Why is mid-point a “fairer” price?

Mid-point is a “fairer” price. Why? At first glance it appears mid-point is fair because both the buyer and seller get price improvement. Sure, both the buyer and seller get price improvement but that can’t be the sole criteria on which to judge fairness. There are many cons at the same time (Figure 1). The mid-point and HFT abuse narrative reminds me of the controversy and problems surrounding the (gaming of) early “Pegging” algorithms. Leaving block trades out of the discussion, the narrative goes that HFTs have an unfair advantage because they can “pick off” non-block institutional orders because of their speed advantage. They have a distinct speed advantage – they have direct feeds and can respond to changes faster.

Can a Limit Order be “Picked-off?”
I have a hard time understanding how an institutional order, or any order for that matter, can be “picked off” if it is passively resting at a limit. Let’s assume that the market is 24-25. You are the 24 bid. HFT sees the market go 24 then 23 offer in a direct feed so they hit the 24 bid and then buy it back at 23. It is important to note that at the time the HFT hit the bid and went 23 bid, the HFT was not assured at any point of covering the sale at a better price than 24. It is a risk that they are taking. Nevertheless, in this case, it is hard to argue that the resting 24 bid got picked off – they got done at their price. It isn’t as if the 24 bid was going to move to 23 – it was a 24 buyer.

Mid-Point
For the price of price improvement (sharing the spread with the contra side), mid-point opens itself up to all sorts of abuse. The mid-point is a sliding price. Let’s assume that the market is 24-25. You are a mid-point order with a top of 25. So you are willing to buy at 24.5. An algorithm can “probe” – send a sell order at 24.5 with the intent of trying to discover if there is a mid-point order. (In full disclosure, in 2005, Bloomberg Tradebook filed and was subsequently granted a patent for methods of finding and extracting hidden liquidity. “System and Method for Trading Financial Instruments Based on Undisclosed Values” US Patent 8099352). It is like the game of Battleship – if the execution comes back, “It’s a hit!” The narrative then goes, the electronic market maker knows from experience that it has a speed advantage and that the executing venue uses the slower SIP feed to price orders. So, when the direct feed goes 23 offer, the electronic market maker hits the (now stale) 24.5 bid before the venue has the opportunity to update and be a 23.5 buyer. The electronic market maker buys at 23 and makes the spread.

It’s not the presence of data latencies that are the problem but the mid-point order type and philosophy that a mid-point is a fairer price that is flawed.

Gaming Early Pegging Algorithm
A mid-point order is a peg – its execution price floats and is “pegged” to the movements of the reference price (NBBO). The issues with mid-point are similar issues that Bloomberg Tradebook had with the early forms of Pegging algorithms (1997). “Anti-gaming” enhancements were (rapidly) made.

The Pegging algorithm enabled clients to place passive orders in the market and have them automatically re-price to the NBBO. The algorithm had detection issues early on. Another algorithm could watch the feed or a human trader could watch the screen and when a new price level was established (displayed), the algo would automatically re-price. During illiquid periods (like pre- and post- market trading) market makers or other algorithms could “walk” the market up, in the process, find the pegging algorithm’s top, and then hit the bid at the higher level and cover when the market came back down. Three anti-gaming randomizations were added to reduce detection:

(1) Random time frequency (after the data feed changed) for re-pricing was introduced to disrupt pattern matching;
(2) Display size was randomized to make the re-priced order appear as a new order entering the market; and
(3) Lit/hidden was toggled.

These are just three examples of the early “anti-gaming” techniques that were added to make pegging a more effective (less detectable) strategy.

Although mid-point does not have a display component, the problems still lie in the peg’s re-price frequency. The re-price mechanism cannot be randomized or toggled; by its very nature the re-price needs to be instantaneous (or it won’t be at the mid) and its accuracy is dependent upon data feed latency. Non-block transactions can still be discovered through probing; illiquid stocks can be “walked” because of a potential lack of control over the “mid-point price strike” mechanism.

Is Mid-Point a Fairer Price?
At first glance, when I heard that mid-point was fairer, I thought of my kids and their early sport teams where “everyone is a winner.” Both sides get price improvement! In addition to being opened up to abuse, there are many “cons” to the fairness argument (Figure 1).

(non-block) Mid-point PROS

(non-block) Mid-point CONS

Everyone is a winner price – Both sides get price improvement.

Reduces market transparency: Mid-point is a non-displayed order; Mid-point orders get done at the “expense” of participants willing to set or display a price. It is a form of (sub) penny jumping; Loss of control: Based off a “sliding” reference price; Highly discoverable: Price is immediately reactive – similar to automatic displayed pegging strategies; Open to gaming: Pricing is dependent on data feeds and can be “gamed” from slow feeds or reference price movement.

Figure 1: Pros and Cons of Mid-point Pricing

Public Policy, Mid-Point and “Penny Jumping”
In the case of a penny spread, it could be argued that the SEC has already ruled from a public policy standpoint that mid-point is not in the best interest of the market. The SEC in Regulation NMS banned sub-penny quoting. For many of the same reasons it should be banned in non-block sub-penny executions.

The main issue is that I see mid-point as another form of “penny jumping” – an issue the SEC addressed in Regulation NMS with the ban on sub-penny quoting. Mid-point orders are priced off the backs of the orders that are willing to display their price – the people willing to be the reference price. Isn’t that unfair how they are able to “sub-penny jump” those orders? What is fairer is when everyone sets a whole penny limit as their price. Declare your intentions and establish the queue. The SEC made a mistake by not banning all sub-penny executions (except for average price and blocks) in Regulation NMS. They should consider coming back now and finishing the job.

What To Do Now?
In the meantime, users of mid-point strategies should consider that “gaming” typically occurs when the mid-point order accepts de minimus (100) share size executions. Low risk 100-share probes described above can be used to find and detect the presence of such orders. Raise the discovery risk by placing a min-fill restriction on the order (greater than the average size). This is a way to try and remain undetected and subsequently pushed around. Analytics like Tradebook’s Strategy analyzer (STAZ) can provide guidance to traders on “optimal” min fill size (Figure 2).