This is a presentation by Kingsley Jones, CEO of Jevons Global Pty Ltd.
Kingsley has been involved in financial markets for close to twenty-five years. During that time, traditional quantitative analysis rose to prominence, as did the importance of technology in trading. In this talk, Kingsley will reprise some of the key ideas that have developed in trading: 1) the importance of order-book analysis; 2) the marriage of technical analysis and behavioural finance; and 3) the role of high performance computing in processing large data sets. We illustrate the confluence of all three on the speaker's own technical indicator: the cost-basis sentiment index. We show how it lies at the intersection of these three trends and show how to speed it up using some mathematical ideas from the theory of parallel computing.
2. IN BRIEF...
RETHINKING TECHNICAL ANALYSIS ... MY JOURNEY
• Origins estimating the cost-basis of the market
• Order books understanding how prices are made
• Linkages tying current sentiment to market history
• Computation pushing performance with parallelism
Rethinking Technical Analysis: New Foundations
and Faster Computation
3. COST BASIS THEORY (2002)
SentimentValuation
Vs.
“Unrealised profits and losses
of investors in stocks drive
investor sentiment.”
4. COST BASIS FOUNDATIONS
When new trading occurs the
Average Cost Basis moves
toward the new prices
The Average Cost Basis is near the
centre of the Volume at Price
distribution
New Average Cost BasisOld Average Cost Basis
11.20
8.90
0.0%
1.0%
2.0%
3.0%
4.0%
5.0%
6.0%
8.00 10.00 12.00 14.00 16.00 18.00
Price
%Volume
11.20
8.90
0.0%
1.0%
2.0%
3.0%
4.0%
5.0%
6.0%
8.00 10.00 12.00 14.00 16.00 18.00
Price
%Volume
Volume
Traded at
Higher Price
𝑡 𝑡 𝑡 𝑡 𝑡−1
6. SUPPORT & RESISTANCE
Investors are far more likely to
have their cost of entry level
at prices where large volume
traded in the recent past
8
10
12
14
16
18
28-Jan-10 28-Apr-10 28-Jul-10 28-Oct-10 28-Jan-11
Time
Price
11.20
8.90
0.0%
1.0%
2.0%
3.0%
4.0%
5.0%
6.0%
8.00 10.00 12.00 14.00 16.00 18.00
Price
%VolumeKey price levels act as support
when reached from above and
resistance when met from below
Volume-at-Price ChartTraditional Price Chart
7. ORDER-BOOK SLOPE (2002)
Source: Jevons Global (2019) and Credit Suisse (2002).
Order book dynamics as the link to technical patterns.
16. MOORE’S LAW AND POWER WALL
CPU clock speed has now plateaued!
Gordon Moore
Source: Intel (2011).
17. AMDAHL’S LAW AND SERIAL WORK
We are now in the Age of Massive Parallelism!
Source: Wikipedia (2015).
18. PARALLEL PREFIX SUM (SCAN)
Source: “Parallel Prefix Sum (Scan) with CUDA” Chapter 39. GPU Gems NVIDIA (2019).
There are work-efficient parallel cumulative sum algorithms.
20. PARALLEL APIS ARE NOW CORE
MPICH CUDA
OpenCLOpenMP
Source: Cray, Ohio State University, Argonne National Laboratory, Nvidia, OpenMP and OpenCL (2015).
21. ORIGINS…THE BEOWULF IN SCIENCE
Source: Jevons Global (2019).
ASCII-RED Pentium Cluster
1990s Beowulf Cluster
Simulation Applications
22. THE NEOWULF AS CLOUD-NATIVE CODE
Source: Jevons Global (2019).
23. THE NEOWULF AS MAKER PROJECT
Source: Jevons Global (2019).
27. SUMMARY...
KEY TAKEAWAYS
• Physics is driving the move to parallelism
• Conventional signal processing is largely serial
• Simple parallelism is possible via compute farms
• How far can we go in parallel back testing?
Technical Analysis can be thought of as nonlinear
signal processing – which we aim to parallelize.