Slide 1

Slide 1 text

Prognoz Market Surveillance

Slide 2

Slide 2 text

PROGNOZ TODAY 2 1 5 A wide range of standard products for public, corporate, and financial sectors Offices in 7 countries, including USA, Canada, Belgium, China, CIS Own training center with strong methodological support of key projects Support of partner professional community around the world More than 20 years experience in the IT and business analytics market More than 1,500 successful implementations for 550 customers in 70 countries worldwide Leading company in international ratings related to business analytics and custom software development In-house unique software – the Prognoz Platform 6 2 7 3 8 4 years’ experience in the IT market successful implementations highly qualified programmers, analysts, and economists customers around the world countries where our offices are located countries we delivered projects to 70+ 550+ 20+ 7 1500+ 1500+

Slide 3

Slide 3 text

GARTNER MAGIC QUADRANTS 3 Business Analytics Advanced Analytics Gartner included Prognoz in the 2015 Magic Quadrant for Business Intelligence and Analytical Platforms and 2015 Magic Quadrant for Advanced Analytics Platforms

Slide 4

Slide 4 text

KEY CUSTOMERS IN FINANCIAL SECTOR

Slide 5

Slide 5 text

PROGNOZ MARKET SURVEILLANCE (TIMELINE) Functionality Features Product Specification  Market abuse patterns recognition (insider trading, pre- arranged wash trades, matched orders, non-competitive trading, market price manipulation, price control etc.)  HFT abusive strategies detection (front-running, quote staffing, quote smoking, layering/ spoofing, price fade, momentum ignition)  Statistical detection of deviations  High-frequency data visualization engine Tradebook and orderbook replay  Market abuse detection: Insider trading and wash sales  Market price manipulation Trading data interactive visualization  Case management  Regulatory compliance Front-end: 2-tier application Data sources: Stock exchanges data (clients trades and orders, tradelogs, orderlogs) References: This solution is in use in Central Bank of Russia

Slide 6

Slide 6 text

BUILD-IN DETECTION MODELS 7 1. Market abuse patterns recognition: ― insider trading ― pre-arranged wash trades ― matched orders ― non-competitive trading ― market price manipulation ― price control 2. HFT abusive strategies detection ― quote staffing ― quote smoking ― layering / spoofing ― price fade 3. Statistical detection of price deviations

Slide 7

Slide 7 text

END-OF-DAY TO INTRADAY DRILLDOWN 9 Drilldown into intraday data: ― sorting and filtering data ― news and event labels ― sorting and filtering event labels ― zooming function ― intraday market activities monitoring ― cross-transactions in group ― key statistics by trader or group 03.11.2011 Intraday 03.11.2011

Slide 8

Slide 8 text

INTRADAY DATA VISUALIZATION 10 Convenient instruments for intraday data analysis: Intraday deals Traders Net position of selected trader Deals

Slide 9

Slide 9 text

HIGHLIGHTING DEALS 11 Buy deals (green points) Sell deals (red points) Visualization of intraday dynamics: ― Labels of deals and events on the timeline ― Net position for trader or group ― Drilldown to orders and counterparty for each deal Cross deals (orange points)

Slide 10

Slide 10 text

ORDER BOOK VISUALIZATION AND REPLAY 12 Visualization of order book: ― Bid-ask spread and orders of selected traders over historical period ― Order book visualization at the selected moment ― Order book replay: tick-by-tick or second-by-second ― Drilldown to list of orders for each price level Historical period Order book Order list Traders

Slide 11

Slide 11 text

EXAMPLE: BID-ASK SPREAD & ORDER BOOK, 60 SEC 13 Orders of selected trader Historical period 60 sec. Current time = 16:48:45.000084 Volume by price levels of selected trades Play mode navigator Bid-ask spread

Slide 12

Slide 12 text

EXAMPLE: BID-ASK SPREAD & ORDER BOOK, 1 SEC 14 16 ms Historical period 1 sec

Slide 13

Slide 13 text

PROGNOZ.SITUATION CENTER 15 1. On-line markets and news monitoring 2. High level market health indicators 3. Interactive drilling down into the detailed trading information 4. Early warnings

Slide 14

Slide 14 text

ALGORITHMS CONFIGURATOR 16 Algorithms configurator 1. High level objective language available for users 2. Binary compliable code (not interpreter)

Slide 15

Slide 15 text

DMZ Enterprise Network Project Project HOW IT WORKS 17 Project Historical Database Oracle API Cache Cache Prognoz.Situation Center Prognoz.TimeLine Project Project Distributed Calculation Engine t Alerts & Statistics Algorithm configurator Architecture benefits for brokers and regulators 1. Sophisticated market abuse patterns recognition 2. Configurable algorithms by users 3. Having isolated DMZ is the stringent info security requirement of many financial institutions 4. Insignificant (for surveillance) latency dramatically decreases costs of solution Batch Files Reserved channel

Slide 16

Slide 16 text

0.1 0.2 0.4 0.8 1.6 3.2 6.4 12.8 25.6 Count Millions Time Number of trades Number of orders Number of orders: ~ 10 M per day (median) ~ 37 M per day (in peak) Number of trades: ~ 700 K per day (median) ACTIVITY OF EQUITY MARKET

Slide 17

Slide 17 text

*CFTC Technology Advisory Committee, 2012 HIGH FREQUENCY TRADING (HFT)

Slide 18

Slide 18 text

“Coscia was accused of entering large orders into futures markets in 2011 that he never intended to execute. His goal, prosecutors said, was to lure other traders to markets by creating an illusion of demand so that he could make money on smaller trades, a practice known as spoofing. Prosecutors said he illegally earned $1.4m (£900,000) in less than three months in 2011 through spoofing.”, The Guardian, HFT layering, November 2015 “A unit of hedge fund Citadel LLC was fined $800,000 by U.S. regulators in June for failing to prevent erroneous orders from being sent to several stock exchanges over a nearly three-year period”, Reuters, HFT stuffing, 2014 “Athena is the regulator’s first market manipulation case against a firm engaged in high-frequency trading, an industry besieged by accusations that it cheats slower investors” , Bloomberg Business, HFT manipulations, 2014 “Navinder Singh Sarao, 36, is fighting extradition to the US where he is facing 22 charges ranging from wire fraud to commodities manipulation, which carry sentences totalling a maximum of 380 years. Mr Sarao is alleged by US prosecutors to have made $40m over four years by spoofing the Chicago futures market. The trader’s activities include making a $900,000 profit on May 6, 2010, when a trading frenzy known as the flash crash saw one of the most spectacular falls ever seen in the equity markets.” , Financial Times, 22 October 2015 MANIPULATIONS

Slide 19

Slide 19 text

TIMELINE: BUILD-IN DETECTION MODELS 23 1. Market abuse patterns recognition: ― insider trading ― pre-arranged wash trades ― matched orders ― non-competitive trading ― market price manipulation ― price control 2. HFT abusive strategies detection ― quote staffing ― quote smoking ― layering / spoofing ― price fade 3. Statistical detection of price deviations

Slide 20

Slide 20 text

Quote smoking - practice of putting a large number of quotes (creation new bids and offers) and then immediately canceling them Best bid Best ask 69 ms QUOTE SMOKING Agent’s asks

Slide 21

Slide 21 text

Possible criteria: • Number of orders per time interval (sec, min, hour) • Median lifetime of order • Best price ratio = count “best price” orders / count agent’s orders • Median minimum distance between agent’s price orders and best prices (best ask/best bid) QUOTE SMOKING

Slide 22

Slide 22 text

Quote stuffing - practice of putting a large number of orders (thousands of messages) and then immediately canceling them for creation delay in other participants. QUOTE STUFFING

Slide 23

Slide 23 text

Possible criteria: • Number of orders per time interval (sec, min, hour) • Order-to-trade ratio • Median lifetime of orders • Range of order’s price QUOTE STUFFING

Slide 24

Slide 24 text

Layering - practice of creation selling/buying pressure in order to make naive investor to move the price. Best ask Best bid Cancellations of orders Agent’s bid Agent’s asks Trade LAYERING/SPOOFING

Slide 25

Slide 25 text

Possible criteria: • High level of order imbalance = count of buy orders / all orders • High number of orders in visible part of order book • Ratio of agent’s volume to visible volume of order book • Ratio of agent’s orders to count of orders in visible part of order book • Median lifetime of orders in each part of order book LAYERING/SPOOFING

Slide 26

Slide 26 text

Price Fade - practice of orders cancellation immediately after the trade on the same venue. Best ask Best bid Cancellations of orders Agent’s asks Trade PRICE FADE

Slide 27

Slide 27 text

Possible criteria: • Number of cancelled orders at the same time • Range of order’s price • Number of orders cancelled before trade (in interval x sec) PRICE FADE

Slide 28

Slide 28 text

TIMELINE: BUILD-IN DETECTION MODELS 33 1. Market abuse patterns recognition: ― insider trading ― pre-arranged wash trades ― matched orders ― non-competitive trading ― market price manipulation ― price control 2. HFT abusive strategies detection ― quote staffing ― quote smoking ― layering / spoofing ― price fade 3. Statistical detection of price deviations

Slide 29

Slide 29 text

CIS, Eastern Europe Moscow +7 495 995 80 76 Western Europe Brussels +32 2 217 19 50 Asia Beijing +86 10 6566 5337 North and South America, Canada, Australia, Africa Washington +1 202 955 55 20 34 CONTACTS