The Last Look…

Data published by two platforms provides an interesting insight into FX market behaviour during market upheaval, reinforcing the sense that, to paraphrase John Gray’s classic manual, LPs are from Mars and customers are from Venus (and yes, I know I could have substituted the planet for a childish joke!)

What we witnessed at the start of last month was a proper market move, something that always seems to happen when retail gets heavily involved in a strategy as they had with the carry trade (food for thought for those who think institutional and retail can live side-by-side?)

Putting aside the retail element, it was informative to see the data published by Euronext FX and Cboe FX on their platform performance during August, as it shows where people go to trade, and how much more defensive the LPs become during uncertain times.

First up, and to reiterate what is sadly a regular refrain from this column, kudos to the two platforms for publishing this sort of data, and where are you the rest? It’s great to publish nice big volume numbers, but there is no sense of how hard, or otherwise, it is to actually trade? When markets go sideways, volumes are up anyway, what prospective customers need to know is where they are most likely to get their trades done?

Away from that gripe, it was notable that data for the last month from Cboe and Euronext, (using the entire month means there has been time for outliers to be smoothed out) shows lower fill rates and longer hold/round-trip times. This is intuitive to anyone who understands the market structure, even in the voice days, it would often take much longer to get a price when things were kicking off, what was interesting to me was how anonymous volumes seem to tick up as well.

On the first point, data published on the respective firms’ websites, show that the while the last look round trip time on Euronext FX to New York was unchanged at 10.6 ms (it has been there or thereabouts for the last five months), to London it blew out to 19.6ms (it has averaged 9.2ms for the year to July, so this could be a mistake, but it is longer either way). Likewise, on Cboe FX the average response time went to 10ms from 9 or 8ms it has been at for most of this year.

Obviously, pipes are busy with trade requests which can slow things, but these extended travel times are probably also the result of LPs slowing things down at their end. To highlight the point, one LP I spoke to observed they are not a “charity” and cannot continue to throw out prices at the same pace they normally do, while one liquidity consumer complained to me that the LPs were “not letting us hit their prices on certain venues”.

We should not expect static fill rates or RTTs, it’s just unreasonable (as are high reject rates and asymmetric RTTs)

Complaints from either side are nothing new, back in the voice days the customer complained they were being “read” by the LP, and the latter complained that the customer was “asking everywhere”, but it does raise the question, ‘how far should LPs go in providing prices?”

A lot of LPs, naturally, push their best prices down their disclosed, bilateral channels when things get lively – they still price to the platforms, but it may not be the “A” stream – and we should not ignore the very successful month that both CboeFX and Euronext FX had in August, perhaps customers go there because they know the performance characteristics upfront?

Slightly slower response times, therefore, are likely, as is the second data point I want to discuss, fill rates.

Cboe’s non-firm fill rate dropped to 88% from 90%, it has pretty much been at 90% and above for all but a couple of months over the past year, while Euronext FX saw its Skew Safe fill rate drop to 80.2% from 83.1% (this is the lowest since the platform started publishing the data in early 2023), while on the platform it plummeted to 72.8%. This fill rate has, until July, when it dropped to 78.3%, been comfortable above 80% since the data was first published.

LPs, therefore, were rejecting more, and again, a different LP tells me they thought some clients were “too aggressive” and that the firm saw reject rates soar for “certain types” of counterparty (hedge funds and prop shops were hinted at). Conversely, while still below the recent average, the Euronext Full Amount streams’ fill rate rose from July’s 93.4% to 93.7%.

There are, I accept, a couple of stark numbers in there, but would ask question, ‘why is anyone surprised by them?’ The complaints I received from some (not all I should stress) consumers just highlights for me the sense of entitlement too many in this business have. It is not the right of anyone to have a guaranteed, tight, firm, price no matter what is occurring in the world. I have been there (a long while ago I know) but market making is not an easy, or nowadays a cheap, business – there are too many who feel that LPs are there to put up and shut up, but these are businesses too.

It does not help, I acknowledge, that some LPs make grandiose claims about how great their pricing is, and always will be. Likewise, it’s hard, as I have noted before, to have sympathy for a business with multi-billion-dollar revenues, but the fact is these firms are not, as my friend observed, a charity – there has to be something in it for them.

This is not a new discord in the industry, of course, it has been going on for decades, but it is interesting to see in empirical terms how the different behaviours manifest themselves. Consumers seem to go to anonymous venues when the smelly stuff hits the fan, and then complain about worse performance metrics, thus completely ignoring the fact that for many of these participants the best prices and liquidity are available elsewhere.

spread is not a good indicator of liquidity; when things go pear-shaped, any bid is a good bid

Someone observed to me that this is the consumers arguing they would prefer the exchange-type environment, but I am not convinced that is a better option. In August, CME group revealed some data for the first week of the month, highlighting how busy things were, but also giving an insight into how different models work. For that week CME says it saw just short of $250 billion per day in FX spot, futures, NDFs and options. Within this were $11.3 billion in one-month NDFs, as well as the highest single volume day on EBS since March 2020 at just over $145 billion. EBS Direct hit a new all-time high at just short of $42 billion and FX Link also hit a new high at $13.5 billion.

Obviously, FX futures and options would have seen a spike, but would it have been anywhere near that seen in OTC, where, for example, Cboe FX saw activity almost double from the previous week? FX Link is also very much a tool aimed at helping OTC traders access futures liquidity, does its new record indicate that the main driver was from the OTC sector?

Without more granular data it is hard to know, but the sense is that the FX market mayhem we saw early last month did have an impact on market functioning, but that it continued to work, albeit less efficiently in some models. Repeating myself from previous columns, spread is not a good indicator of liquidity; when things go pear-shaped, any bid is a good bid, but most of the bids appear to exist in a disclosed environment.

This is not a revelation, but this data should reinforce the idea that in different market conditions, different styles of trading are required. We should not expect static fill rates or RTTs, it’s just unreasonable (as are high reject rates and asymmetric RTTs), and in a firm environment, there are still misses (rather than rejects) that cost money.

It is very sad that other venues operating an ECN/CLOB environment don’t publish their performance data – as noted, it’s less than helpful for a customer who may be considering using that venue – because it is through data like this that we can establish expectations.

If the empirical evidence is there, then people won’t get upset by slower RTTs and higher reject rates in certain market conditions and things just operate in a smoother fashion. Sales desks at LPs and relationship managers at platforms won’t have to waste valuable time calming an unreasonable and irate client down, and those clients will also benefit from a better understanding of how the FX market really works.

The pity is, too many of the latter will refuse to listen – the sense of entitlement runs deep…

colin@thefullfx.com