Bringing Out the Bots
Can’t we just automate that?
Auto-X is powerful. The potential to do something hands free. Except for the auto-mobile. That requires quite a bit of attention still, despite best efforts.
Auto trading is enticing because trading is hard. Here’s an idea, now go make some money. A couple conditions, simple pattern recognition - ring the cash register. But ask any quant developer and they’ll tell you it’s a full time effort to run the machines.
Investors with longer horizons are demonstrably keen to hand off their jobs too. A diversified ETF solves many allocation pains in a few clicks, and life cycle rebalancing in target date funds makes it even easier. Lucky for most that works.
Options investors have to keep on clicking. The beauty of a derivatives strategy is nipping and tucking returns through contracts that expire and offset. That’s a process, not a checkbox. It requires regular pruning, adjustments, pauses, and realignments as the market’s seasons change.
If you want to harness the full power of options, you need the systematic rigor that algorithms deliver, combined with a conductor that guides a symphony through each crescendo.
The blend of man and machine is a spectrum where different operators are willing to hand off different parts of the job. But every single trading and investing problem I’ve tackled with options has required some combination of these tools. The scientist’s rigor and romantic’s aesthetic.
Trading pits are a tactile, olfactory rich, uniquely human melee. But they have always been an innovation forge where individuals extended their wits through electronic interfaces. It didn’t take long after the CBOE listed puts for someone to program a TI calculator with Black, Merton, and Schole’s famous formula. Fast forward twenty five years and there were still piles of big money tickets jammed between cathode ray tube monitors. Scream and elbow in the crowd - then punch it into the system.
Once I’d been trusted not to fubar lunch runs, my first real job as a clerk was scanning the log files of AQTOR. AQ stands for auto-quoting of course, the rest some dear reader surely knows.
Market makers exist to provide liquidity via their bids and offers. Quotes with price and size. Verbally in open outcry, via instant message, or gatling gunning up sixteen exchange APIs. While the first two clearly require a human being, electronic liquidity was ripe for the machines to take over.
This fancy Swiss application was designed to do the heavy lifting of spinning out all those messages into the newly developed electronic platforms. Most of them without anyone lifting a finger.
If you want to package the dealer's job into an .exe, the program must perform two primary functions. It needs a fair value for the security in question, and a confidence interval about how to price the risk of trading that in either direction.
Each of these functions can be further broken down into components, and assessed for automation potential. Run through each of the greeks (i.e. pricing sensitivities) and delta is easy to adjust mechanically as stock moves. The price changes as time passes via theta, and clocks are also easy to read.
Things get more difficult as we move into the volatility space. Reverse engineering the market price given other assumptions is where most traders, even market makers, can stop. That’s the implied part of IV. If you want to automatically make changes to that, you need to either keep solving the market price, or have your own set of signals and logic to disagree and change. Easier said than done.
A good liquidity provider is using some systematic adjustments based on relationships between other options or trading pattern recognition. But even with all the GPUs that NVDA can pump out, it takes a lot of if/then statements to consider the nuanced soup of factors that a human does. There’s my kind of nerd behind every dozen monitors.
Coming up with a price for the option can be automated to whatever degree you’re comfortable with. As does solving function number two - coming up with the price for risk. A quote is a buy AND sell level, so in addition to pricing an option value, a spread must be added to fund the cost of liquidity. Managing, hedging, paying port fees to be there, and most of all accounting for error.
Again we can lean on the market to provide a good baseline price (width), but undifferentiated competition does not stay in the game very long. The list of potential datapoints to use in coming up with the perfect quote is endless. A small count of them will get you a majority of the way there, but when to listen to which ones quickly gets subjective.
I learned to market make by letting the tools do the work they’re best at, and focusing where a human adds value, and can put the pieces together. Nudging that spread just a bit tighter or wider. When orders spend longer processing, or there’s a funny new quoting behavior, it takes a subjective skill to adapt and realign.
Now as an advisor, I am using this same playbook to capture opportunity. Churning through market data to derive indicators and track spread values over time is good work for silicon. Disciplined order entry is perfect for an algorithm. Let the bots handle closing positions so I can focus on meaningful conversations and data analysis.
Harvested Financial was founded to bring the power of derivatives to everyone. I think they’re really cool! But I also understand that the level of wonkiness to appropriately trade and manage these is tedious. There’s the operational pitfalls of regular execution and managing assignments, dividends, and rolls. Adjusting strategies or issue selections for different market environments doesn’t make it any easier.
Options for all requires a lot of solves that rot their value proposition when packaged up for ease. My thoughts on the ETF wrapper for this can be succinctly summarized as grocery store sushi. But automation and systematic opportunity capture in equity options are important tools for retail and professional traders alike.
If we combine a human knack for smelling when something is fishy, with a trading algorithm that does the heavy lifting, we have something powerful. I want to call it my newest offering, but it’s exactly how I was taught to trade.
This combination is why I’m partnering with PeakBot to integrate their trading tools into my portfolio management and advisory practice. Clients will now have access to a set of strategies that are automatically traded, but hand curated according to different objectives and environments.
One of the most tempting strategies for buy side options traders is selling iron condors. Placid realized volatility most of the time lulls you into premium slumber. But volatility isn’t always a good sale. The persistence of VRP isn’t neatly captured by continuously whacking a four legged spread at some predetermined interval.
Sometimes Jerome Powell gets handed the mic. Sometimes the vol is just too damn high. My grizzled take on the sirens of fat premium is that they’re never sweet enough to account for the cliffs below.
The wheel also spins a lucrative narrative as a way to turbo charge your stock picks. That can very well be true – just not always or everywhere. The choice of names matters, and even raging momentum might not compensate you enough for the left tail. When to hold them and fold them is a difficult question without an easily programmable answer.
Because the dealers fight so hard to price and provide liquidity, there’s little worry about whether the spread is fair. Retail traders capture opportunity by choosing when to implement different strategies, and selecting the best issues for them. Those two levers are the source logic to manage a basket of algorithmic options bots. Hands on the wheel of a powerful set of tools.
It took our AQTOR log files ten minutes to open up because they were so verbose. The most common log lines by far were the price updates. Every tick was captured – until there was a two minute gap when the machine bricked on a stock gap, and options needing price updates queued into oblivion.
All of this was happening as YKJ chin wagged with his specialist. Few and far between were the most meaningful updates – a vol curve shift or a rate adjustment. As the rev/con got bounced back and forth between the CBOE and the AMEX, gauging just how negative to move the NYX borrow was pure art. What you learned off screen.
For all the orders and trades a bot makes, the human operator has very few changes. But those are the ones that matter. Watching short theta decay waiting for an exit is painful. You’ll find me spending my time reading between the lines of a watchlist, gently nudging the allocations towards opportunity.