I mentioned in a recent column the increasing noise around fragmentation in equity markets, specifically end user frustration and unhappiness with it, but it is noticeable to me that the same discussion is gaining more volume in FX. Now I need to stress that this is not me (for about the 10th time) predicting consolidation in the public platform world, rather I suspect that the evolution, which started last year, towards supporting fewer venues, will gather pace as 2020 progresses and as such winners and losers will emerge.
I am not sure the “losers” will disappear, as I have noted many times in the past, there are plenty of venues that do enough to survive and whose growth prospects are pretty limited. What intrigues me is not so much who will thrive and who will wither – that is a very difficult question – but more what type of model will do best?
To me the market seems to be heading towards a bifurcated structure – on one hand we have the all-to-all broker model and on the other a disclosed relationship model. Those who sit in the middle are, I suspect, going to struggle as market participants look very carefully at their connections and LPs become more choosy over who they stream their good prices to. In this world, the betwixt and between model creates ambiguity and gives the participant a reason to cut it.
Behaviour will also be an issue. On a consenting adults, no last look broking model, participants know what to expect and will price and trade accordingly. Similarly, on a disclosed model, they will know who they are pricing and can monitor behaviour (just as the customer can monitor LP behaviour). The problem with the blended, multi-faceted model is that it is harder to track behaviour – and that risks LPs in particular, putting that venue in the “too hard” basket.
This may not have been important in the past, but as LPs use data analytics more positively, the concept that they will blindly follow a client to a venue will become outdated. The world, not least the LPs themselves, has woken up to the fact that liquidity is a very precious commodity and as such, there is going to be a more balanced relationship between LP and LC.
I also think that the changing importance of client segments is going to play a role – as LPs start to look more closely at their relationships there will be an eagerness to meet the current darlings – real money – on a disclosed basis, while the former stars of the client portfolio – active traders – are pushed to the broker model where the LP has less commitment to price aggressively.
I raise this issue because there has been a lot of chatter about last Friday’s activity in FX markets and how, according to some, the volume was more spread out than previously. In the past, we have seen a rush to the primary markets when the smelly stuff has hit the fan, it is being suggested to me that last week indicated that is less the case now. I am not so sure, but am willing to embrace the suggestion, after all, you can’t argue with new records being set at several venues. We should not be surprised, however, that a bunch of platforms hit new peaks in terms of ADV, FX markets have always traditionally been busier in crisis environments, indeed it is a little heartening to this FX geek to find out that the market does still react to news for longer than a few seconds! Equally, although some are making a big deal of it, I don’t think we should read anything into the fact that the primary venues did not hit new records – they were around in the heyday of the ECN when 150-200 yard days were almost a norm, whereas so many of the challengers were not.
That said, there were some impressive numbers to be had – but there were also some that made you think. As Profit & Loss reported, on the ECN/broker side, CboeFX and Euronext FX hit new peaks at $75 billion and $45 billion, respectively, while 360T (and my understanding is that the GTX ECN venue remains the dominant spot venue) hit $47 billion.
What was of interest to me – and may shoot my theory down – is how Cboe’s “firm” liquidity remained responsible for about a third of the overall volume as it is most days. I would have thought that more customers would have preferred to have traded against a “firm” price than one subject to last look (indeed the exchange’s non-firm fill rate last Friday dropped from around the 87% mark to just over 81%). Activity on other firm streams remained strong, sources familiar with the matter say that Refinitiv’s Matching (which is no last look of course) handled approaching 90 yards on Friday, double what it normally would do, while CME had a huge day, approaching 200 yards (I have not been able to get specific EBS data for Friday by time of going to press, but I understand that for the week as a whole, volume was 57% above the 2019 average with G10 being 80% higher).
Furthermore, at 24 yards, CboeFX liquidity was significantly surpassed, I am told, by LMAX Exchange, which hit 36 yards (itself a new record for the venue). To my earlier point about simpler, more straightforward models, speaking to a senior e-trader this week about this I was told simply, “you know where you stand with no-last look”, in other words, for some business you need “see price, hit price”.
I am reluctant to talk specifics around this, not least because one swallow does not make a summer and we need much more data, but who would have thought just a few years ago that in terms of firm liquidity the ADV ladder would read CME, EBS and Matching followed by LMAX Exchange? (you can put your hand down now David Mercer!)
Going back to CboeFX, I suppose if the non-firm prices are that much tighter a 6% increase in reject rates may be worth the price for some customers, although I wonder if the latter actually use any of the several excellent tools out there that can give them a cost of reject calculation?
Turning to the disclosed model it is a much murkier picture, mainly because so few actually publish volumes on a monthly basis, let alone daily, and when they do, they are often across all FX products, not just spot. That said, again I think there is a case to argue that participants are leaning towards the model that is that little bit simpler, in this case the fully disclosed, dedicated stream.
The model in this case, that seems to be gathering the most traction remains FXSpotStream, which is inevitable because it wasn’t that long ago that I may have suggested volumes were plateauing at the platform! Of course, my point was really that the platform’s growth trajectory had to inevitably flatten out at some stage, but it’s nice to know the Midas touch still works!
More seriously, in the absence of other data, FXSS becomes the bellweather for the disclosed, relationship model and last week was eye-opening in that the platform hit a new record above $86 billion, which is not only a serious number, but also suggests to me that volatile conditions lead to those customers not interested in the broker model going to a venue where the LP knows who they are pricing. We should not be surprised because this model brings accountability, as I noted earlier. A genuine LP must be more comfortable streaming to a known counterparty within a limited liquidity pool and should be able to demonstrate that the available liquidity is deeper and at a slightly tighter price than elsewhere. These LPs also need to be grown ups about the relationship and understand that sometimes a customer’s timing will be ideal and they should not be punished for that with wider or shallower pricing for a period of time, but generally speaking this model works well.
For the consumer, the aforementioned tighter price in larger amounts (and this can be, to a limited degree, in competition) gives them more certainty over their execution when it gets busy – which is what it’s all about, just look at how liquidity still dominates customer thinking. Equally, the consumer needs to be adult about how they trade, running over people is so 2000 (as latency arbitrage is so 2006), therefore they need to execute responsibly.
Of more interest to me with the disclosed model is the reject rate. Obviously CboeFX is a different model so we shouldn’t compare fill rates, but I understand that on FXSS last Friday the fill rate was 95%, very close to what I am told is a 97% fill rate for February as a whole (it was 96% for the last week of February). It would be interesting to see how this stands up against other venues, but of course we can’t because they don’t offer this data (cue my semi-annual moan about said lack of transparency!) Notwithstanding that, it is significant to me that the fill rate remained steady – that suggests LPs comfortable with the model.
Having noted that, I have to say that I believe more needs to be done. Someone asked me recently what I thought the ideal fill rate should be on a disclosed venue and I replied, in all seriousness, 99.5%-plus. Of course, the nirvana is 100% but I understand you have to make the odd allowance for technical issues such as a breakdown in connectivity and/or a breach of credit limits due to a multiple hit (and even there the customer needs to be asked why they were taking advantage of a dedicated stream by spraying the market?).
I think in all the time I have been writing about the multi-dealer platform industry I have nailed two trends; the dilution of the position of the primary venues (and as long as the market continues to consume their data, they don’t really have to worry too much) and the rise of disclosed trading. The latter is, without doubt, in the ascendancy, especially in this age of accountability, and FXSS’ numbers support that, with year-on-year and month-on-month increases above most other venues (of all ilks).
That is not to say, however, that anonymous venues are going to struggle. Certainty of execution is likely to become an even bigger thing, especially if this return to volatility is sustained, and while that points, as I have argued, to knowing who you are dealing with, it is also very much about accessing a firm, preferably no-last look, liquidity pool that is robust and that can demonstrate its worth in the toughest of conditions.
This all points, in my view, to a future where people pick one of two models on which to trade – and the key is both are simple. We have been through a period of complexity, the return of volatility is likely to merely highlight the benefit of the straightforward approach – and the winners will be those who excel at one thing and decline to be all things to all people.