The Last Look…
Posted by Colin Lambert. Last updated: June 8, 2021
As a hot topic over the past three years, data has had a good run – people see it as having a positive influence on the market and a shining example of how advances in technology bring better conditions. With these advances, however, have come challenges to certain business models in the industry and I guess it’s up to the readers to decide whether this is a good, or bad, thing.
Although so many challenges in this industry are faced by banks and platforms, for the buy side too, there are issues that emerge from time to time, and one of them in recent years has been improvements in data science. Many funds use data science to help their decision making, but another side of the advance has provided, if not a threat, then a question over their modus operandi in FX markets.
I had two interesting chats with people in the hedge fund world recently, one that suggested the firm was struggling to maintain a panel of LPs, the other that the panel needed widening – go figure! The larger of the two funds was moving to a more bilateral relationship model with their LPs, the smaller, (but still sizeable) fund felt they were being too aggressively “read” by their LPs’ skews.
For larger funds, the advance in data and analytics, especially from independent sources, means the days of banging the market around are, if not at an end, probably very close to it. A good friend of mine in the e-FX space has long maintained that ultra-aggressive hedge funds were getting away with predatory behaviour because they still had the support of the bank’s sales desk, or more pertinently, the head of that desk who probably grew up as part of the generation that thought all flow was good and the more of it the better. Without doubt there has been an element of conflict between trading and sales when it comes to certain clients – it goes back decades – but my conversant senses those days were at an end.
I was recounted by the larger fund, a story of a meeting late last year with an LP, it was a long-standing relationship and the meeting was, apparently, “pretty aggressive” in its tone. The LP came armed with data, nothing new it has to be said, but for the first time it was from a third party and it laid out the challenges of dealing with the hedge fund’s flow.
Since then the fund in question has had similar conversations with a number of their top LPs – they have also come to the conclusion that the ECN channel is definitely not for them – “too many eyes and ears” as it was put to me. The upshot is a reduction in panel size and the fund itself starting to use execution quality analysis.
My takeaway from the chat was that we had an aggressive, some would say predatory, market participant, who has been forced to change their approach to the FX market thanks to advances in data and analytics.
The latter surprised me a little, I would have thought it had been in play for years, but it seems, as an aggressive player, market impact was a benefit – until it got out of hand and became too large – and as such it never worried about how well the trade was done!
My takeaway from the chat was that we had an aggressive, some would say predatory, market participant, who has been forced to change their approach to the FX market thanks to advances in data and analytics. If this is indeed the case then I would argue it is a good outcome for the market as a whole (especially the LPs still picking up body parts after being mangled by the fund for the 10th time that week) and highlights the beneficial effects of data.
My other conversation appears to have highlighted the negative impact of better data and analytics, however. The smaller hedge fund had analysis which indicated that LPs were very good at predicting their trades and were streaming quite aggressive skews at the appropriate times. This was prompting some issues when it came to entry and exit levels, especially in quieter markets where a pip or two can make a difference over time.
My first instinct was it is perfectly acceptable for an LP to skew a client, this is why the client has choice, apparently however, this fund, with a panel of four LPs, felt all their LPs were doing it. Again, I am not sure there is a problem here beyond the client perhaps looking at their strategies and checking they make enough to compensate for their obvious predictability.
Interestingly, as a digression, the fund in question started asking the LPs to execute “at best” orders, but found the latter were reluctant. This could have something to do with best execution requirements, more likely in a situation such as that, the LP couldn’t make money out of the flow. The fund in question was pushed to the LP’s algo suite apparently, with the appropriate fees.
This fund, therefore, has seen the negative side of data science and has broadened its panel of LPs to help shield the business (again, they eschewed the ECN route due to signalling issues) from prying computers.
I find this story interesting because it does highlight two different effects of the same change. It has also, it could be argued, brought one counterparty “into line”, and prompted two different firms to change how they operate on a day-to-day basis.
I also think it highlights another change – lower risk appetite for customer flow. Without doubt the LPs in question are happy to take the flow on their terms, which is, in itself, a change in tone, but they are acting more like brokers every week that passes. I still believe that for customers whose first priority is hedging and not market direction, the FX market has plenty of risk absorbers happy to provide a service – for the more active, directional firms, it is a different story.
My sense is the broadest application for data science at this time is for defensive, or negative, purposes
In the past, there is no doubt that a bank trading desk would take certain hedge fund flow and trade with it – into the bargain reinforcing the fund’s direction of travel – but now it is such a grey area in terms of conduct, that most desks are happy to execute the flow, provide the flow analysis to the clients (anonymised of course) and walk away with P&L (hopefully) intact. The challenge is, this flow used to be profitable because the trading desk went with it, there was often an initial loss on covering the trade.
We know that data science is changing the nature of the FX market, but the change is being exacerbated by a reduction in risk appetite and LPs’ willingness – even permission – to take a deliberate market position. Combined, the two trends have seen LPs become even more like brokers – why else would there be a prolonged discussion about riskless principal? – and that represents a threat to the good market function.
By taking out the risk appetite you remove any of the benefit of seeing the hedge fund flow described here. This in turn means those hedge funds find the FX market a less appealing place to do business and, in time, reduce their activity (probably going to crypto).
There is no right and wrong answer here, but in an ideal world data science does root out predatory behaviour on the part of some liquidity consumers, and at the same time the FX market, by having a decent risk profile, provides a better landscape in which it can be used.
My sense is the broadest application for data science at this time is for defensive, or negative, purposes – to help find a reason to hold a trade longer in the last look window or to stop providing risk transfer services to certain clients. Wouldn’t it be nice if it could be used in association with a greater risk appetite? Certainly if that occurred then some LPs might find they need to change their business model back from broking to risk taking.