It’s about time, that the practice in FX of liquidity providers having a ‘last look’ before accepting a trade – a legacy from the old days of phone quoting, when the dealer took one ‘last look’ at pricing before accepting a trade – may at last be on its way out.
In the electronic platform era, last look, was a way to encourage banks to increase their liquidity provision and provide request for quote (RFQ) and executable streaming prices (ESP) where they may not know who was asking for the price, or who might hit their streaming prices.
In order to protect themselves from what some call toxic flows (here and here), the Liquidity Providers (LPs) were offered some degree of protection when quoting pricing by being given one ‘last look’ when someone attempts to deal on their quote, before deciding whether or not to accept the trade – ie they can decide if they like the trade before accepting it.
The problem with platforms that support last look is that clients are misled by the illusion of firm pricing, and in reality cannot rely on being able to trade on prices with certainty (unless of course they hit the ‘wrong side’, in which case the trade will certainly be accepted).
The practice actually distorts the true picture of available liquidity on that platform, which is then multiplied, either by the same LP quoting across multiple platforms, or other LPs using that quote as an input for their own electronic pricing – thus last look is closely linked to what is sometimes called the liquidity mirage, where prices appear to be firm, but like a mirage, when you reach them (or in this case try to trade on them), they disappear as the LP has their last look.
Next week (10th June) sees the release of the Fair and Effective Markets Review (FEMR), which was set up last year by the Bank of England and the Financial Conduct Authority to look into practices within the FICC markets and is expected to include recommendations into methods and practices that govern FX execution and will no doubt lead to new rules that govern how FX is executed.
It’s therefore interesting to see that two leading FX platforms have announced changes in their trading protocols to address the practice of last look. FXall and Hotspot FX (now owned by BATS Global Trading), have both taken the decision to introduce new trading protocols that will dramatically reduce the ability for LPs to have last look.
Hotspot has reduced the time a liquidity provider has to accept/reject a trade from 200ms to 100ms and increased the ratio of quotes to trades to 85%
Commenting on the changes, William Goodbody, head of FX at Hotspot, said,
“Last look is a widely used practice in the industry. To make it work, it needs a clear set of guidelines.”
Whilst Phil Weisberg, Global Head of FX at Thomson Reuters said,
“Following the new Matching rules we set last year that encouraged high standards of behavior in primary markets, the updated FXall operational procedures define the same rigorous standards for both our RFQ and Streaming Price trading protocols, where participants transact on a disclosed, relationship basis, as well as our Order Book platform, where participants trade anonymously using either firm or Provisory Liquidity.”
Some top-tier banks and electronic market makers that have invested in highly robust and efficient pricing engine and risk management capabilities do provide trading with no last look, and they benefit from increased deal flow.
One of the major reasons for the success of the new EBSDirect service is that it’s based on disclosed relationship pricing, and as such the LP will know who is on the other side of the price, and in general will be more likely to offer pricing without last look – ie firm dealable liquidity – which after all is what price takers want from a platform.
Filed under: FX, Paul Blank |
SingleDealerPlatforms.Org 
Good post. A couple of questions:
What are your thoughts on the impact of this change on traded spreads?
And on overall profitability of liquidity providers?
Hi John,
Thanks for the comment.
In terms of any individual dealers spread, one would assume that if time in force and quote/ hit ratios increase (ie they must commit to pricing more often and have less time to reject a trade), that in order to protect themselves they would widen their quote, or more aggressively skew pricing to reflect their interest.
The ECN best bid-ask may not change in ‘normal market’ conditions, as there are many dealers to provide liquidity. However, in times of stress and one way moves (such as the EURCHF move in January), I can only imagine that pricing will become far more volatile and we may see more liquidity gaps until pricing settles down.
Forcing dealers to trade when they don’t have a natural interest can make them gun shy, although this is far less of an issue for the larger players that internalise high levels of flow, and can more easily support quoting with no last look.
I think it will encourage more relationship based pricing on a disclosed basis, where banks are more happy to quote provided they know who the client is, and the type of flows they can expect.
In terms of profitability, conventional wisdom would say that it should fall, but for those that have invested in the technology to support trading with no last look, they may well increase profitability as they increase their liquidity provision.
Just some thoughts,
Paul
An interesting column given you work for a GUI company. GUI’s by design are never co-located. So should your GUI clients or their liquidity providers take the risk on the market moving in the time between the GUI finger click and the order hitting the LP’s server?
HI Chris,
As always, you raise good points.
Assuming I have understand you correctly, as we know, some platforms have looked to redress the balance and level the playing field for GUI users by introducing speed bumps and randomisation pauses, to remove the latency advantage of co-location.
However, that doesn’t remove the question of who takes the risk on ‘slippage’ for GUI users.
Here I tend to think that with relationship based pricing, especially where the client executes the trade on a bank’s SDP, that the bank should be rewarding that flow by providing firmer prices with less chance of rejection due to slippage. Although, the bank should of course continue to review the terms of pricing for clients that continually sought to ‘pick off’ slow dealer quotes.
That relationship should also hold between the bank and it’s own liquidity providers.
In my view, relationship flow should be rewarded with firm pricing, whether that’s bank to client, or bank to their LPs.
Kind regards
Paul
Chris. Do you mind if I ask who you think should carry that risk?
I should mention I am an industry professional commenting personally and not on behalf of my firm! A firm mind you that does not, and never has, employed a last look hold time. The decision and the reasons are for another time.
John if you agree there is no such thing as a ‘free lunch’ then either side can carry the risk – at the right cost. The liquidity consumer can accept a certain % of rejections and or some hold time in return for the LP maintaining the tightest price possible. Or the LP can widen the price to build in the slippage – maybe Paul’s GUI client lives in outer Mongolia and is using a 56k modem. Some GUI clients are very well informed. Maybe they hit a number of banks at the same time, maybe they usually use a primary market and only (cherry pick) banks when they cannot get the liquidity elsewhere. Each of these things comes at a cost and personally I think given so many variables that the LP should be the one with the right. Remember no one is policing the liquidity consumer today!
Something has to give. Unless you are co-located and your technology is able to take 1000 updates a second (or more) there will always be latency that someone needs to pay for,
Back to Paul’s absolutely genuine client in Mongolia. If I sat down with him and said “Look you are almost a second away from the market. My model suggests I can give you the tightest price I can show anyone with an 85% fill rate. I can give you the same price but give me an extra 1/5th of a second to try and fill the other 15% and I am sure I can then fill you 90% of the time”. I think the guy in Mongolia would accept that deal. if he is genuine what does he have to lose? 200 ms is the blink of an eye. He will not even notice it.
This isn’t just theory – it happens successfully today on trillions of dollars of business.
Chris – an interesting answer, thanks for the considered reply.
The value of the embedded option contained in an executable quote and the latency trade-off is something that is often not well understood by some buy-sides.
When you look at the relative price tag attached to sell-side versus buy-side technology for the vast swathe of the real-money market it’s pretty clear which side is better able to carry the risk.
Chris is absolutely right. The last look is to protect against the tail events, embedding this in the spread is already available to customers right now in central limits order books (eg Ebs) they choose to have the rejects and the tight spreads.
How do the wonderful Hedgies to it? 1000 quotes a
Second with an immediate out… That is not providing liquidity.
The original article is naive to say the least. Careful what you wish for….
It seems that there is alway a tradeoff between tighter spread and ‘last look’ to balance the risk and benefits among all parties. As Paul pointed out in the article that the cause of the ‘last look’ is from the toxic flows, which comes mostly from api trading models. And technology is never perfect and its improvement from the LPs are only relative to the counter parties’ over time. It seems that the dark pool provides an answer to it.
https://www.linkedin.com/pulse/true-benefits-fx-dark-pool-rutie-zhou