MTPS: A MULTI-TIME-FRAME TECHNICAL PATTERN AND STRATEGY BACKTESTING SYSTEM

MPhil Thesis Defence


Title: "MTPS: A MULTI-TIME-FRAME TECHNICAL PATTERN AND STRATEGY BACKTESTING 
SYSTEM"

By

Mr. Diwen XU


Abstract

Strategy backtesting which is a process to take some user-defined trading rules 
and market information as input, simulate behaviors of market participants 
under these trading rules and finally report the strategy performance. Strategy 
backtesting systems are the most important applications of algorithmic trading 
since it provide individual or professional traders flexibility to design and 
test their customized ideas. Technical analysis is the most popular way for 
financial instruments traders to predict future price movement by researching 
historical data records. However, technical analyses highly rely on a various 
of technical indicators, patterns and personal charting heuristics, which makes 
it very difficult to be transformed into algorithmic trading strategies. Most 
technical traders especially for individual traders, on the other hand, are 
lack of knowledge about computer programming, and find it hard for them to 
script their strategies by coding. Therefore, a strategy backtesting system 
allowing users to describe complex and advanced technical strategies via pure 
graphical user interface (GUI) is in very highly demand.

In this thesis, we design and implement a multi-time-frame technical pattern 
and strategy backtesting system, that is MTPS, to address the problem of 
describing complex technical patterns on GUI and applying these patterns in 
algorithmic trading strategies. By observing the cases that a large amount of 
profitable technical patterns are composed of sequences of characteristic 
candle bars, we introduce a hierarchical MTPS strategy description system that 
promote users to describe their technical patterns by describing sequences of 
technical events. This strategy description system endow users who are not 
knowledgeable about computer programming a much higher flexibility to include 
complicated and advanced patterns in their algorithmic strategies. In addition, 
we also provide users interfaces in MTPS to include technical indicators from 
multiple time-frames in one strategy which is a unique feature that most of the 
state-of-the-art strategy backtesting systems can not achieve. Besides, we 
present present the complete architecture of MTPS including the implementation 
details of all the software components and how they are organized in an 
event-driven design pattern. Moreover, we propose an efficient pattern matching 
algorithm for MTPS to minimize the delay between market data arrival and order 
placement. Last but not the least, experiments are conducted showing that MTPS 
do capable to describe 3 complex and advanced technical pattens and backtesting 
strategies based on these 3 patterns to generate profit surpassing the 
"buy-and-hold" strategies.


Date:			Friday, 27 July 2018

Time:			2:30pm - 4:30pm

Venue:			Room 3494
 			Lifts 25/26

Committee Members:	Prof. James Kwok (Supervisor)
 			Prof. Gary Chan (Chairperson)
 			Dr. Wilfred Ng


**** ALL are Welcome ****