The SEC’s decision may not be the end of the fight. Last month, attorneys for Nasdaq argued that the SEC could be sued if it approves IEX, saying the SEC would first have to change its own rules to explicitly allow for a speed bump.
To this, the SEC issued an interesting response: addressing concerns about the legality of speed bumps, the SEC separately said that delays of less than one millisecond are consistent with its Regulation NMS.
According to the SEC [1], IEX's 350 microseconds delay is negligible, and thus the market is automated and the quote is protected. The practical interpretation is that the SEC has set a ceiling for what it deems the speed race among HFT firms (with fiber optics, microwaves, lasers...)
So with this decision, the SEC has capped what technological advancement in trading can achieve going forward, as now a 350ms delay will become the norm, while anything below 1 millisecond is deemed a "de minimis" delay. Which is no good for HFTs, where microseconds can mean all the difference between profit and loss.
The IEX delay is uniform and deterministic, so if you're faster than your competitors by 1 microsecond in sending an order to IEX in response to an external event, you still win. The race that you can't win anymore is with IEX's pegged orders, which re-price without the delay (but they're always non-displayed).
To add to your comment here, IMHO the situation is now worse from a technical perspective on multiple fronts:
* IEX is still FIFO, as you point out about the race in response to external events. If Katsayuma really wanted to wipe out latency arb, he should have introduced some randomization and batching into the matching engine. I'm not sure how that would hold up from a regulatory perspective though.
* IEX quote updates will be 350us slower than any other update because of the additional input latency, which means that their protected quotes will always lag behind the market on a price flip. The amount of time that a price level is still protected on the SIP, yet practically gone, will increase as a result. This will increase the amount of time that DAY ISO orders are necessary for establishing the new price level, and therefore will benefit the HF shops which can leverage DAY ISOs the best. This also possibly creates a moral hazard for noncompliance.
* Other exchange owners have expressed interest in adopting a shoebox-like system, because of the expectation that delayed quotes will _force_ more market participants to trade on that exchange. If multiple systems start generating lagged quotes, the first problem continues to get worse.
* IEX's pegged order type relies on the SIP. But now that IEX itself will become a part of the SIP, the shoebox delay will have a feedback effect of lagging its own peg updates. Likewise, if other updates start delaying their output, IEX pegs may become less useful. I'm not exactly sure what the net effect of this will be (after a few years of other exchanges adapting)... but I don't expect it to be "good" for the system as a whole.
In general, I believe that artificial introduction of latency into the National Market System seems only to serve special interests: advanced technological traders (usually HFT itself), and the exchanges which enact the artificial latency.
Comments
The SEC’s decision may not be the end of the fight. Last month, attorneys for Nasdaq argued that the SEC could be sued if it approves IEX, saying the SEC would first have to change its own rules to explicitly allow for a speed bump.
To this, the SEC issued an interesting response: addressing concerns about the legality of speed bumps, the SEC separately said that delays of less than one millisecond are consistent with its Regulation NMS.
According to the SEC [1], IEX's 350 microseconds delay is negligible, and thus the market is automated and the quote is protected. The practical interpretation is that the SEC has set a ceiling for what it deems the speed race among HFT firms (with fiber optics, microwaves, lasers...)
So with this decision, the SEC has capped what technological advancement in trading can achieve going forward, as now a 350ms delay will become the norm, while anything below 1 millisecond is deemed a "de minimis" delay. Which is no good for HFTs, where microseconds can mean all the difference between profit and loss.
[1] https://www.sec.gov/divisions/marketreg/automated-quotations...
The IEX delay is uniform and deterministic, so if you're faster than your competitors by 1 microsecond in sending an order to IEX in response to an external event, you still win. The race that you can't win anymore is with IEX's pegged orders, which re-price without the delay (but they're always non-displayed).
To add to your comment here, IMHO the situation is now worse from a technical perspective on multiple fronts:
* IEX is still FIFO, as you point out about the race in response to external events. If Katsayuma really wanted to wipe out latency arb, he should have introduced some randomization and batching into the matching engine. I'm not sure how that would hold up from a regulatory perspective though.
* IEX quote updates will be 350us slower than any other update because of the additional input latency, which means that their protected quotes will always lag behind the market on a price flip. The amount of time that a price level is still protected on the SIP, yet practically gone, will increase as a result. This will increase the amount of time that DAY ISO orders are necessary for establishing the new price level, and therefore will benefit the HF shops which can leverage DAY ISOs the best. This also possibly creates a moral hazard for noncompliance.
* Other exchange owners have expressed interest in adopting a shoebox-like system, because of the expectation that delayed quotes will _force_ more market participants to trade on that exchange. If multiple systems start generating lagged quotes, the first problem continues to get worse.
* IEX's pegged order type relies on the SIP. But now that IEX itself will become a part of the SIP, the shoebox delay will have a feedback effect of lagging its own peg updates. Likewise, if other updates start delaying their output, IEX pegs may become less useful. I'm not exactly sure what the net effect of this will be (after a few years of other exchanges adapting)... but I don't expect it to be "good" for the system as a whole.
In general, I believe that artificial introduction of latency into the National Market System seems only to serve special interests: advanced technological traders (usually HFT itself), and the exchanges which enact the artificial latency.