The Last Look…
One aspect of having a public, and rather robust stance on certain issues – in this case last look – is that you do receive plenty of ideas about how to solve the problem, and sometimes, as was the instance recently, they are off-the-scale crazy, even in my book!
What I thought was a light-hearted remark at a social meet, dissolved into a serious discussion about whether the “solution” could work or not (and yes, a cold beverage was involved). What was suggested? Simply that LPs should compensate consumers for the cost of rejects!
Now I am no fan of last look, but even to me this sounded bizarre, for a start, how would the cost be calculated? My friend had that nailed down – use an independent data service, that can provide data on a millisecond basis (they also pointed out that, as I have written multiple times previously, that the average reject takes place over multiple milliseconds, in some cases over a second, therefore the data doesn’t even need to be that quick). There was also a suggestion that the data could be aligned with the average hold time for rejects, thus introducing double jeopardy for the LPs.
My immediate problem with such an idea is that it would reward bad behaviour from certain liquidity consumers, who happily operate with 80%-plus reject rates – imagine how much an LP would be paying out to an aggressive hedge fund for example? It would also trigger chaos in the FX market as everyone reaches for the calculator rather than focus on the risk.
In fairness to my friend, they did also suggest that reject improvements should be taken into account when calculating, for example, a monthly bill for rejects, but I think we all know the nature of that beast means that the LPs will get a minimal rebate at best.
As I argued my corner, bizarrely for the last lookers, the idea shifted slightly, to one where only LPs with asymmetric hold times were punished – something that would make the process even more complicated. Monthly data, which is admittedly widely available, would be used retrospectively, to notify an LP they were on the hook for the next month’s rejects.
We need to deal with the issue of asymmetry around last look, and I don’t mind the idea of punishing asymmetric hold times. I don’t have a problem even, with punishing or inhibiting the consumers who act in a predatory fashion – but surely there has to be a better way than this?
As we continued back and forth, another concession was that LPs would not be charged, if the consumer’s reject rate was over 20% – something indicating that they trade in a predatory fashion. Again, fine in principle (I suppose), but surely the LPs will merely tighten their criteria further to ensure any consumer remotely toxic (in their view) will be rejected more?
My friend argued, in summary, that such an approach would lead to LPs holding more risk on their books, because they would reject less and they would probably back the analysis that tells them certain flow is good after minutes, for example, rather than robotically internalise it. They also pointed out that this would, inevitably, lead to wider prices for predatory clients as LPs react to losing their safety net when quoting them.
I must confess, I do like the latter outcome, because I think that too many LPs have historically allowed some clients to behave badly because they “need to see the flow”, which is, of course, rubbish – especially if the move has already happened. Who “needs” to see it; the sales desk for something to talk about? They’ll already be too late. The trading book to go with it? Ditto.
We do still need to deal with the issue of asymmetry around last look, and I don’t mind the idea of punishing asymmetric hold times. I don’t have a problem even, with punishing or inhibiting the consumers who act in a predatory fashion – but surely there has to be a better way than this?
So, three observations to close with. Firstly, who wouldn’t want to be in the pub with me for such riveting conversation? Secondly, you will be surprised to hear that my “friends” often like to trigger me on last look and this could have been a wind-up (it didn’t feel it though); and thirdly, AI, which triggered the discussion, has a lot to answer for.
