Dates and Locations

TBC

London, UK

London Chamber of Commerce >>

Please visit our Course Diary to see dates and locations for all of our UK and overseas courses.

Course Outline: Backtesting, Optimization & Strategy Development

Developing a trading strategy is a journey through a minefield. What often appears to be good is not, and what seems to be a poor choice turns out to be a success. Much of this can be resolved by understanding the fundamental concepts of strategy selection and design, backtesting and optimization, and risk control. It is a case of theory versus reality. Once you are comfortable with the concepts, choices are clear and the results come easily.

This two day workshop is designed for traders and investment managers of all asset classes and time scales who are looking to test and optimize their own trading strategy or model. Covering in-depth backtesting strategies and an introduction to optimization, delegates will use Microsoft Excel and backtesting software (PC supplied) to solve back-testing and optimization problems using real market data.

Session 1 - Strategy Choices
Theory vs reality:
The Black Swan coin flipping example; Full disclosure by financial institutions; off-balance sheet accounting; The case for full automation
Sound premise
Validation versus discovery
Time horizon and strategies
Strategy trade-offs:
Trend; Mean reversion; Arbitrage
Market noise -- friend or foe?:
Measuring market noise; Ranking of markets, emerging markets by noise; Noise defines profits
It’s not the method, it’s the market:
A trend example; There is no “best method”
Add-ons:
profit-taking and stop losses (examples)
Complexity versus transparency:
The case for loose pants
Price shocks (example):
Affect longer holding periods; Best defensive strategies
Misnomers:
Low vol does not mean low risk; Small losses does not mean small drawdown
Expectations:
Expecting the future to be like the past; Good portfolios can be build from average results; Engineering the risk out of a strategy

Session 2 - Data and Markets
Continuous data:
But not if using percentage calculations; Allow for cost of rolls if long holding period; Continuous 3-month contracts
Using cash FX for testing
Convert results in $PL to NAVs at a target volatility for consistency

Session 3 - Traditional Testing (Examples: macrotrend and mean reversion)
Length of test
Selecting the test period
Use the same start date, equal to the longest wind-up
Impact of costs that are too high or too low:
Costs affect the short end; Eliminating a good strategy
In-sample and out-of-sample rules:
Assigning the test periods; Checkerboard method; Random periods for multiple strategies
Optimizing by market, by sector, by portfolio
Multiple parameters and rules
The power of optimization:
Interval search; Monte Carlo optimization; Genetic algorithm
Expressing the results; returns, ratio, drawdown, kurtosis, semivariance
Seeking robustness -- evaluating the results, % good results, kurtosis
Sensitivity testing
Price shock bias

Session 4 - Going Out-of-Sample
Expectations: In-sample, out-of-sample, actual trading
Choosing the best parameters
The final test with all data
Results are too good
Results are not good enough
Longevity and retesting

Session 5 - Risk Control
Price shocks, before and after:
Be out of the market as much as possible; Post price shock strategy
It’s not the market, it’s the money
You can’t eliminate the risk, you can only move it around
Be prepared to earn less and risk more
Manual allocations versus portfolio optimization
Rebalancing
Is it still working?
t-test; Common sense