Why we stay small

Ask the industry how serious a trading firm is and it will answer with numbers. Headcount. Capital under management. Petabytes of data on licence. Desks, and how many asset classes they cover between them. The reflex that follows is always the same one: to get better, get bigger. That usually means another team of researchers hired, another data feed licensed, and before long a desk standing in a market you were not in yesterday, on the assumption that more of everything eventually adds up to more skill.

We do not believe that, and we built the firm on the opposite bet.

Orcas is proprietary and systematic. We trade the firm’s own capital in liquid markets around the world. We take no outside money, and we mean to stay small on purpose. Not small as a phase on the way to something respectable. Small as the actual plan. When people hear that, they usually assume it is a polite way of saying we have not raised yet, or that we are amateurs who will grow up once the cheque clears. It is neither. Size is a choice, and for the work we do we think most of the industry has made it backwards. The case for being big is real, and I want to put it fairly before I argue with it.

What size actually buys

A large book with a lot of capital behind it can run strategies a small one simply cannot afford. Some edges are capacity-hungry: they only clear their fixed costs at scale, or only pay when you can warehouse a lot of risk and sit on it. If your model needs a billion dollars of exposure before it means anything, then you need a billion dollars.

Coverage is real too. Fifty desks across equities, rates, commodities and credit spread the firm’s risk over a lot of loosely related bets, so no single bad regime takes the whole thing down. When one book is having a miserable year, another is quietly carrying the P&L. That is plain arithmetic.

And specialisation genuinely lets a person go deep. Give someone nothing to think about but Japanese government bond microstructure for five years and they will know things a generalist never will. Throughput sits on top: more researchers run more experiments at once, and more survive to trade.

This is a real machine and it works. The big multi-strategy firms are not the product of some collective confusion. They are extremely good at a particular game, and that game has physics of its own. I am not going to pretend the machine is broken, only that it is a different one, running on assumptions that quietly stop holding at our size.

The specialisation tax

Here is where the physics change. Every specialist you add is also a boundary you add. The researcher hands a signal to the engineer. The engineer turns it into a running system and passes that to the trader, who takes the position; risk inherits whatever comes out the far end. Every one of those arrows is a handoff, and each handoff is a place where meaning leaks out.

What leaks is context. When the researcher found the effect, they knew a hundred small things about it that never made it onto the page: which regimes it fell apart in, and the ugly fortnight of March data they had quietly thrown out. The engineer picks up the ticket with none of that and builds exactly what it says, which is not quite what was meant.

You can be precise about why this gets worse the longer the chain runs. Treat the researcher’s real understanding as a signal and the written spec as a lossy encoding of it. The engineer never gets the understanding, only the encoding, and rebuilds their own version of it against different assumptions, so error creeps in. They encode again for the trader, who rebuilds again, and the errors do not add up, they multiply, because each stage faithfully decodes the previous stage’s mistakes as intent. It is the photocopy-of-a-photocopy problem, or a message whispered down a line of children: what comes out the far end is stated with total confidence and is quietly wrong, and nobody in the middle did anything visibly wrong. A tidy information-theoretic account of exactly this exists, and I keep meaning to write it up properly, but this note is not the place for that detour.

The point for us is plain. We keep the whole loop - idea, backtest, live trading, risk - inside a very small number of heads. The person who found the effect is the one who puts it into production and watches it trade. Nothing gets translated across a boundary, because there is no boundary to cross. The context never has to be written down and shipped, because it never leaves the room. Depth bought through specialisation is a tax large firms pay in lost context, and it is one we would rather not pay at all.

Every hire is also a cost

There is a second thing people forget when they tot up the benefits of headcount. Every person you add buys throughput and charges you in communication. Two people need one conversation to stay aligned. Ten people need forty-five. The links between people grow much faster than the people do, and past some point most of a day goes into keeping everyone pointed the same way instead of into the actual work.

You feel it in the meetings that exist only to prepare for other meetings. It shows up as decisions that need four people in a room and therefore slip from now to next Thursday. Past a certain size the firm starts spending a striking amount of its energy just keeping itself coordinated, rather than on the markets it is meant to be trading. That is not anybody being lazy. It is just what a decision costs once it has to travel through more heads before it can be made.

At our size a decision is often one person thinking hard for an afternoon and then acting. When a market does something strange on a Tuesday, we can change our minds by Tuesday afternoon. A large firm with the very same insight has to route it through its process, and by the time that clears the opportunity has usually gone. Being able to decide quickly is an edge in itself, and it is one of the few that gets worse rather than better as you grow.

The edges that do not scale

This is the part that matters most, and the hardest to wriggle out of, because it has nothing to do with culture or speed. It is arithmetic, and it runs the wrong way for a large book.

When you buy or sell, you push the price against yourself. Buy a little and the market barely notices. Buy a lot and you walk the price up as you go, so your average fill comes in worse than the price you saw when you decided to trade. That cost is market impact, and it grows with the size you push through. A large book cannot switch it off: to put serious capital to work it has to trade in serious size, and that size moves the very market it is trying to trade against.

Now set that beside how much edge there actually is to capture. Most edges are shallow. A mispricing might be worth a handful of basis points, with room for only a few million dollars before it closes up. For a small book a few million dollars is a real position and those basis points are a real return. For a book that has to deploy billions, the same opportunity is a rounding error it could not fill without paying away the whole edge in impact.

So a small book lives somewhere a large one cannot physically reach. We can take capacity-constrained trades that are simply too small to be worth a giant’s time. We can sit patiently in a position for weeks that a large book could never build without telling everyone watching the tape exactly what it was doing. The edge per dollar is highest precisely where the dollars are fewest, and that is where we choose to fish. That is the reason we are small.

Whoever forms the idea owns it

Being small changes what a person is actually on the hook for. When the same few people carry an idea the whole way from first hypothesis to live position, there is nowhere for responsibility to hide. The person who believed the effect was real is the person whose number turns red when it is not, and the one whose phone goes off at three in the morning when the position does something it was never supposed to.

In a long chain, responsibility spreads out and thins. The researcher can suspect the engineer misbuilt it, the engineer can suspect the spec was wrong, the trader can suspect risk sized it badly, and each of them can be a little bit right, which means none of them has to be wholly wrong. That is comfortable, and the comfort is the problem. You do not fight for an idea you are able to walk away from.

When you own the whole thing, the incentive only points one way. You lose interest in ideas that merely look clever in a backtest and gain interest in the ones you would stake your own name and your own sleep on. It changes what you pick to work on before you have written a line of code. Skin in the game is the org chart here: there is no wall to throw a model over, because whoever built it is standing on both sides of the wall.

What this costs us

I would be lying if I made staying small sound like a free lunch, so here is the bill.

We say no far more often than we say yes. Whole markets, whole families of strategy, are off the table for us, not because they are bad but because they would need more hands or more capital than we are willing to carry. Every capacity-hungry edge a large firm can harvest, we watch sail past. That is a great deal of money we choose, deliberately, not to go after.

We carry real key-person risk, and I am not going to dress it up as anything else. When the whole loop lives inside a few heads, those heads are the firm. If one of us walks under a bus, there is no bench of ten interchangeable people to slot into the gap. A large firm barely breaks stride when someone leaves. We would feel it for a long time.

And there is no cavalry. When something breaks at two in the morning there is no overnight desk in another time zone to hand it to, no room full of engineers to throw at the problem until it surrenders. There is us, and whoever happens to be awake. We accept an operational fragility that a bigger firm simply has the bodies to insure against. I do not want to give the impression that we have priced these costs and shrugged them off. We feel them, and we have decided they are worth what they buy us.

Small is the strategy

So this is not a waiting room. We are not a small firm rehearsing to be a large one, biding our time until we are allowed to raise capital and hire an army like everyone else. Do that and we would be trading away the one thing that makes us worth anything, in exchange for a machine that thousands of people already run far better than we ever could.

We intend to still be here a very long time from now, compounding through conditions we cannot yet picture and on venues that do not exist yet. A big firm is betting it can out-resource whatever the future turns out to be. Our bet is the opposite one: that nothing profitable stays profitable for long, and that you outlast that churn by staying light enough to step out of its path, while a heavier firm is still building the muscle to try to absorb it. Whatever is still trading on the far side of all that is likelier to be a handful of people who own what they do and can change their minds by Tuesday afternoon.