’ve missed an opportunity to pay the theta bill. Over hedging a frontspread is buying high and selling low. Letting a frontspread run to 4x your gamma however, would be taking a pretty big swing at a large unrealized loss reversing its course.