Nanex 04-Apr-2014 ~ Direct vs SIP Data Feed
The animation below shows how a trade or quote sent on an exchange makes its way
to the SIP (Securities Information Processor) and a Direct Feed used by High Frequency Traders (HFT).
Reg NMS requires that an exchange (A) make its data available to the
SIP (C) at the same time or earlier than to direct feed subscribers (B). Here's the relevent text from
Rule 603(a)(2) requires that any SRO, broker, or dealer that distributes market information must do so on terms that are not unreasonably discriminatory. These requirements prohibit, for example, a market from making its "core data" (i.e., data that it is required to provide to a Network processor) available to vendors on a more timely basis than it makes available the core data to a Network processor.
This is the same rule that NYSE broke and was
fined $5M by the SEC in September 2012. We have a
nice write up summarizing this fine as it applies to the SIP. Unfortunately,
this practice continues at other exchanges. In the animation below, note that the information
sent to the SIP has to travel significantly farther distances (40 miles vs 1000 feet),
on a slower network (1 GBps vs 40 GBps) with a protocol that adds more latency (TCP
vs UDP) than the same information sent to the direct feed. Sometimes this latency on
the input side of the SIP shows up in SIP data as fantaseconds (a term we coined to
describe trades printing before quotes). See this
well documented example.
The animation also shows something that many aren't aware of: the original timestamp
gets stripped, and replaced with a fresh timestamp when the SIP transmits it to a subscriber!
Watch the timestamp in the box get stripped when it enters, and replaced when it leaves,
the circle labeled "SIP Tape A". It's important to know that the timestamping mechanism
used by the SIP isn't very accurate: we
documented one instance where it was found to drift by as much as 16 milliseconds
- which means the "fresh" SIP timestamp could actually be earlier than the original
event. Keeping original timestamps is crucial for constructing
audit trails, or for detecting system delays, which is why it's integral to
this solution which allows HFT and non-HFT to coexist.
Understanding the Animation
Reg NMS requires that exchanges provide core data (trades and quotes) to the SIP as
fast or faster than direct feeds. In the animation above, that means a trade or quote
originating at an exchange (labeled A) must arrive at the SIP (C) no later than it arrives
on a direct feed (B).
The animation starts with a trade (or quote) in the symbol "F", timestamped by the exchange
at exactly 9:45. One network sends it to direct feed recipients (B) which are all 1000
feet away, and the other network sends it to the SIP (C) which is about 40 miles away.
When the trade arrives at the SIP, the timestamp is removed and aggregated with trades
from other exchanges (not
shown). Finally, at the point where the SIP transmits the
trade to a SIP subscriber (blue circle) it gets a new timestamp based on the SIP clock
(which could be ahead or behind the original exchange's clock - see this
documented example of a 16 millisecond drift).
This new timestamp, in effect, will hide any delays between the exchange (A) and the
blue circle from SIP users. During the flash crash,
delays of over 30 seconds occurred in many stocks and were impossible to detect,
because of the altered timestamps. Had the original timestamps from the exchange remained,
everyone, not just High Frequency Traders, would have been aware they were trading on
Sending data faster to HFT is against regulations. Period. Changing timestamps is just
plain wrong, and one could argue that SIP subscribers are being denied the true timestamp
that HFT enjoy. Keep in mind that nearly all brokerage reports on trade execution quality use the
SIP and therefore, an altered timestamp.
Nearly 40% of trades are priced based on the SIP - this include practically all retail
trades and most dark pools. Even Goldman Sach's Dark Pool is executes based on SIP prices.
When SIP prices lock or cross (due to slow input from one of 13 exchanges or
even FINRA), it causes internalizers (retail trades) to stop trading until the
condition is resolved: a phenomenon
made clear during the flash crash.
Furthermore, speed differences between the direct feed and the SIP can lead to other
undesirable market behaviors, such as momentum ignition, which have become
quite common (we detailed one market-wide momentum ignition event
here). Simply put, when prices suddenly move, those who can act earlier
(HFT using direct feeds), will profit at the expense of those who cannot, such as internalizers
and Dark Pool that are based on the SIP. With nearly 40% of trading based on the SIP,
the profitability of this manipulative strategy can be great enough that the cost of
inducing price shifts (momentum ignition) is worth the economic and regulatory risk.