EXANTE Algorithmic Trading: Practical Aspects

39
Algorithmic trading: practical aspects EXANTE Ltd. exante.com.mt [email protected] Moscow 2011

description

Slides for speech of EXANTE Managing Partners Vladimir Maslyakov and Anatoliy Knyaze , entitled "Practical aspects of algorithmic trading and high-frequency trading", on TradeTech Russia 2011 Presentation highlights the problems associated with the development of a model (pre-trade analysis), the launch of the strategy (trading) and the post-trade analysis, as well as an overview of the algorithmic trading in general, and a small glimpse into the future.

Transcript of EXANTE Algorithmic Trading: Practical Aspects

Page 1: EXANTE Algorithmic Trading: Practical Aspects

Algorithmic trading:

practical aspects

EXANTE Ltd.

exante.com.mt

[email protected]

Moscow 2011

Page 2: EXANTE Algorithmic Trading: Practical Aspects

I. Algorithmic trading

II. Develop the model

III. Launch the strategy

IV. Analyze the results

V. Trends

Page 3: EXANTE Algorithmic Trading: Practical Aspects

Algorithmic trading

Automated trading

HFT

Page 4: EXANTE Algorithmic Trading: Practical Aspects

Automated trading

Buy-side Sell-side

Stat arbitrage

VWAP Market Making / HFT

Trend

following

Arbitrage

Smart order

routing

Pricing

Page 5: EXANTE Algorithmic Trading: Practical Aspects
Page 6: EXANTE Algorithmic Trading: Practical Aspects

I. Algorithmic trading

II. Develop the model

III. Launch the strategy

IV. Analyze the results

V. Trends

Page 7: EXANTE Algorithmic Trading: Practical Aspects

Data Hypothesis Model Testing

Page 8: EXANTE Algorithmic Trading: Practical Aspects

Historical Data

Width

Instruments

Venues

Corp. actions

News

Depth

Past period

Resolution

Order book

Counterparties

Correctness

Splits and divs

Gaps

Timestamps

Validation

Page 9: EXANTE Algorithmic Trading: Practical Aspects

Data Hypothesis Model Testing

Page 10: EXANTE Algorithmic Trading: Practical Aspects

A priori knowledge

Gut feeling

Empirical Fundamental

Page 11: EXANTE Algorithmic Trading: Practical Aspects

Visualization

Data volume

Speed

Math

Иллюстрация с panopticon.com

Page 12: EXANTE Algorithmic Trading: Practical Aspects

RTS Index and S&P Index, 2010-10-11

16:40 16:50 17:00 17:10 17:20 17:30 17:40

RTSI SPX

Page 13: EXANTE Algorithmic Trading: Practical Aspects

Data Hypothesis Model Testing

Page 14: EXANTE Algorithmic Trading: Practical Aspects

Model

Alpha Risks Transaction costs

Page 15: EXANTE Algorithmic Trading: Practical Aspects

EDGE ?

Page 16: EXANTE Algorithmic Trading: Practical Aspects

Model: math

Prototype

Page 17: EXANTE Algorithmic Trading: Practical Aspects

Our experience: R

Domain Libraries Open and free

Slow No realtime Open and free

Page 18: EXANTE Algorithmic Trading: Practical Aspects

Data Hypothesis Model Testing

Page 19: EXANTE Algorithmic Trading: Practical Aspects

Testing

Data

• Historical data

• Modeling market impact and order flow

• Realtime

Prototype

• R / Python/ Java

• Cluster / cloud

• GPU

Results

• Alpha

• Risks

• Transaction costs

Page 20: EXANTE Algorithmic Trading: Practical Aspects

I. Algorithmic trading

II. Develop the model

III. Launch the strategy

IV. Analyze the results

V. Trends

Page 21: EXANTE Algorithmic Trading: Practical Aspects

Инфраструктура

Page 22: EXANTE Algorithmic Trading: Practical Aspects

Speed

Low-latency

Ultra low-latency

Sub millisecond

Depth

L1

L2

Raw

Coverage

Americas

Europe

Asia

Realtime data

Page 23: EXANTE Algorithmic Trading: Practical Aspects

Strategy

NYSE MFG

LSE

Data Orders

JP

Strategy sandbox

Robot 4 Robot 3

Robot 2 Robot 1

Language

High-level (C++, Java, C#, etc)

DSL (Slang, etc)

Visual (diagrams)

Infrastructure

Client

Server

Cloud

Control

Manual

Automatic

GUI

Page 24: EXANTE Algorithmic Trading: Practical Aspects

Arbitrage example

GAZPRU (MICEX)

On new tick: ogzd_rub = convert(ogzd, usd_rub) spread = normalize(ogzd_rub/gazpru) changedSpread()

OGZD (LSE)

USD/RUB (FOREX)

LIMIT (LSE)

London Server (Telehouse)

Filled (size)

MARKET (MICEX)

Filled (price)

On change spread: if (spread > threshold) place_limit(OGZD, price, size)

On limit fill: If (limit_is_filled) place_market(GAZPRU, size)

Arbitrage strategy

Parameters: threshold

Page 25: EXANTE Algorithmic Trading: Practical Aspects

VWAP example

On new tick: vwap = recalculateVwap(trades) execute_vol = recalculate(average_volume, volume) executeOrder(execute_vol)

MARKET (MICEX)

Moscow Server (MacomNet)

Filled (size)

on market fill: our_vwap = update(price, size) vwap_delta = our_vwap - vwap

VWAP strategy

SBER bid/ask (MICEX)

SBER volume (MICEX)

SBER trades (MICEX)

Parameters: average_volume

Page 26: EXANTE Algorithmic Trading: Practical Aspects

I. Algorithmic trading

II. Develop the model

III. Launch the strategy

IV. Analyze the results

V. Trends

Page 27: EXANTE Algorithmic Trading: Practical Aspects

Gather results data

Market snapshot Orders

Latency Strategy parameters

Data

Page 28: EXANTE Algorithmic Trading: Practical Aspects

Export the results data

Excel R, Matlab

Visualization Model

Export

Page 29: EXANTE Algorithmic Trading: Practical Aspects

Compare with the model

Page 30: EXANTE Algorithmic Trading: Practical Aspects

Optimize the parameters

Model

Testing

Trading

Results

Page 31: EXANTE Algorithmic Trading: Practical Aspects

I. Algorithmic trading

II. Develop the model

III. Launch the strategy

IV. Analyze the results

V. Trends

Page 32: EXANTE Algorithmic Trading: Practical Aspects

Adoption CME GLOBEX Vol, % Msgs, %

E-mini S&P 500 Futures 51.66 69.93

EuroFX Futures 69.32 83.41

Eurodollar Futures 51.29 64.46

Crude Oil Futures 35.34 71.24

Foreign Exchange Buy-side, % Sell-side, %

Order Routing 25 92

Time-slice 25 15

Liquidity 42 46

Alpha 92 39

FX Hedging 25 39

Streambase 2011 Special Report on FX

Algorithmic Trading and Market Dynamics July 15, 2010

60

90

0

10

20

30

40

50

60

70

80

90

100 FORTS

Vol, % Msg, %

Estimated by FORTS 09.2011

Page 33: EXANTE Algorithmic Trading: Practical Aspects

Dodd-Frank

SEC 15c3-5

MiFID II

Swap Execution Facility.

Eliminate naked access to exchange.

Crossing networks, derivatives, HFT.

Page 34: EXANTE Algorithmic Trading: Practical Aspects

Algotrader: a new breed

Technology

Mathematics

Finance

Page 35: EXANTE Algorithmic Trading: Practical Aspects

Strategies

ULL DMA

5μs / km

< 100μs / algo

106 msg / sec

HFT

Buy-side or sell-side?

Market making

Liquidity search and aggregation

Stat arb

Multi-asset trading

FX, Eqty, Debt, Derivs

Europe, USA, Asia

Feeds and execution

Fast and reliable data

Page 36: EXANTE Algorithmic Trading: Practical Aspects

Technologies

Software Overclocking FPGA

Multi-core GPGPU Cloud x32 x200 x30000

Page 37: EXANTE Algorithmic Trading: Practical Aspects

Our experience: cloud

Power Diversity Cost control

Engineering Compliance Latency

Page 38: EXANTE Algorithmic Trading: Practical Aspects

Service!

Page 39: EXANTE Algorithmic Trading: Practical Aspects

Anatoliy Knyazev

[email protected]

Vladimir Maslyakov

[email protected]