STOCK INDEX AND FOREX DAYTRADING SYSTEMS CREATED THROUGH AUTOMATIC DESIGN
Daytrading System: A Challenge
Daytrading System design presents a particular challenge to the Trading
System designer. In Daytrading System design, no positions are held overnight,
however many trades per day may be taken. In high frequency Daytrading,
trades may be accomplished every few seconds and the Trading System attempts
to capture just a tick or two of market movement. These trading opportunities
appear and disappear very quickly so latency, measured in milliseconds,
becomes critical. Backtesting of low latency Daytrading Systems becomes
difficult and questionable due to computational limitations and data windows
available. Professional algorithmic traders are focused more on this low
latency type of trading verses the retail trader who focus more on the high
latency, low frequency type of Daytrading where only perhaps 1 to 10 trades
per day are taken.
The expected value of any Trading System is given by:
EV = PW*AW-PL*AL
Where:
EV = Expected Value
PW = Probability of a win
PL = Probability of a loss
AW = amount won
AL = Amount Lost
If EV is positive, the aggregation of trades over the test interval is positive and the Trading System is profitable. Slippage and Commission are not included in these calculations. Note that PW and PL may float over a wide range resulting in some Trading Systems with a high PW having a negative EV and vice versa. The specific values of the resultant Trading System during the in sample training or development period is of secondary interest. Many mathematical constructs can curve fit data and produce an excellent in sample backtest. The ability of the studied Trading System to maintain its trading statistics “Out Of Sample” is material. For Daytrading Systems that are created using an automatic design mechanism, the out of sample testing becomes critical.
Trading System Lab
Trading System Lab tests out of sample “during” the Genetic Programming design process, thus the researcher or designer can view how the Trading System is progressing both in sample and out of sample while the Trading System is being auto designed by the algorithm. Good out of sample results throughout the design or training interval may be indicative that the Trading System is progressing down a path which will result in robust further out of sample performance.
Daytrading Systems should have several sub systems within them. RMESA and
Big Blue Daytrading Systems both use multi pattern sub systems. For the
most part two major types of sub systems are employed. The first pattern
is a breakout pattern capturing Daytrading movements that trend from the
open to the close. The second pattern is the countertrend pattern, capturing
market movements that result in a strong reversal pattern during the day.
Additional patterns may include gap patterns, counter-countertrend patterns
and other micro movement intraday patterns. RMESA and Big Blue use some
variant of support and resistance or pivot point measures of intraday prices.
These levels are widely watched and as such, may be a self-reinforcing trading
point. Indeed, Daytrading systems that use these levels have produced some
excellent back and forward tests. Additionally, news or reports that occur
during the day may be integrated using real time news feeds and the Daytrading
Systems may be modulated based on the nature of the intraday news event.
Overall profitability of Daytrading Systems are largely impacted by factors
including Slippage and Commission, data and data feed issues, consistency
of trading, etc.
Daytrading System design is unique, however many of the same approaches
used in the design of overnight trading systems may be applied to the design
of Daytrading Systems. Stress testing the Trading System on out of sample
data remains one of the most important items that should be accomplished
and the Trading System Lab platform stress tests the Daytrading System on
out of sample data during the automatic design process. With design, testing
and coding being mainly accomplished automatically, the Daytrading System
designer can focus on important “big picture” elements of his
design. -Mike Barna, President, TSL