Alpari’s onslaught against ‘last look’ execution, our opinion and its tremendous response: Exclusive insight from the non-bank industry

Last week, FinanceFeeds reported that Alpari’s US division has invoked a lawsuit against Citigroup for its ‘last look’ execution practices, something which has been a moot point for several years among non-bank institutional and retail electronic trading companies (and something that I have reported on extensively and see this as a milestone in the direction of common sense – Ed).

On Friday last week, an opinion on this which was published by FinanceFeeds detailed why this is something very important and gave full insight into why the battle against the Tier 1 FX dealers by a non-bank institution over execution is a very important point in the future of the industry’s structural and commercial methodology.

Our opinion generated a tremendous response from a cross section of large companies in the non-bank ECN sector as well as those facing the Tier 1 banks, one of which is most certainly of value to interested parties.

In this case, FinanceFeeds has assured anonymity to the party concerned, as well as the name of the institution represented, and also insists that no commercial partnership exists between FinanceFeeds and the company or individual whose views are represented here. This particular opinion is from a middle/senior executive within a non-bank institution in North America.

Here is the full response:

In statistical and absolute dollar PnL terms, 100-300ms last look does very little on benign order flow when makers have good pricing systems and attempt to fill all profitable orders. Most of the venues cap last-look at 100-200ms and require fairly high average fill rates for market-makers. Hotspot publishes their requirements, so they’re worth referencing.

In other words, in practical terms, last-look doesn’t do much of anything at all when handling retail, small, or GUI order flow. If a new order is randomly submitted against a $20 spread, it’s not very probable that the order goes from being worth $20 for the maker to <$0 w/in 100ms. That’d require a fairly large jump in price for these markets, which is pretty unlikely during any random 100ms window of time.

Put another way, if a maker were to show pricing at, say, a 0.2pip spread (typical), taker orders were randomly in time against that spread, and they maker only filled orders that appeared profitable after 100ms, I’d bet they’d probably fill >98% of orders. The 2% rejected would probably be close to 0 value anyhow, so not a big PnL impact for either maker or taker. I’ve seen first hand examples of essentially this exact scenario.

Last-look was largely popularized by banks to be able to mitigate losses to a relative handful of HFT takers exploiting network latency differences. If, say, I had a 5ms faster line than bank makers from NYC to LON, I could just hit LON makers every time I learn of significant price movement in NYC first before they update their prices. This caused most banks to set up last-look to be able to reject exactly these sorts of orders.

I believe last-look is now more often mitigating an effect caused by liquidity aggregators: Typically, makers can give better pricing for small orders because they won’t affect inventory/risk much, and in a sense they don’t impact the market price much.

Larger orders need to be given wider prices because they effectively cost a maker more to fill. Consequently, makers show tighter pricing for small amounts and wider pricing for large amounts. This is how you’d price varying order sizes if you were just trying to process order flow at break-even (e.g., where marginal price equals marginal cost).

Aggregators are then allowing takers to bundle up small quotes from many makers (or the same makers showing the same pricing over multiple venues) and hit them all at the same time. Effectively, they’re letting large institutions aggregate and hit the pricing intended for small orders (say, $500k) when they’re impacting the market with a large order (say, $20MM).

That might seem great from the taker’s perspective at first. Rather than trickling out small orders based on your needs, you might then choose to wait until you have built up a very large interest.

If maker’s are pricing at a $20 spread, and you think that by sweeping every maker in the market you can impact the market price by $30, then you’ll sweep them and execute at a profit. You then might continually do this, whipsawing the market and taking money from your LPs until they widen prices, stop pricing to the market at large, find a way to stop pricing you in particular, or use last look to reject orders that appear to be part of one of these sweeps.

This type of sweeping behavior can be bad for the market. It can allow takers to induce unneeded volatility (bad for everyone) by trading in very large amounts not because they need to, not because they have an economic interest to trade, and not because they’re pricing new information into the market. At tight enough (say, retail) spreads, you could execute large sweeps to consistently take money from LPs by maximizing your market impact whilst minimizing your execution cost.

As I said, over time, makers will widen out your prices, stop pricing you, or try to find a way to speed bump or reject your orders. Only by blocking or rejecting your orders can they can keep pricing other friendlier flow tightly. You’d find over time that your aggregator’s previously tight spreads would deteriorate as makers begin to refuse to price you, and you’ll continually find yourself in search of new maker’s or venues not yet wise to your trading behavior.

Purely engaging in this sort of sweeping behavior can be, in a way, a special sort of market manipulation that saps liquidity from the marketplace, incurs a cost to LPs and thereby other takers via wider spreads, induces volatility, and provides no economic value or service to anyone.

I’m not suggesting very many firms do this explicitly or intentionally, only that it’s become a common practice and game of cat and mouse between makers and takers inadvertently due to the growing popularity of trading through liquidity aggregators, who, by design, aggregating pricing only appropriate for small orders to sweep as part of large orders,

Generally, it’s these large sweeping orders and latency arbitrage strategies tying to execute against stale quotes that last-look makers rejects. Natural (say, retail) order flow won’t get many rejects whatsoever, so the ability to reject would be of little to no value for most orders, and those that are rejected would be at a value be distributed close to 0 value for both maker and taker.

In any case, rejects are clearly labeled on every venue I’ve seen, so it should be easy to calculate your reject rate and the cost incurred due to rejects for any taker.

Generally, makers without last-look can’t provide competitively tight pricing compared to those using last-look. Many makers and venues will provide both a last-look and firm pricing feed, and allow takers to use either.

Of course the last-look feed will have better pricing, because they’re able to reject the worst X% of orders (I’d bet almost all from sweeps and attempts against stale prices), so the average value of order flow would be higher, and they’ll tend to pass-through their savings from fewer losing orders by providing tighter spreads, effectively charging takers less.

Consequently, the popularity of last-look is a major factor for the tightening of spreads over the last 5 years. In some sense, this reflects a savings via reduced parasitic loss incurred to makers being passed through as savings to takers, not padding makers’ profits, because market making in FX is intensely competitive so most any savings for makers will be reflected in their spreads/prices.

Generally, makers need the ability to either block and choose to not trade with takers engaging in practices harmful behaviors, or to reject orders that appear to have been targeting stale prices (or prices about to be updated prior to receiving the new order).

One way that some venues have tried to do this is to allow makers to simply block any interaction with takers that they deem toxic. I’m sure you’re familiar with the that sort of “liquidity management”.

The other is to move the last-look functionality from the maker to the ECN. Effectively, this means the ECN just speed-bumps the taking orders, doesn’t inform the makers of any pending matches, waits, say, 5ms for the maker to pull or back up his price if desired, then attempts to match the orders. This eliminates the possibility of the maker using knowledge of the pending match to trade elsewhere (guarantees pre-trade anonymity / data leakage), prevents any potential for abuse of last-look by makers, and still protects makers from potentially abusive practices on the part of takers.

Things like that are implemented a few places…

As for some of the other things you mentioned in your opinion last week.

1. EBS Live/Ultra has little to nothing to do with last-look. It’s simply their latest version of their premium market data — selling a better view of their market and thereby a considerable advantage in the market to an exclusive few firms for a high subscription price.

I’m not personally a fan of a venue in some sense auctioning off an advantage to a subset of participants and believe they should make market data available at a minimal cost for all participants.

That aside, EBS Select/Direct is now a large portion of the total EBS spot trading and utilizes last-look heavily… Due to client adoption, it’s grown from ~0 to 25% of their spot FX volume over the last few years. They’ve been trying to grow this aggressively for years, so I wouldn’t point to them or anything they’ve done as somehow pushing back against last-look.

2. Almost all other large or small venues (Hotspot, FXall, Currenex, Fastmatch, Gain, etc) offer both firm and last-look trading pools. They’re quite clear about the differences and requirements.

Reject rates and delays by LP for last look pools are generally provided on a daily, weekly, or monthly basis. Due to client adoption, it seems the large majority of volume has shifted to last-look on most venues offering both over the last several years. It seems most takers are happy to accept a small chance of a rejected order for tighter pricing.

3. As someone familiar with LMAX’s market, I don’t personally put any faith in the accuracy of their marketing, public, or regulatory/policy statements.

They make more public statements than almost every other ECN combined, so they may seem like a good source of material for your articles, but I’d take what they say about their own platform, others, or the industry with a large grain of salt. They talk their book quite a bit…

4. The FX Global Code of Conduct is well known and referenced within the industry. It reflects last look practices prominently because they’re prominent and accepted within the industry, so I don’t know that I’d try to make it seem like this law suit is the industry’s dirty secret finally getting out.

5. A handful of legal firms resurrected the legal corpse of brokerage to sue some banks over disclosures regarding a common and well known practice within the industry.

Maybe those banks did or have abused last-look with respect to how they did or didn’t inform their clients about it, but if that’s the case don’t you think the BARX case would have inspired others to have filed suit themselves rather than this mob of law offices huddled around the corpse of a business that hasn’t been in operation for years (Alpari)? Additionally, if the suit is about a failure to disclose last-look functionality, isn’t that distinct from the merits of last-look itself?

Anyhow, I think we all want a better more fair marketplace and want to shine a light on any sort of abusive practices. I just wanted to share my perspective with you as it seemed relevant to your article. Hopefully the lawsuit (and your article) help to encourage the market to move towards greater transparency and fairness for all participants.