First time I made a decision to move into e-trading was once I was working at Lehman Brothers on the e-procurement facet. You possibly can see the ecommerce groups beginning to develop and the emotions on the buying and selling desks began to vary in how liquidity was managed. My selections within the e-trading area the place all the time constructed on expertise regardless that the mechanics of the trades between buyside and sellside the place relationship primarily based. Some within the e-trading area nonetheless argue relationships are key to liquidity…however are they?
Within the noughties a seller knew which company treasurer would commerce on the London repair, which macro fund would present measurement into New York, and which hedge fund would pull liquidity the second volatility picked up. Costs had been shaped as a lot by relationships and judgement as by provide and demand. Away from the buying and selling facet the precise equipment of which buying and selling expertise was chosen from an array of distributors was additionally typically dictated by relationships between consumer and vendor.
As we enter 2026 I feel you’ll be able to argue that world has largely disappeared or is within the technique of disappearing. FX liquidity is not negotiated; it’s parametrised. It arrives as streams, curves, confidence scores and reject chances. What used to reside in a seller’s head now lives in configuration information and datasets.
This shift from voice to code is not only a narrative about automation. It’s a story about how FX liquidity itself was redefined. From human discretion to machine-readable parameters and the way that change reshaped buying and selling behaviour, market construction, information economics and the regulatory setting.
The voice market: liquidity as judgement
Within the voice-driven FX market, liquidity was inherently contextual.
A quoted value relied on:
- Who was asking
- How a lot they needed to commerce
- Why they had been buying and selling
- After they had been more likely to come again
A seller may widen a value not as a result of volatility had elevated, however as a result of they suspected the consumer had info. They may present measurement selectively, lean on inside flows, or warehouse threat primarily based on expertise quite than fashions.
Importantly, liquidity was elastic. It could possibly be negotiated, delayed, reshaped, or withheld completely. The idea of a “agency value” was versatile, and execution high quality was inseparable from relationships. Information existed — but it surely was secondary. The first sign was dialog both on the cellphone or on the Bloomberg/Reuters chat.
Digital buying and selling didn’t take away discretion — it encoded it
The primary wave of digital FX buying and selling didn’t get rid of seller judgement. It translated it. Single-dealer platforms (SDPs) allowed banks to stream costs to purchasers, however these costs had been nonetheless formed by:
- Consumer tiering
- Historic behaviour
- Inside threat limits
- Seller instinct
What modified was the interface, not the decision-making. Liquidity grew to become steady quite than episodic, but it surely was nonetheless deeply relationship pushed. The essential shift got here later, when multi-dealer platforms (MDPs) and algorithmic execution compelled liquidity to change into comparable. In 2000 FXConnect, FXAll, Currenex and 360T launched and extra MDPs adopted. As soon as costs from a number of banks appeared facet by facet, discretion needed to be expressed in a method machines may course of. That required parameters.
Parametrisation: the second liquidity grew to become machine-readable
In FX, this occurred progressively however decisively. Sellers started to specific liquidity by:
- Unfold widths
- Skew changes
- Measurement tiers
- Timeouts
- Reject logic
- Final look thresholds
- Consumer tiering ranges (Gold, Silver, Bronze, and so on)
Every parameter encoded a chunk of human decision-making:
- How a lot threat am I keen to take?
- How assured am I on this value?
- How poisonous do I feel this circulation is?
- How briskly do I wish to reply?
Liquidity stopped being a dialog and have become a perform.
From the buy-side perspective, this was transformative. As a substitute of negotiating, merchants may probe. They may ship RFQs, stream requests, and youngster orders to deduce liquidity circumstances statistically quite than socially.
Execution algos then formalised the brand new language of liquidity
The rise of FX execution algorithms accomplished the transition.
Execution algos required liquidity to be:
- Observable
- Predictable
- Quantifiable
An algo can’t “sense the market” the way in which a human dealer as soon as did. It wants inputs. In consequence, liquidity needed to be damaged down into measurable parts:
- Fill likelihood
- Market affect
- Slippage distribution
- Latency sensitivity
This compelled each side of the market right into a suggestions loop. Banks tuned their parameters to guard towards hostile choice. Purchase-side corporations measured these parameters implicitly by analysing outcomes. Over time, liquidity grew to become one thing inferred from information outputs quite than proven explicitly by relationships. On this sense, execution algos didn’t simply devour liquidity they contributed to reshaping it.
As liquidity grew to become parametrised, buy-side corporations tailored by changing into information companies
Why “liquidity” in FX not means what it used to
In a parametrised market, liquidity shouldn’t be depth on the high of e-book. It’s a conditional likelihood.
A decent value is barely significant if:
- It’s agency
- It survives latency
- It fills on the anticipated measurement and value
- It doesn’t disappear on an execution try
Because of this FX liquidity typically seems to be plentiful till it isn’t. Throughout regular circumstances, parameterised liquidity performs effectively. Fashions are calibrated on steady regimes, reject charges are low, and spreads behave predictably.
Underneath stress, parameters flip:
- Measurement thresholds drop
- Final look home windows tighten
- Skews widen asymmetrically
- Streams pause completely
What disappears shouldn’t be liquidity itself, however the assumptions embedded within the parameters. Is that this when relationships matter or when the info aligns?
The buy-side response: measuring what can’t be seen
As liquidity grew to become parametrised, buy-side corporations tailored by changing into information companies. The final 10 years have seen this enhance quickly throughout the buyside, though maybe much less so with company treasurers.
Buyside Execution desks started to seize and analyse:
- Quote-to-trade ratios
- Reject causes
- Time-to-fill distributions
- Venue-specific efficiency
- LP behaviour by pair, measurement, and time of day
This information grew to become was key to a change in mindset on the buyside. Over time, refined buy-side corporations stopped asking “the place is the very best value?” and began asking:
- The place is essentially the most dependable liquidity?
- Which LPs behave persistently in stress?
- How does liquidity decay as measurement will increase?
These questions can solely be answered statistically — one other signal that liquidity had change into abstracted from human interplay.
Venue evolution: from execution to information infrastructure
The parametrisation of liquidity is also put ahead as one other strategic cause exchanges acquired FX platforms. I’ve written on this in earlier articles however offers akin to CME and EBS, LSEG and Refinitiv, Deutsche Börse and 360T and BidFX and SGX. They weren’t merely about quantity. They had been about proudly owning the rails on which parameterised liquidity flows.
Venues more and more act as:
- Normalisers of heterogeneous liquidity
- Distributors of analytics
- Repositories of historic behaviour
In a world the place liquidity is outlined by parameters, the flexibility to standardise, measure, and replay these parameters turns into strategically important. Execution is ephemeral. Information persists.
What was misplaced — and what was gained
The parametrisation of FX liquidity introduced plain advantages:
- Decrease transaction prices
- Larger transparency
- Scalability throughout areas and time zones
- Decreased reliance on particular person sellers
However one thing was misplaced as effectively.
Human discretion as soon as absorbed ambiguity. A seller may select to point out liquidity regardless of uncertainty, primarily based on judgement. Parameterised methods are much less forgiving. When uncertainty rises, the default response is commonly to withdraw. Because of this FX liquidity can really feel binary for some purchasers: plentiful till out of the blue absent.
The following part: adaptive parametrisation
The way forward for FX liquidity is rarely going to be a return to voice, nor a easy extension of present algos. It’s adaptive parametrisation.
This consists of:
- Dynamic skewing primarily based on real-time circulation toxicity
- Machine studying fashions for reject likelihood
- Venue choice that adapts intra-order
- Suggestions loops that replace parameters repeatedly
However even right here, the core fact stays, that liquidity continues to be being expressed by parameters. The distinction is that these parameters at the moment are adjusted quicker and with extra information. The market has not change into much less human — it has change into human judgement at scale, encoded in methods. Among the eFX platforms have seen this and anybody dialled into wanting behind the press releases can see that strategically some platforms are effectively upfront of others when it comes to future proofing.
We haven’t even approached how AI will have an effect on these MDPs and the buying and selling/liquidity on them. It’s a topic for a bigger piece however take a look at what among the sharpest minds within the biz say about AI and its affect on buying and selling corporations operationally. See the remarks just lately within the press from Citadel CTO Umesh Subramanian and Schonfelds Ryan Tolkin.
Conclusion: liquidity as a design alternative
FX liquidity didn’t disappear. It was redesigned. What was as soon as negotiated grew to become calculated. What was as soon as implicit grew to become specific. What was as soon as private grew to become statistical. Understanding trendy FX markets requires understanding this transformation. Not simply how algos work, however how liquidity itself was was code and what meaning when the assumptions behind that code are examined.
Readers can see extra of John McGrath’s articles on his Substack web page: johnjmcgrath.substack.com