The U.S. Securities and Trade Fee (SEC) is publicly contemplating modifications to the Order Safety Regulation (OPR, Rule 611).
We lately investigated how usually the market truly locks and crosses. It was uncommon and really short-lived.
At present we’ll have a look at how usually trades are printed on the Securities Info Processor (SIP) at costs that look like trade-throughs. That is additionally uncommon, however might be primarily because of the period of time it takes to route the transaction to SIP.
What are you speaking about?
For those who have a look at SIP, you’ll be able to usually see that after the Nationwide Greatest Bid and Supply (NBBO) has modified, trades are arriving on tape on the outdated (outdated) worth.
Theoretically, if the NBBO is “protected”, you can’t commerce at a worse worth than the NBBO. So what is going on on?
Chart 1: In lots of markets, trades attain consolidation tape after BBO change
The diagram above from a research in Europe reveals what we’re speaking about. The research discovered that many of the trades had been latency arbitrage alternatives, with orders despatched to darkish swimming pools being routed utilizing quicker microwave hyperlinks.
Nonetheless, this could additionally happen because of the physics of routing orders. This meant that when the BBO was up to date, the midpoint fill was “moved” to the unified tape. This is without doubt one of the the explanation why pre-trade integration tape is unattainable resulting from physics.
Solely a part of the midpoint is printed after NBBO is modified
The primary query we ask is: How usually does one thing like this truly occur in the USA?
And the reply is nearly none.
As the information in Determine 2 reveals, roughly 19.4% of all off-exchange trades happen on the midpoint. Nonetheless, solely 0.5% are on the “outdated” midpoint, the place the NBBO has already modified, and are hardly ever reported outdoors of the NBBO.
Determine 2: Midpoint trades account for nearly 18% of all off-exchange trades, with solely 0.04% printed by way of NBBO
Have been these trades made earlier than the darkish pool confirmed the NBBO modifications?
The following query is how “outdated” are these commerce prints?
For those who monitor the output of all commerce reporting capabilities that occurred within the “outdated” NBBO Mid after the NBBO was modified, you will note that the majority of them arrive inside 2,500 microseconds (2.5 milliseconds). To place this into perspective, that is all very quick (our eye blinks are 250 milliseconds), however not quicker than the velocity of sunshine.
Graph 3: Distribution of “stale” midpoints from off-exchange venues
The truth that exercise will increase instantly after a market change appears to substantiate that the darkish pool midpoint is being actively investigated for liquidity concurrently the sunshine market is being eliminated.
You could keep in mind our earlier work the place we confirmed how buying and selling accelerates between venues (and SIPs). There, we noticed trades and quote updates hustling, transferring, and reacting for about 1 millisecond. Due to this fact, it’s fairly doable that a number of the “outdated prints” had been truly traded earlier than the darkish pool discovered of the NBBO modifications.
Contemplate the next instance.
Focusing solely on the Nasdaq itemizing implies that transactions will must be reported to Carteret’s SIP.
We assume that almost all darkish pool transactions happen within the Secaucus knowledge middle (Kind ATS-N signifies this to be true). Additionally assume that every one orders and SIP messages are despatched over fiber (that is essentially the most conservative). Lastly, to simplify this thought experiment, let’s assume that the darkish pool is utilizing a SIP feed for NBBO updates slightly than a direct feed (I do know this is not true in all instances, however this creates the longest response window).
When SIP NBBO is up to date, it takes roughly 143 microseconds over fiber for the NBBO replace to reach on the Secaucus darkish pool and replace the median worth throughout the darkish pool.
If a transaction occurred simply earlier than that, it might take a further 143 microseconds on the fiber for that “sluggish” transaction to reach at Carteret’s SIP. Additionally, calculation time should be thought of. As a information, SIP takes roughly 15 microseconds to course of a quote change.
This represents a spherical journey time of roughly 300 microseconds, shaded blue within the graph above. You possibly can see that the window accommodates many late-arriving trades, however not practically all.
Troublesome to find out if stale transactions happen and there’s latency arb
In actuality, all of those outputs are a lot quicker than the period of time a darkish pool can wait to submit a commerce (10 seconds).
Nonetheless, it’s clear that some studies point out that NBBO updates have been despatched to the darkish pool, processed, and returned effectively past the 300 microseconds it takes for them to return again.
What we’re seeing is that even with OPR in place, we find yourself with stale tradeprints. It’s troublesome to find out whether or not latency arbitrage is going on.
