Algo Trading Presentation PDF
Algo Trading Presentation PDF
Algo Trading Presentation PDF
1
Why is Algorithmic Trading Exploding in the industry?
2
Average Reported Buy-Side Trading Volume by Geography
2006 2005
3
Adoption of FX Electronic Trading Globally
2006 2005
4
Current and Expected FX e-Trading Activity Globally
Based on responses from 628 instituitions trading electronically in 2004, 679 in 2005, 877 in 2006
Greenwich Associates
5
Algo & e-FX proliferate growth in the search for Alpha
6
How are algo trading revenues split today?
7
Algo is the vehicle, Market Data the Fuel but what is the cost?
Cost Variables:
Market Data STP Protocols
Real Time Data Compliance
Historical Data MiFID
Connectivity APIs
Infrastructure / Network Algo Maintenance
Latency Tools Algo Creation (Quants)
ECN brokerage IT specialists
E-sales Client Analysis
Back Office Marketing
Security Ticket Processing
Prime Brokerage Order System etc…
=
??
What you do know is that the future of
your business is at risk if you don’t
do it….
8
Using Algorithms to Increase Business
9
Using Algorithms to Increase Business
10
Unlocking Artificial Intelligence
Since mid 20th century, researchers have heavily contributed in the
development artificial intelligence.
Much like neural organization, it is not about the complexity of the basic
constituents but more about the complexity of their assembly
Most Algorithmic topics focus on simple inference engines that link
ensembles with decision trees resembling a living intelligent design.
The backbone of Algorithmic is machine learning which is an extension
of statistical science. Most of you have already dealt with this concept
by implementing a simple regression calculation!!!
Today’s Algorithmic trading is involves neural networks, fuzzy logic,
pattern recognition…
Make no mistake, there is no magic or wizardry.. It is always a more
evolved form of regression technique using more/less complex filter
design…
Again, complexity is more in the circumvolution of these algorithms.
11
Algo – Magic or Science ??
12
Algo Trading Strategies
•
¾ SignalProcessing: Mathematical extension of technical analysis based
on the art of filtering to eliminate noise and discern trading patterns
¾ Pattern
Recognition: Enable the machine to learn, adapt and re-act
when patterns arise creating revenue opportunities.
• Note: Single strategies are vulnerable: It is in the mix and diversification that algos
show strengths and consistent productivity
13
Trading Strategies
14
Algo Trading Applications
15
Cycle of Algo trading
1. Idea derivation is predominantly based on market intuition and practice
2. Discover quantitative support to model initial instincts. Surprisingly this
discovery comes from very diverse industries & scientific fields
3. Sketch & Prototype research algorithms
4. Backtest and evaluate robustness and productivity
5. Improve production via cross fertilization of existing algorithmic library
6. Test algorithms in semi-vivo environment to detect corner cases and
validate in/out sample results
7. Design and implement production version
8. Monitor real time signals: Algorithmic baby-sitting
9. Refine models constantly with live observation/adjustments.
10. Recycle technology to other asset classes and diversify frequency and
candidates
16
Dangers related to Algo Trading
No human intervention: need to hunt for corner cases as programs
have no conscience of improper trading activities. Time is spent in
design regarding: security, limits breaches, controls, etc…
IT becomes an unprecedented dependency (servers, networks,
hardware, data feeds, latencies)
Your success relies on a100% error-free code (memory leakage,
infinite loops.etc). The risk of running a bugged algorithm can be
enormous especially if in-vivo at high frequency.
Uptime monitoring requirements have to be high and attending
resources must be highly educated
The technology is highly portable and risks of information property
infringement are high by nature
Market volatility lowers by wide spreading of algorithmic activity and
chances of contagious systemic collapses are increased
No universal recipe and promotion of human risk managers to
activate/deactivate automates (mean reversion, trends)
As in the Aviation industry, even if auto-pilots are capable of landing
airplanes smoother then human pilots, the need for experienced
supervisors is required in case of a major systemic failure (9/11).
17