The stampede of one hundred horses
Vol #223: April 17th, 2025
I asked three different AI models to help me with an age-old internet question - “Would you rather fight 100 duck sized horses, or 1 horse sized duck?”
ChatGPT, Claude, and Grok unanimously chose the duck-sized horses. The reasoning sounds like the same training data spun with a personality branded thesaurus. Each cite the weak (“laugable”) kicking power of the tiny horses, and called the horse sized duck a “real menace”, “truly terrifying”, and “a prehistoric level threat”.
The original question was actually posed offline, from Mr. Brannick of Manchester, UK to his local paper’s readership in 2003. “My friends and I would like people’s views on who would be victorious in a fight between 100 duck sized horses and one horse sized duck. I go for the team effort, but am beginning to doubt my choice.”
There are many hypothetical questions here, and a lot of room for interpretation. The wealth of human ingenuity doesn’t disappoint. The integrity of a horse sized duck’s skeletal structure has been thoroughly questioned. The Science Museum of Western Virginia argues that similar animals existed during prehistoric times. Either way, there’s lots of kicking and some very large beaks.
The best answer I’ve read so far has been New York Times columnist Nicholas Kristoff.
“Definitely one horse-sized duck. Then I'd distract it with some cracked corn and, as it gobbled it up, I'd jump on its back and take it for a flight.”
Whether or not I had some duck feed handy, I too would prefer to manage one adversary than many small ones. Never mind the fact that horses are much smarter animals than ducks; no matter how small they are, every single horse presents a different vector of threat.
The paradox here is how threat level scales proportionally to size. Room for discussion exists because it doesn’t, and other than a handful of Jurassic era specialists, we have no idea what kind of havoc a horse sized duck could wreak.
As an options market maker, every order is a threat. As immediate compensation you receive a positive mark from the difference between your trade price and the midpoint, and possibly some transaction fee credit if the order is posted on an exchange with maker/taker pricing. Now the risk is yours.
That change in open interest - be it opening, closing, or rolling - is a threat because it brings new information to the market. The price discovery process is continuously in motion thanks to those who both take and make liquidity. Whether you sold low or bought high doesn’t matter, your position is simply a casualty in the search for fair value.
While most of the keyboard warriors in the forums (and thus the LLM imitation engines) choose the tiny horses as foe, every single market maker I know would rather have one big duck of an order. Apparently foreign policy opinion columnists got the memo too.
A hundred one lots, do not equal one hundred lot. Nor do ten ten lots. One big block is much less threatening.
For as long as I’ve been in the trading business there’s been talk about how the exchange floors are going away. The early 2000s saw the best trainees sent upstairs to trade the all electronic ISE exchange. Firms increasingly abandoned their specialist posts, and orderflow was migrating to instant messages. Crossing mechanisms were also developed that promised to handle large orders.
Yet as OCC volumes boom (a record 102M contracts on April 4th!) the floor has become more important than ever. For participants that have horse sized orders, there is no better way to get a fill. A broker might be an adversary, but he can contextualize and facilitate executions, doing everything he can to be a big duck and not small horses.
Electronic orderbooks are fabulous for rapidly and transparently disseminating all of the trade and order information. This allows for many more strike and expirations to be listed, and more underlying products to have options listed. It also makes options much more accessible.
While this is all very good for liquidity in the big picture, it does have externalities. When you combine the fact that capital is fragmented across both venues and securities, with algorithms that are reading every single trade detail and pricing it against a host of other factors, the top of the book becomes jittery.
When a single small order comes in, either in relation to the top of book or the typical size in the product, it will typically trade without making a dent in the market implied pricing. If it’s retail tied to payment, it will likely be improved. Deltas could be completely absorbed and not passed through to the underlying if the order is small enough.
Mechanically speaking, the risk has been transferred. The liquidity providing counterparty now has long or short options at a given price level, taking up some portion of their risk budget. They, and everyone else who reads the OPRA feed, will be just a little bit less likely to buy and sell at that level again. (Unless something is distinctly mispriced.)
As liquidity absorbs the orderflow, prices shift. Each of these events creates a trigger to potentially adjust and rebalance whether it's a one lot or ten lot. A trade is a trade, with information behind it. The one lot buyer might only have a few thousand dollars in their brokerage account, but they are a sentient trader with an opinion.
When multiple orders contain the same signal, that becomes confirmation of a shift in value, be it volatility, rates, or skew. Regardless of their unique sizes, several nudges in the same direction drives conviction.
The large order is just one voice. It might be well capitalized, but that comes with its own constraints. Many of the largest pools of capital operate within fairly constrictive guidelines about when and how they trade. That in fact reduces the information content of a thousand lot.
Execution is a trilemma between price, size, and time. Pick two of the three. For a good price you have to be patient, and on a quick fill you won’t get much size. The daily waltz between broker and dealer is each side trying to squeeze out just a little bit more.
When we consider the time dimension of price formation, the information quotient of order count becomes even more stark. While markets are built to process all of the updated points of light, even in a world of AI agents and complex algorithms that can’t happen immediately. The response cascades from a thousand if/then statements are prone to over-reaction. Activity happening in repeated succession is a far stronger signal than dollars.
The search for value is a continuous process, but it is far from evenly distributed. A very large order will be analyzed for toxicity, compared to the current price levels, and quoted all at once. Once filled, the markets can adjust and reprice for the next inbound piece of information.
The biggest threat of the duck sized horses is that they can come at you all at once. In a kung-fu movie setup where each villain waits their turn, the miniature equines would be very manageable. For the market maker who is getting dinged on downside here, upside there, all in rapid succession, the only course of self defense is to rapidly adjust price.
Markets adjust slower than FIX messages get processed, and when the race condition of orderflow outpaces the digestion of the same, it's a gastrointestinal event. All the small orders could amount to nothing, but you won’t know that until it’s over.
No matter where the orderflow comes from, whether its clydesdales or shetland ponies, the same process will happen. Liquidity providers aren’t in the business of ignoring signals, no matter how small they are.