Markets Without Reset

By Kareem Harras, Head of MENA‍

Infrastructure carries opinions. Every system a broker runs - the aggregation layer, the monitoring, the capacity agreement, the bridge - encodes beliefs about how markets behave. Nobody writes them down. They are built in by whoever designed the thing, in whatever year that designer learned what normal meant.

For some time now we have been quietly reshaping our own systems as those beliefs expired. This year the market began sorting providers by how much of that work they had already done.

The deepest belief of all was too safe to even notice: the market stops. And while markets genuinely did stop, the stopping itself performed real work, at no cost to anyone. The Friday close froze every book for forty-eight hours, so risk paused on a schedule. The weekend was a maintenance window, so anything that creaked could be repaired while nothing traded. The Monday open concentrated all repricing into one moment every desk was staffed for — violent, but violent by appointment. That is the honest reason the old regime felt easier: part of every firm's risk management, and part of its engineering, was done by the calendar, free of charge.

Three things changed: the market's character, the flow's character, and the economics underneath both. 

The market's character changed. The pauses are being retired instrument by instrument - continuous CFDs across the industry, CME's around-the-clock gold and oil, LSE 24 on the way - so books that once flattened on a schedule now carry risk through every hour. And volatility stopped subsiding: the World Gold Council found gold's volatility reaching the top fifth percentile of readings since 1971. That single fact invalidates a design decision sitting inside most infrastructure - the assumption that stress is a mode you enter and leave, so a system only has to survive it briefly. When the exception becomes the resident condition, anything that degrades under load degrades daily. 

The flow's character changed. Clients used to spread themselves across a broker's instrument list. Now they crowd - most of the book's real exposure sits in one instrument at a time, and which instrument it is keeps changing: metals this quarter, energy or indices the next. A broker ends up hedging what is effectively one large position, and by the time depth has been sourced for it, the crowd has moved somewhere else. 

The economics changed too. Resilience is expensive. Credit lines, spare hedging capacity, redundant systems, staffed weekends - these are what allow a provider to keep quoting when conditions turn, and they all cost money before they ever earn any. The problem is that years of competing on price have left providers with very little margin per trade to pay for them. So when a feed is unusually cheap, it is worth asking what was left out to make it cheap - the spare capacity, the redundancy, the weekend cover. The discount and the fragility tend to have the same source. 

Capacity has a second problem. Hedging demand does not grow steadily; it arrives in jumps - quiet for months, then several times normal size inside a single session. An arrangement sized for the quiet months breaks at the peak, and the broker usually learns this mid-event, in the form of revised terms: wider spreads, lower limits, higher margin requirements, all arriving at the moment the book most needs the old ones.

Technical debt, in a market that never closes 

Engineering has a precise term for an assumption that stays in the system after reality has falsified it: technical debt. That is all technical debt ever is - yesterday's beliefs, still operating.

While the market kept its pauses, that debt was cheap to carry, because the pauses paid the interest on it. Nothing traded over the weekend, so nothing ever had to be repaired while it was running. No position could move for two days, so gaps in monitoring cost nothing for two days out of seven. And all repricing waited for the Monday open, so it arrived while every desk was staffed and watching.

Withdraw that payment and the cost lands on the firm. The conditions that expose deferred work now run continuously, and there is no quiet hour in which to do it. Repayment falls due at the least convenient moment - usually mid-event, in front of clients.

Which is the argument for paying it down early. We have spent years doing exactly that, retiring our own assumptions before the market got to them. Depth that stays on the book when a pricing source degrades, with routing that moves flow to alternatives in milliseconds. Surveillance that never assumes a weekend. Capacity terms built to hold across every session the market trades. Each was a decision made while the case for it was still theoretical, which is the only time such decisions are cheap. 

This year put those choices to the test in public, and our spreads held, our quotes stayed available, and our fills landed - through a Fed decision, an exchange outage and the worst day gold has had since 1983. That is what infrastructure built for this regime looks like when the regime arrives. 
‍

One question

If you are reviewing your arrangements this autumn, put one question to every provider you speak with:

How much of your infrastructure still assumes the market stops?

The answers will sort your shortlist faster than any pitch. We will be answering it ourselves, at Forex Expo Dubai in September.