Explaining the notion behind algorithmic trading by showing electronic trading prevalence. Describe high and low frequency trading and introduce a low frequency trading model based on cross-sectional data of closing prices of four US liquid stocks.
5. WHAT IS ALGORITHMIC TRADING?
COMPUTERIZED
ALGORITHMS
MATHMATICAL
MODELS
FAST
CONNECTIONS
REPLACE FLOOR
SPECIALISTS
LIGHT SPEED
ORDER
EXECUTION
CUSTOMIZED
TRADING
ROBOTS
6.
7.
8.
9. HIGH FREQUENCY TRADING
• “Only the companies that understand technology properly and invest in it regularly
will survive in a new environment with new rules.” –Agustin Rubini
• HFT is measured in milliseconds (1ms = 10−3 sec)
• It is even measured in micro seconds (1microsec = 10−6 sec)
• Adds a new definition to AT:
• Fast connections: Microwave towers, Fiber Optics,…
• “Proximity” to Stock Exchanges’ Data Center’s Shorter cables!
• Sophisticated models to understand other trading algorithms strategy
• Accused of quote stuffing and spoofing (Phantom Offers)
• Algo steps out of market when costly Flash Crash
11. AN HFT SCHEMA
• Presentation by :Brad Katsuyama
(2016)
• HFT enables aster access to price
changes
• HFT capitalizes on the
inefficiencies of other traders,
investors and FIs
• Proximity matters in HFT
• 100,000 AMD shares market order
• 2 milliseconds from First to Last
Stock Exchange
• It takes HFT firms only 476
microseconds
12. • Slow down fastest connections
• Orders arrive simultaneously at
exchanges
• NYSE gets order first & BATS
last
• From 2 millisec to 290 microsec
time variance
• Time calculated using live trading
messages
13.
14. LOW FREQUENCY TRADING
No need for very fast
connections
Less than 1 minute Implemented using:
1.Data : Time Series
Historical Prices
2.Platform/BackTesting:
MetaTrader, Quantopian….
3.Design Theory: Strategy for
Algorithm or Trading robot
4.Programming Language :
MetaTrader uses MQL -
Quantopian uses Python
15. DESIGN THEORY
• 1. Moving Averages
• 2.Machine Learning to ride the market volatility
• 3. Sentiment Analysis
16. SENTDEX DATABASE FOR SENTIMENT
Free data availability
01 Jan 2013 - 30 Oct 2018
Assesses the sentiment of companies by pulling from over 20 sources such as Wall Street Journal,
CNBC, Forbes, Business Insider, and Yahoo Finance
17. MY DESIGN THEORY
• Use Moving Averages: The short-rum MA (21 days) & Long-run MA(252
days)
• Use Sentiment scores from Sentdex database (-3 to 6)
• Spread between 21-MA and 252-MA
• Anchor bias theory: Investor are anchored to the long term MA and rely on
unreliable information to adjust to stock price movements.
• When 21-MA >> 252-MA & Positive sentiment score BUY
• When 21-MA << 252-MA & Negative sentiment score SELL
• I choose four stocks from US Equities market: Nasdaq – Facebook – Apple -
Amazon
18. OBSERVATION OF AN ~ 7-DAY DELAY
•Sentiment scores impact stock’s prices
after approximately after 7 days with
respect to the distance between MA!