Note: Descriptions are shown in the official language in which they were submitted.
81800301
SYSTEMS AND METHODS FOR QUANTIFYING TEMPORAL
FAIRNESS ON ELECTRONIC TRADING VENUES
CROSS-REFERENCE TO RELATED APPLICATIONS
[001] This application claims priority to U.S. Provisional Patent Application
No.
62/090,568, filed December 11, 2014, entitled "Method for Quantifying Temporal
Fairness on
Electronic Trading Venues," and is a continuation-in-part application of U.S.
Patent
Application No. 14/533,543, tiled November 5, 2014, entitled, "Ideal Latency
Floor".
FIELD OF THE INVENTION
[0021 The invention specifies systems and methods for quantifying temporal
fairness on an
electronic trading venue by analyzing a population of time deltas that
indicate whether, and/or
by how much time, a given market participant has beaten at least one other
market participant
in sending (or being sent) similar messages to (or from) the venue.
BACKGROUND OF THE INVENTION
[003] Operators of electronic trading venues have obligations to ensure the
markets they
operate are fair. While fairness in the context of financial markets is a
broad and multifaceted
concept, on an electronic trading venue implementing time-priority based rules
for processing
orders (or more generally messages), one important form of fairness that
manifests in the
relative times at which participants are sent (or themselves send) similar
data to and from the
venue is temporal fairness.
1004] In general, a financial market may be considered (or defined) to be
temporally fair if
all participants in that market have the opportunity to both receive and act
upon information
material to it at substantially the same time. Roughly speaking, and relating
to the "receiving"
part of this general definition, an electronic trading venue may be considered
temporally fair if
no market participant is habitually being sent market data updates before any
other
participant.
1
CA 2946578 2018-03-22
CA 02946578 2016-10-27
PATENT
Attorney Docket No. 45XP-215265
Also roughly speaking, and relating to the -acting upon" part of this general
definition, an
electronic trading venue may be considered to be temporally fair if delays
associated with the
venue's processing of each message it receives do not cause one participant's
orders to habitually
"overtake- another's. In this context an order is said to "overtake" another
if the former was
received by the venue earlier than the latter, but the latter was ultimately
processed by the venue
(e.g., against the instrument's limit order book or other similar structure)
before the former.
Temporal fairness oftentimes is elusive due to the nature of communication
networks, such as
various network latencies or other network performance issues that result from
modern
communication networks.
[005] Ensuring electronic trading venues achieve temporal fairness as it is
defined above
requires careful design and implementation of the software and hardware
components
comprising the venue. However, careful design and implementation alone are not
enough to
prove the venue is actually achieving temporal fairness. Rather, proving the
venue is temporally
fair requires taking appropriate measurements of actual data flowing in and
out of it (usually at
the network-level), and employing an appropriate methodology to analyze that
data. Further, this
measurement and analysis process should be performed repeatedly and preferably
regularly, over
time, to ensure that changes to the venue's software and hardware components
(such as
upgrades, addition or removal of participants/instruments, reconfigurations
and so on) do not
degrade the venue's temporal fairness.
[006] Conventional approaches to measuring temporal fairness suffer a number
of drawbacks.
Some such approaches use techniques and measurements that are thought¨though
perhaps not
theoretically or empirically proven to be correlated with actual temporal
fairness. One such
approach may involve identifying a group of participants all of whom have
similar response
times, and comparing their individual successes in winning -races" to take (or
aggress) prices
when those participants compete with one another on the venue. Another such
approach may
similarly involve identifying a group of participants, and then attempting to
ascertain which are
advantaged over others by counting the number of times each receives price
updates before the
others. These conventional approaches seem to quantify temporal fairness in
terms of
participants, by identifying which such participants are in some way(s)
temporally advantaged
over others. They do not however seem to quantify temporal fairness as a
scalar measurement
2
2243496NA
CA 02946578 2016-10-27
PATENT
Attorney Docket No. 45XP-215265
with unit time.
[007] For many reasons it is desirable to be able to quantify temporal
fairness as a scalar
measurement, having unit time. By having such a measurement (with unit time),
a precise delay
period for a mechanism such as a latency floor that "batches up" and
(effectively) reorders
messages flowing into and out of the venue may be inferred. Such a mechanism,
its delay period
being informed by the value of that measurement, may provide temporal fairness
on a venue that
would otherwise be considered temporally unfair. Further, such a scalar value
may
unambiguously be used to determine if temporal fairness has improved, worsened
or remained
steady, e.g., by comparing the relative magnitudes of two such measurements.
[008] These and other drawbacks exist with conventional systems for
quantifiably measuring
fairness in electronic trading venues, thereby motivating the invention
described herein.
SUMMARY OF THE INVENTION
[009] The invention addressing these and other drawbacks relates to systems
and methods for
quantifying temporal fairness on an electronic trading venue by analyzing a
population of time
deltas that indicate whether, and/or by how much time, a given market
participant has "beaten"
at least one other market participant when those participants send (or are
sent) similar messages
to (or from) on the venue.
[010] For instance, the system may obtain market activity information
associated with a first
market participant and a second market participant. One such form of market
activity
information may indicate messages received by the venue from market
participants relating to an
instrument traded on the electronic trading venue. Such messages may relate
to, without
limitation, orders (e.g., new order requests, cancel requests, replace
requests, etc.). Another such
form of market activity may indicate messages sent by the venue to the
participants relating to an
instrument traded on the electronic trading venue. Such messages may relate
to, without
limitation, market data updates (e.g., credit screened snapshots of the
instrument's limit order
book, unscreened snapshots of it, etc.).
[011] The system may further determine, based on the market activity
information, a first
plurality of time deltas that each indicate an amount of time in which the
first market participant
3
224349600_1
CA 02946578 2016-10-27
PATENT
Attorney Docket No. 45XP-215265
has temporally beaten the second market participant with respect to a form of
market activity
related to an instrument traded on the electronic trading venue.
[012] The system may, based on the market activity information, a second
plurality of time
deltas that each indicate an amount of time in which the second market
participant has
temporally beaten the first market participant with respect to that same form
of market activity
related to the instrument traded on the electronic trading venue.
[013] The system may determine a quantified level of temporal fairness as a
scalar value having
unit time with respect to the first market participant and the second market
participant on the
electronic trading venue based on the first plurality of time deltas and the
second plurality of
time deltas. This scalar value with unit time determined by the system may be
such that the
magnitude of the value indicates the degree of temporal unfairness, with
larger values indicating
a higher degree of temporal unfairness and smaller values indicating a lower
degree of temporal
unfairness. A scalar value of zero may indicate temporal unfairness has been
eliminated and thus
temporal fairness has been achieved.
[014] The system may, in analyzing a first and second plurality of time deltas
for a pair of
participant's market activity on an instrument relating to market data update
messages, determine
a minimum value for "fmkidaia", given finktda,a > 0, such that the number of
values strictly greater
than fdata in the first plurality of time deltas plus "X" is approximately
equal to the number of
values strictly greater than f
--rnictdata in the second plurality of time deltas plus "X". Typically, the
value of "X- will be determined by counting the number of values across the
both pluralities of
time deltas that are less than or equal to f
-mktdata> and halving that count.
[015] The system may, in analyzing a first and second plurality of time deltas
for a pair of
participant's market activity on an instrument relating to order messages,
determine a minimum
value for "fords", given fords 0, such that the number of values strictly
greater than f
-ords in the
first plurality of time deltas plus "Y" is approximately equal to the number
of values strictly
greater than fords in the second plurality of time deltas plus "Y". Typically,
the value of "Y" will
be determined by counting the number of values that are less than or equal to
fords across a third
and fourth plurality of time deltas related to that same pair of participants
and instrument, and
halving that count.
[016] The system may, on a per instrument basis, determine a quantified level
of temporal
4
224349600_1
81800301
fairness with unit time for an instrument on that venue, by finding a single
minimum fmkrdara
value and separately a single minimum fords value such that the criteria
described above is met
for each pair in a plurality of pairs of participants on that instrument (and
not just a single pair
as described above), and summing those two minimum "f' values together. This
sum may
subsequently be used to inform the "floor value" or "delay period" of an Ideal
Latency Floor
or other such mechanism deployed on the venue for that instrument.
Alternatively, for
simplicity, the maximum such sum across a plurality of instruments on the
venue may be used
to inform the value of an Ideal Latency Floor, and that maximum sum may
reflect the
temporal fairness of the venue as a whole.
[01 6a] According to one aspect of the present invention, there is provided a
method of
quantifying temporal fairness on electronic trading venues, the method being
implemented on
a computer system comprising a cluster of commodity computing hardware having
one or
more physical processors programmed with computer program instructions that,
when
executed by the one or more physical processors, cause the computer system to
perform the
method, the method comprising: obtaining, by the computer system, market
activity
information associated with a first market participant and a second market
participant;
determining, by the computer system, based on the market activity information,
a first
plurality of time deltas that each indicate an amount of time in which the
first market
participant has temporally beaten the second market participant with respect
to market activity
related to an instrument traded on the electronic trading venue; determining,
by the computer
system, based on the market activity information, a second plurality of time
deltas that each
indicate an amount of time in which the second market participant has
temporally beaten the
first market participant with respect to the market activity related to the
instrument traded on
the electronic trading venue; determining, by the computer system, a
quantified level of
temporal fairness with respect to the first market participant and the second
market participant
on the electronic trading venue based on the first plurality of time deltas
and the second
plurality of time deltas; batching, by the computer system, a set of orders
received within a
time period determined based on the quantified level of temporal fairness;
randomly selecting,
by the computer system, an order from the batched set of orders; and
processing, by the
CA 2946578 2020-01-20
81800301
computer system, the randomly selected order before remaining orders in the
batched set of
orders.
[01613]According to another aspect of the present invention, there is provided
a system of
quantifying temporal fairness on electronic trading venues, the system
comprising: a computer
system comprising a cluster of commodity computing hardware having one or more
physical
processors programmed with computer program instructions that, when executed
by the one
or more physical processors, cause the computer system to: obtain market
activity information
associated with a first market participant and a second market participant;
determine, based on
the market activity information, a first plurality of time deltas that each
indicate an amount of
time in which the first market participant has temporally beaten the second
market participant
with respect to market activity related to an instrument traded on the
electronic trading venue;
determine, based on the market activity information, a second plurality of
time deltas that
each indicate an amount of time in which the second market participant has
temporally beaten
the first market participant with respect to the market activity related to
the instrument traded
on the electronic trading venue; determine a quantified level of temporal
fairness with respect
to the first market participant and the second market participant on the
electronic trading
venue based on the first plurality of time deltas and the second plurality of
time deltas; batch a
set of orders received within a time period determined based on the quantified
level of
temporal fairness; randomly select an order from the batched set of orders;
and process the
randomly selected order before remaining orders in the batched set of orders.
[017] These and other objects, features, and characteristics of the system
and/or method
disclosed herein, as well as the methods of operation and functions of the
related elements of
structure and the combination of parts and economies of manufacture, will
become more
apparent upon consideration of the following description and the appended
claims with
reference to the accompanying drawings, all of which form a part of this
specification,
wherein like reference numerals designate corresponding parts in the various
figures. It is to
be expressly understood, however, that the drawings are for the purpose of
illustration and
description only and are not intended as a definition of the limits of the
invention. As used in
the specification and in the claims, the singular form of "a", "an", and "the"
include plural
referents unless the context clearly dictates otherwise.
5a
CA 2946578 2020-01-20
= .
81800301
BRIEF DESCRIPTION OF THE DRAWINGS
[018] FIG. 1 illustrates an exemplary system for quantifying fairness of an
electronic
trading venue, according to an implementation of the invention.
[019] FIG. 2 depicts an exemplary temporal fairness analyzer, according to an
implementation of the invention.
[020] FIG. 3 depicts a process for quantifying fairness of an electronic
trading venue,
according to an implementation of the invention.
[021] FIG. 4 is a schematic diagram that depicts quantifying temporal inter-
component and
intra-component temporal fairness on an electronic trading venue, according to
an
implementation of the invention.
5b
CA 2946578 2020-01-20
CA 02946578 2016-10-27
=
PATENT
Attorney Docket No. 45XP-215265
DETAILED DESCRIPTION OF THE INVENTION
[022] The invention described herein relates to a system and method for
quantifying temporal
fairness on an electronic trading venue. From the perspective of causation, an
electronic trading
venue may be considered to be temporally fair if properties of the components
comprising the
venue (and therefore under the control or ownership of that venue's operator)
do not themselves
cause systemic bias in determining which participants win races to "make" or
"take" prices on
the venue. Such systemic bias can manifest in market data distribution,
particularly if one market
participant is habitually being sent their market data before another. Such
systemic bias can also
separately manifest in the processing of order messages received by the venue,
particularly if one
market participant's orders are habitually "overtaking" another's, i.e., the
former participant's
orders being received by the venue earlier than the latter participant's, but
the latter's being
processed for matching by the venue before the former's.
[023] Mathematically, an electronic trading venue may be considered to be
temporally fair if
the following two conditions¨one relating market data distribution, the other
to the venue's
receipt and ultimate processing of orders ___________________________ are met
when applied all pairs of participants on each
instrument that trades on the venue. Relatina, to market data: for each pair
of participants on each
instrument, the number of times the first participant receives (or is sent) a
market data update
before the second should be approximately equal to the number of times the
second participant
receives (or is sent) a market data update before the first. Relating to
orders: for each pair of
participants on each instrument, the number of times an order sent by the
first participant
overtakes one sent by the second, should be approximately equal to the number
of times an order
sent by the second participant overtakes one sent by the first. It is to be
noted that both of these
conditions can be only "true" or "false", and do not quantify the extent to
which temporal
(un)fairness exists.
[024] The systems and methods described herein may quantify temporal fairness
as a scalar
value with unit of time on an electronic trading venue. When this scalar value
is larger in
magnitude it may indicate a higher degree of temporal unfairness; a scalar
value of zero may
indicate temporal fairness has been achieved. Additionally, upon finding a non-
zero scalar value
for temporal fairness on a venue, a venue operator may choose to deploy an
Ideal Latency Floor
6
224349600_1
CA 02946578 2016-10-27
PATENT
Attorney Docket No. 45XP-215265
on the venue with the "floor value" set to that non-zero scalar value (or
larger). to rectify the
venue's temporal unfairness.
[025] The systems and methods described herein may specify, on a per
instrument basis, that
two minimum values¨one for orders "fords", and separately one for market
dataf
"-mktdata"¨
be found such that the two equations set forth below are satisfied when
applied to each pair of
participants selected from a plurality of them on the instrument.
[026] Pertaining to market data, for a pair of participants on an instrument,
given f.õ1.
-÷mtdata > 0, the
systems and methods described herein may identify a minimum value for fmir A
--t-ata be found such
that the number of values strictly greater than f
mktdata in a first plurality of time deltas plus "X" is
approximately equal to the number of values strictly greater than fmladasa in
a second plurality of
time deltas plus "X". Typically, though not necessarily, the value of "X" will
be determined by
counting the number of values across both that first and second plurality of
time deltas that are
less than or equal to fmkidata, and halving that count. The construction of
these pluralities of "time
deltas" for market data is specified by the invention and discussed
subsequently herein.
[027] Pertaining to orders the invention specifies that for a pair of
participants on an
instrument, given fords > 0, the systems and methods described herein may
identify a minimum
value for fords such that the number of values strictly greater than fords in
a first plurality of time
deltas plus "Y" is approximately equal to the number of values strictly
greater than fords in a
second plurality of time deltas plus -Y". Typically, though not necessarily,
the value of "Y" will
be determined by counting the number of values that are less than or equal to
fords across both a
third and fourth plurality of time deltas related to that same pair of
participants on that
instrument, and halving that count. The construction of these pluralities of -
time deltas" for
orders will be described herein.
[028] Pertaining to market data, the first and second plurality of "time
deltas" for a pair of
participants on an instrument may be constructed as follows. At each market
data update for an
instrument, separate messages may be sent to each participant in the pair
reflecting a point-in-
time view of the state of the market (e.g., bids and offers) on that
instrument. If in that update,
the first participant is sent his before the second, then amount of time that
elapses between the
sending of the first participants message, and sending of the second's is
inserted into the first
plurality of time deltas. Alternatively, if the second is sent his message
before the first, then the
7
224349600_1
CA 02946578 2016-10-27
PATENT
Attorney Docket No. 45XP-215265
amount of time that elapses between the sending of those two messages is
inserted into the
second plurality of time deltas. The process of inserting values (specifically
differences in
elapsed time) into the pluralities of time deltas may be repeated over a
plurality of market data
updates; the selection of such updates will be discussed herein. Further,
since on some venues
each participant may be sent multiple messages even at a single update, the
systems and methods
described herein may identify which message is to be considered, which will be
described herein.
[029] Pertaining to orders, the first and second plurality of -time deltas"
for a pair of
participants on an instrument may be constructed as follows. If the venue
receives an order for an
instrument from the first participant before a similar such order for that
instrument from the
second, but the venue ultimately processes (e.g. against the instrument's
limit order book) the
second's order before the first's then the elapsed time between the processing
of the two orders is
optionally (depending on the specific embodiment) added to the elapsed time
between the receipt
of the two orders, and the resulting sum is inserted into the second plurality
of time deltas. When
the second participant's order is received before the first's similar such
order, but the first's is
ultimately processed by the venue before the second's, the analogous
computation occurs and the
resulting value is inserted into the first plurality of time deltas. The
process of inserting values
into the pluralities of time deltas may be repeated over a plurality of
orders, the selection of
which is specified by the invention and is discussed subsequently herein. The
systems and
methods described herein may further specify what it means for orders to be
"similar", which
will be described herein.
[030] Pertaining to orders, the third and fourth plurality of "time deltas"
for a pair of
participants on an instrument may be constructed as follows. If the venue
receives an order for an
instrument from the first participant before a similar such order for that
instrument from the
second, the amount of elapsed time between the receipt of those two orders (in
some
embodiments, additionally summing this value with, or instead using, the
amount of elapsed time
between their ultimate processing by the venue) is inserted into the third
plurality of time deltas.
When the venue receives an order for the instrument from the second
participant before a similar
such order from the first, the analogous computation occurs, and the resulting
value is inserted
into the fourth plurality of time deltas. The process of inserting values into
the pluralities of time
deltas may be repeated over a plurality of orders, according to the same
criteria chosen for the
8
224349600_1
CA 02946578 2016-10-27
PATENT
Attorney Docket No. 45XP-215265
selection of orders using for the first and second plurality of time deltas
for orders, as previously
described.
[031] Relating to the construction of market data time deltas above, the
systems and methods
described herein may receive from a user a selection of specific criteria,
such that in each market
data update there is a maximum of one such message sent to each participant
that is to be
considered based on the criteria. Non-limiting examples of criteria for
selecting such market data
updates, include but are not limited to, the first credit-screened view of the
limit order book sent
to a participant in an update, or the first unscreened such update sent to a
participant, and so on.
For the avoidance of doubt, in the case where a pair of participants is not
both sent a message
meeting the selected criteria within a given update, there will be no value
inserted into either's
plurality of time deltas relating to that update (and pair).
[032] Relating to the construction of order time deltas above, the systems and
methods
described herein may receive from a user a selection of specific criteria that
determine whether
two or more orders are "similar" to one another on an instrument. Non-limiting
examples of such
criteria include, but are not limited to: all orders sent on the instrument
regardless of their limit
price, side or quantity; or, orders of the same side (i.e., buy or sell) sent
on the instrument; or, the
first order sent by a participant in a specific race to "make" or "take" a
price on the instrument,
as the term "race" is used in U.S. Patent Application No. 14/533,543, filed
November 5, 2014,
entitled, "Ideal Latency Floor," and so on.
[033] Relating to both the construction of market data time deltas and the
construction of order
time deltas the specific orders or market data updates to be included in the
methodology may be
selected by the user of the invention according to specific criteria. Non-
limiting examples of
such criteria include, but are not limited to: events (orders or market data
updates) occurring over
a given time horizon such as an entire trading day, or an entire trading week,
or a particularly
busy period of time during the day over a number of days, and so on. All, or a
sampling of such
events (orders or market data updates), that occurred during the specified
time may be
considered.
[034] The term "market participant" (or simply "participant") is intended to
be broadly
construed to refer to any entity that receives (through a computing device)
market data from the
venue, or sends (through a computing device) order-related messages to the
venue, including. but
9
224349600_1
CA 02946578 2016-10-27
PATENT
Attorney Docket No. 45XP-215265
not limited to: a firm that conducts business on the electronic trading venue,
a credit entity
associated with such a firm (a single firm may have a plurality of credit
entities), or a user
(human or otherwise). In the foregoing text, the user of the invention will
select the specific form
of participant to which methodology will apply.
[035] Exemplary System Architecture
[036] FIG. 1 illustrates an exemplary system 100 for quantifying fairness of
an electronic
trading venue 110, according to an implementation of the invention. System 100
may include
electronic trading venue 110 (used interchangeably with -venue 110" for
convenience), one or
more market data distributors 130, one or more market participants 140, and/or
other
components.
[037] Venue 110 may include a credit engine 112, a matching engine 114, a FIX
gateway 116,
a switch 118, a network tap device 118, a temporal fairness analyzer 120, a
database 122, a
network traffic capturer 124, a market data distributor 130, and/or other
components. Market
data distributor 130 may each receive information about matches and the
Central Limit Order
Books (CLOBs) and credit, and sends, to market participants 140, a credit-
screened and
unscreened view of the CLOB for the instruments that trade on venue 110.
[038] Credit engine 112 handles credit relationships and credit lines among
market participants
140. Each matching engine 114 handles a subset of instruments traded on venue
110, and
maintains the CLOBs or other similar structure for those instruments. A
matching engine 114
causes two orders to match if they are of opposite sides (buy and sell), are
price compatible, and
if, according to credit engine 112, the two participants submitting the orders
have credit with one
another.
[039] FIX gateway 116 may each receive and send Financial Information eXchange
("FIX")
protocol messages such as new orders requests, execution reports, and so on.
More than one of
each of the above components of venue 110 (e.g., multiple market data
distributors 130, multiple
FIX gateways 116, etc.) may be used to achieve adequate performance (e.g.,
response times,
throughput, etc.) through distributed (e.g., parallel) computing.
[040] Market participants 140 (also referred to interchangeably as
"participants 140" for
convenience) are each entities that conduct business on venue 110 (represented
in system 100 as
devices used by such participants to interface with venue 110).
224349600_1
CA 02946578 2016-10-27
PATENT
Attorney Docket No. 45XP-215265
[041] The various components illustrated in FIG. 1 may communicate with one
another via one
or more communication links (represented by lines between such components).
The
communication links may include network links through a network described
herein. The
various communication links are shown for illustration and not limitation, as
one or more
communication links between components not otherwise illustrated may be used
as well.
[042] In an exemplary operation, market participants 140 send in orders
(including cancels,
amendments, etc.) to venue 110 over the network, and receive market data
updates from the
venue over the network. In either case, the traffic maybe hi-directional
(e.g., order-related traffic
may be sent from venue 110 to participants 140 as execution reports for
orders, and market data-
related traffic maybe sent from participants 140 to venue 110 as subscription
requests for specific
instruments). Order-related network traffic is routed by switch 118, which may
be implemented
as a conventional network switch device, to one or more of the FIX gateway
116. Market data-
related network traffic may be provided by market data distributor 130 and is
routed by switch
118 to participants 140.
[043] Network tap device 101 may allow network traffic to flow through it, but
also may
forward a verbatim copy of all the traffic that flows through it to the
network traffic capturer 124,
which may attach timestamps to each of the packets it receives in the network
traffic from the
network tap device 101. Network traffic capturer 124 may subsequently write
those packets, or
information contained within them, to venue database 122.
[044] Network traffic capturer 124 may perform additional processing on the
network traffic
before it is written to the venue database 122. For instance, the network
traffic capturer 124 may
extract from the network packets the application-level messages (market data
updates and FIX
messages, or certain fields contained within those messages) and write them to
separate files. In
some instances, network traffic capturer 124 may write one file per instrument
per trading
session.
[045] A mapping from IP addresses to individual participant 140 (since a
single participant may
use a plurality of IP addresses to connect to venue 110), and mappings from IP
addresses to
specific components in venue 110 (e.g., to a certain FIX gateway 116 or market
data distributor
130 may also be written to venue database 122. The IP address visible in the
network traffic at
the point in the network where the network tap device 101 is installed may be
different than the
11
224349600_1
CA 02946578 2016-10-27
PATENT
Attorney Docket No. 45XP-215265
IP address of a given component to which switch 118 is attached (e.g., the FIX
gateway 116, the
matching engine 114, a credit engine 112, or market data distributor 130),
because of Network
Address Translation (sometimes referred to as "NAT'ing) that the switch 118
may perform.
Further, and although not shown explicitly as a connection (line) on the
diagram above, matching
engine 114 may populate venue database 122 with the sequence in which they
actually received
messages (orders), noting that reordering of messages (orders) may occur as
network packets
flow through the switch 118 and FIX gateway 116 components up to the matching
engine 130
(and indeed other network devices and software that may exist anywhere between
switch 118
and matching engine 114 may also cause reordering).
[046] Temporal Fairness Analyzer 120
[047] FIG. 2 depicts an exemplary temporal fairness analyzer 120, according to
an
implementation of the invention. Temporal fairness analyzer 120 may be
configured as a server
(e.g., having one or more server blades, processors, etc.), a computer (e.g.,
a desktop computer, a
laptop computer, etc.), and/or other device that can be programmed to quantify
temporal fairness
of electronic trading venue 110. In an implementation, temporal fairness
analyzer 120 may be
configured as a cluster of commodity computing hardware programmed by various
computer
program instructions. For instance, the cluster may execute ApacheTm Hadoop
software. In
these implementations, database 122 may be implemented via the Hadoop
Filesystem (HDFS) on
this cluster and the instruction(s) that programs temporal fairness analyzer
120 may be
implemented to use the MapReduce API and HDFS-client API provided by Hadoop,
and in the
JavaTM programming language.
[048] Regardless of its particular implementation or configuration, temporal
fairness analyzer
120 may include one or more processors 212 (also interchangeably referred to
herein as
processors 212, processor(s) 212, or processor 212 for convenience), one or
more storage devices
214 (which may store various instructions that program processors 212), and/or
other
components. Processors 212 may be programmed by one or more computer program
instructions. The computer program instructions may include, without
limitation, a temporal
fairness quantifier application 220, a latency floor application 222, and/or
other instructions that
program temporal fairness analyzer 120 to perform various operations, which
are described in
greater detail herein. As used herein, for convenience, the various
instructions will be described
12
224349600_1
CA 02946578 2016-10-27
,
PATENT
Attorney Docket No. 45XP-215265
as performing an operation, when, in fact, the various instructions program
the processors 212
(and therefore temporal fairness analyzer 120) to perform the operation.
[049] Quantifyine Temporal Fairness in an Implementation
[050] In an implementation, a software program (e.g., temporal fairness
quantifier application
220) may be written in a high-level programming language such as JavaTM to
compute the f
mktdata
and fords values on a per instrument basis. The program may take as input
(e.g. via command line
or configuration file), the instrument(s) on which it is to be run, whether it
is to be run on orders
and/or market data, the time period over which it to be run, the criteria for
selection of market
data update messages and/or the criteria for identifying similar orders, and
the specific
participants of which all pairs of such will be subject to the analysis.
[051] The program may utilize a map data structure (e.g., java.util.HashMap)
to store, in main
memory, the pluralities of time deltas utilized by the methodology. For market
data analysis, the
keys of such this map may be the 3-tuple of [instrument, participantl,
participant2]. For order
analysis, the key may be the 4-tuple of [instrument, participant 1,
participant2, flag], where "flag"
taking the value of "true" indicates the corresponding value associated with
the key pertains to
time deltas for similar "overtaking" orders only, and where "flag" taking the
value "false"
indicates the corresponding value associated with the key pertains to time
deltas associated with
similar orders in general. The data structure used for the keys of the map may
be a list of strings
(e.g., instances of the class java.util.LinkedList<String>).
[052] Regardless of whether the previously described map is used for market
data and/or
orders, the value corresponding to each key will be participantl's time deltas
relative to
participant2's. The data structure used for the map's values may be a list of
integers (e.g..
java.util.ArrayList<Integer>). If the timestamps are captured as real numbers,
the program may
store the differences (i.e., time deltas) between them using as integers by
using a different
precision and scale, e.g., for reasons of space and performance. For instance,
a real number
difference expressed in seconds may be multiplied by 1,000,000 to obtain a
rounded, non-zero
integer difference in microseconds).
[053] As the program reads from the file or database containing data from the
network traces,
that data being either timestamped order messages and/or timestamped market
data messages,
sorted by timestamp, upon such a pair of messages meeting the criteria for
selection specified as
13
2243496003
CA 02946578 2016-10-27
PATENT
Attorney Docket No. 45XP-215265
input to the program, the previously described map will be referenced. In
particular, the
participants for the first and second messages will be discerned from each
message, either
through the IP address associated with the message's source or destination, or
through the
message's payload (e.g., the SenderCompId field on a FIX order message). The
instrument will
also be discerned, likely from the payload of the message itself (e.g., the
Symbol field on a FIX
order message). In this manner, all pairs of participants that sent or
received messages subject to
the analysis will be encountered. The first time such a pair of participants
is encountered for the
instrument and analysis type, the key will be inserted into the map along with
a new instance of
list containing the time delta. On subsequent times that pair of participants
for the analysis type
and instrument is encountered, through processing a later pair of messages,
the list (value) of the
map will be fetched from the map by way of the key, and the time delta added
to that existing
list.
[054] In some venue implementations the program may be able to determine which
market data
update messages belong to a given update through examination of the payload of
the message
(e.g., by looking at the "market by price sequence number" in a Reuters
Foundation API
message). Once all the messages for a given update have been read into main
memory, and
processed against the earlier described map, they may be discarded from main
memory so as to
ensure the program does not run out of memory.
[055] In some venue implementations the program may be able to determine the
order in which
the order messages were processed against the venue's limit order book by
examining the integer
id in the OrderId field of the FIX message response. In this manner, by
relating the inbound FIX
order messages to their outbound FIX response messages the program may be able
to determine
both the ordering in which the orders were received, and the ordering in which
they were
ultimately processed by the venue. Further, a message's timestamp along with
the Price, Side,
Symbol and SenderCompId fields of the (FIX) message's payload may enable the
program to
determine, for instance, which orders were similar to one another, per the
input parameter to the
program specifying the same.
[056] Upon completing the reading of the message data for the time period
specified, and
populating the map with all the time deltas associated with that period, the
program may attempt
to find a minimum "f" value(s) for orders and/or market data. For a given pair
of participants on
14
224349600_1
CA 02946578 2016-10-27
PATENT
Attorney Docket No. 45XP-215265
an instrument and for an analysis type (i.e., orders or market data) it may do
so by generating a
sorted set (e.g., using java.util.TreeSet) of all the unique values that
appear in that pair of
participants lists of time deltas, and adding to a zero value to that sorted
set. The program may
then iterate over that sorted set of unique time delta values and test each
one using the equations
for "approximate equality" described earlier in this document that reference
fords and fmktdata= This
process of "testing" each candidate f value from the sorted set involves
evaluating the left hand
side and the right hand side of the equation with each specific f value. This
testing process is
described further below.
[057] Since a computer does not inherently have a notion of "approximately
equal" per the
previously described equations involving fords and frniddata, in an
implementation a statistical test at
a predefined level of statistical confidence may be applied, to determine if
the two sides of the
equation are indeed approximately equal to one another. The statistical test
may compare two
"actual" values to two "expected" values for each "f" value tested, at a
predefined level of
statistical confidence. With regard to this test, the two "actual" values may
be those returned by
the right and left hand sides of the "approximately equal" equation,
respectively. The two
"expected" values passed into the statistical test may each be half the sum of
the "actual" values.
If the statistical test utilized (e.g., a chi-squared test) returns a value
greater than the predefined
level of statistical confidence (e.g. 0.01) then the f-value that generated
these "actual" values is a
candidate for the minimum f value, and it may be said that for this f value
the test for
approximate equality has "passed" (cf. failed).
[058] As the program iterates over all distinct time deltas for a pair of
participants on an
instrument for an analysis type, it may generate and store the ranges of f-
values which passed the
statistical test for approximate equality. Upon iterating over all pairs of
participants on each
instrument for an analysis type, and calculating ranges of candidate f-values
for each pair, the
program may find an overall minimum f value for that analysis type and
instrument, by
examining those ranges and finding the minimum value that falls within the
ranges of all such
pairs, noting that any f value bigger than the largest value found in a pair's
range is also
implicitly passes the statistical test for that pair. Depending on the
analysis type, the value found
to satisfy all ranges for that instrument and analysis type may be the final
output of the program
for that instrument, i.e., the minimum f
-mktdata or minimum fords=
224349600_1
CA 02946578 2016-10-27
PATENT
Attorney Docket No. 45XP-215265
[059] The program described above may be parallelized by implementing it as a
Mapper-only
(i.e., no reduce phase) Hadoop MapReduce Job, whereby each Mapper processes a
single
instrument and single analysis type. Such parallelization may be required to
ensure the program
does not run out of memory, and runs to completion in a satisfactory amount of
time.
[060] Using the Minimum "f" value to inform the Value of a Latency Floor
[061] In an implementation, latency floor application 222 may use the minimum
f value
(determined by summing the minimum fmktdata with the minimum fords for an
instrument) to
inform the value of a latency floor in order to attempt to make the electronic
trading venue 110
more fair.
[062] In the context of an electronic trading venue 110, a latency floor (also
known as
randomization mechanism) can be thought of as a limited exception to the time-
priority rules for
processing messages. At short timescales messages are generally not processed
in the order in
which they are received. At longer timescales messages received earlier are
still processed before
those received later. The "value" of the latency floor is what distinguishes
the short timescale
from the longer timescale.
[063] An appropriately-designed latency floor may be used to remedy temporal
fairness issues
on an electronic trading venue, by removing the advantage a participant gains
from receiving
prices earlier than others and/or by having his orders subjected to longer or
shorter amounts of
time between receipt and processing by the venue than others. If a latency
floor is implemented
by a venue the value of the latency floor can be informed on a per instrument
basis by analyzing
all pairs of a plurality of participants on each instrument in the manner set
forth above to find
two minimum f values: one for market data (f
.---mktdata) and the other for orders The
value of
(fords).the floor can then be set on a per instrument basis by combining the
two minimum f values (e.g.
by summing them).
[064] Advantageously this may result in a value of the floor that is no longer
than it needs to be
to remedy temporal fairness on the venue (noting that negative outcomes maybe
associated with
unnecessarily large latency floor values). Alternatively, for simplicity, the
floor value may be set
on a venue as a whole, or on an instrument group basis, with the value being
set to the largest
sum of minimum f values in market data and orders on the instruments in the
group.
[065] The floor may be set to an even larger value than the sum of the two
minimum f values if
16
224349600_1
81800301
the operator of the venue additionally wants to establish a "minimum response
time" for all
market participants to have a chance of winning a race to make or take a price
(a further
discussion of response times and equal chances of winning is described in U.S.
Patent
application serial number 14/533,543). In this "minimum response time"
approach the value of
the floor may be informed by the sum of the minimum f values on the market
data and orders for
an instrument, plus the minimum desired response time. For simplicity, it may
also be performed
on an instrument group basis by choosing the maximum sum of minimum f values
for the
instruments in that group, and adding to that sum the desired minimum response
time to yield the
floor value.
[066] FIG. 3 depicts a process 300 for quantifying fairness of an electronic
trading venue,
according to an implementation of the invention.
[067] In an operation 302, process 300 may include obtaining market activity
information
associated with a first market participant and a second market participant.
Market activity
information may indicate messages received from market participants relating
to an instrument
traded on the electronic trading venue. Such messages may relate to, without
limitation, orders
(e.g., new order requests, cancel requests, replace requests, etc.).
[068] In an operation 304, process 300 may include determining, based on the
market activity
information, a first plurality of time deltas that each indicate an amount of
time in which the first
market participant has temporally beaten the second market participant with
respect to market
activity related to an instrument traded on the electronic trading venue.
[069] In an operation 306, process 300 may include determining, based on the
market activity
information, a second plurality of time deltas that each indicate an amount of
time in which the
second market participant has temporally beaten the first market participant
with respect to the
market activity related to the instrument traded on the electronic trading
venue.
[070] In an operation 308, process 300 may include determining a quantified
level of temporal
fairness with respect to the first market participant and the second market
participant on the
electronic trading venue based on the first plurality of time deltas and the
second plurality of
time deltas.
[071] FIG. 4 is a schematic diagram that depicts quantifying temporal inter-
component and
intra-component temporal fairness on an electronic trading venue, according to
an
17
CA 2946578 2020-01-20
CA 02946578 2016-10-27
PATENT
Attorney Docket No. 45XP-215265
implementation of the invention. FIG. 4 depicts how the systems and methods
described herein
may quantify the temporal fairness of individual components of the electronic
trading venue.
[072] As illustrated in FIG. 4, a FIX gateway component (also referred to
herein as a "gateway"
and each illustrated in FIG. 4 as a GW1, GW2, and GW3 ¨ although three FIX
gateway
components are illustrated, other numbers of FIX gateway components may be
used) of an
electronic trading venue may receive orders from market participants
(illustrated as Pl-P8,
although other numbers of market participants may provide orders as well).
Market Participants
(Pl-P8) are shown connected to these gateways by way of solid line arrows,
which do not
necessarily indicate unidirectional communication links.
[073] To quantify the intra-gateway temporal fairness of GW1, system 100 may
calculate the
minimum fords value as disclosed herein by selecting for the analysis all
pairs of participants from
the plurality of participants connected to that gateway. Specifically, for
this gateway, these pairs
would be: {P1,P2), {P1,P3} and { P2,P3}. The intra-gateway minimum fords for
GW1 may be
computed as 205 s (or some other number) and this number may be visualized by
placing it
inside the circle representing GW1 in the figure. The same approach can be
taken to quantify and
visualize the temporal fairness associated with GW2 (Otis) and GW3 (17 s) as
shown in the
figure.
[074] To quantify the inter-gateway temporal fairness, system 100 may
calculate the minimum
fords value by selecting all pairs of participants connected to the pair of
gateways of interest, using
the participant from the first gateway as the first element in each pair of
participants, and the
participants from the second gateway as the second element in the pair. For
instance, if the
gateways chosen are GW1 and GW2 then the pairs from which the minimum fords
value should
be computed are: { Pl,P4} ,1P1,P5 ,{ P2,P4},{P2,P5 ), {P3,P4} and 1P3,P5 . If
the minimum f
-ords
value is computed as 23 s then a dotted line maybe drawn between GW1 and GW2
showing this
value, as indicated in FIG. 4. Note that if the dotted line is given a
direction, by way of an arrow
as shown in the diagram, it may indicate which gateway is "dominating" the
other e.g., because
more orders from the former overtake the latter. As depicted in the diagram,
other inter-gateway
fairness values can also be calculated and displayed in a similar manner.
[075] It is to be appreciated that while FIG. 4 shows inter- and intra-
component temporal
fairness values for order-related venue components, that this same approach
can be used for
18
224349600_1
CA 02946578 2016-10-27
PATENT
Attorney Docket No. 45XP-215265
market data-related venue components. By selecting specific participants based
on the market
data component they are connected (either directly or transitively if there
are multiple "layers" of
such components), system 100 may similarly quantify temporal fairness for
market data
distribution components too. The approach specified in the invention where a
minimum value for
fnaktdata is computed from all pairs of those selected participants¨selection
of which is based on
the market data component each is connected to¨would be used.
[076] It is to be further appreciated that the approach and visualization-
style exemplified in
FIG. 4 may allow the operator of the electronic trading venue to rebalance or
reconfigure which
participants are connected to which specific venue components, or for
instance, to re-implement
those venue components using different technology, so as to minimize the
overall minimum fords
and fmktaaia values on that venue.
[077] Although illustrated in FIG. 1 as a single component, computer system
110 and end user
device 140 may each include a plurality of individual components (e.g.,
computer devices) each
programmed with at least some of the functions described herein. In this
manner, some
components of computer system 110 and/or end user device 140 may perform some
functions
while other components may perform other functions, as would be appreciated.
The one or more
processors 212 may each include one or more physical processors that are
programmed by
computer program instructions. The various instructions described herein are
exemplary only.
Other configurations and numbers of instructions may be used, so long as the
processor(s) 212
are programmed to perform the functions described herein.
[078] Furthermore, it should be appreciated that although the various
instructions are illustrated
in the figures as being co-located within a single processing unit, in
implementations in which
processor(s) 212 includes multiple processing units, one or more instructions
may be executed
remotely from the other instructions.
[079] The description of the functionality provided by the different
instructions described
herein is for illustrative purposes, and is not intended to be limiting, as
any of instructions may
provide more or less functionality than is described. For example, one or more
of the
instructions may be eliminated, and some or all of its functionality may be
provided by other
ones of the instructions. As another example, processor(s) 212 may be
programmed by one or
more additional instructions that may perform some or all of the functionality
attributed herein to
19
224349600_ 1
CA 02946578 2016-10-27
PATENT
Attorney Docket No. 45XP-215265
one of the instructions.
[080] The various instructions described herein may be stored in a storage
device 214, which
may comprise random access memory (RAM), read only memory (ROM), and/or other
memory.
The storage device may store the computer program instructions (e.g., the
aforementioned
instructions) to be executed by processor 212 as well as data that may be
manipulated by
processor 212. The storage device may comprise floppy disks, hard disks,
optical disks, tapes, or
other storage media for storing computer-executable instructions and/or data.
[081] Database 122 described herein may be, include, or interface to, for
example, an OracleTM
relational database sold commercially by Oracle Corporation. Other databases,
such as
InformixTM, DB2 (Database 2) or other data storage, including file-based, or
query formats,
platforms, or resources such as OLAP (On Line Analytical Processing), SQL
(Structured Query
Language), a SAN (storage area network), Microsoft AccessTM or others may also
be used,
incorporated, or accessed. The database may comprise one or more such
databases that reside in
one or more physical devices and in one or more physical locations. The
database may store a
plurality of types of data and/or files and associated data or file
descriptions, administrative
information, or any other data.
[082] The various components illustrated in FIG. 1 may be coupled to at least
one other
component via a network, which may include any one or more of, for instance,
the Internet, an
intranet. a PAN (Personal Area Network), a LAN (Local Area Network), a WAN
(Wide Area
Network), a SAN (Storage Area Network), a MAN (Metropolitan Area Network), a
wireless
network, a cellular communications network, a Public Switched Telephone
Network, and/or
other network. In FIG. 1, as well as in other drawing Figures, different
numbers of entities than
those depicted may be used. Furthermore, according to various implementations,
the
components described herein may be implemented in hardware and/or software
that configure
hardware.
[083] The various processing operations and/or data flows depicted in the
figures are described
in greater detail herein. The described operations may be accomplished using
some or all of the
system components described in detail above and, in some implementations,
various operations
may be performed in different sequences and various operations may be omitted.
Additional
operations may be performed along with some or all of the operations shown in
the depicted flow
224349600_1
CA 02946578 2016-10-27
PATENT
Attorney Docket No. 45XP-215265
diagrams. One or more operations may be performed simultaneously. Accordingly,
the
operations as illustrated (and described in greater detail above) are
exemplary by nature and, as
such, should not be viewed as limiting.
[084] Other implementations, uses and advantages of the invention will be
apparent to those
skilled in the art from consideration of the specification and practice of the
invention disclosed
herein. The specification should be considered exemplary only, and the scope
of the invention is
accordingly intended to be limited only by the following claims.
21
224349600_1
