Execution Timing Critical Factor in System Performance
By Lou Mendelsohn
Most technical trading systems today are updated each day after the markets have closed, meaning that trading signals that frequently are triggered by closing prices cannot be executed until the opening on the next trading day.
This approach presumes that execution timing has little or no effect on trading performance and cannot or should not be a controllable system parameter, subject to testing and commodity specific optimization. But so far, assumptions underlying this approach have not been substantiated and confirmed through valid historical testing.
With the proliferation of computerized systems generating automatic signals for execution on the open and with most systems not being able to forecast trading signals prior to the close, traders may be exposing themselves unnecessarily to excessive execution slippage and price gapping as their trades are executed on the open.
Some trading systems report profitability statistics based on historical testing which assumes closing price executions, even though the systems lack the capability of forecasting the signals prior to the close. Therefore, the signals could not have been executed on the close as indicated. That means significant distortions are introduced into performance results, invalidating the profitability claims.
Timing big factor
Technical evidence suggests that execution timing does have a substantial effect on the resulting performance of technical trading systems, independent of the logic and parameters that trigger the trading signals.
To demonstrate this, the Profit Analyst history tester of the ProfitTaker trading system was used to display performance results for four identical tests on T-bills for the trading period from Feb. 1, 1982, to Feb. 25, 1983. The contracts tested were June, September and December 1982 and March 1983, with three rollovers occurring during this period.
The identical “logic block” and system parameter values used for all four tests were selected for display purposes only because they result in few trading signals being generated. This made it possible to display all of the trading signals for the four tests on one trading signals report. The only difference among the four tests was the following four combinations:
- Enter on close/exit on close.
- Enter on open/exit on open.
- Enter on close/exit on open.
- Enter on open/exit on close.
Enter and exit signals can be either to buy or sell. The enter is used to put on a position while the exit designates the closing out of an existing position. For example, with the enter on close/exit on close combination, if the enter signal is a buy on the close, then the exit is a sell on the close.
Only a small portion of the trading history report is included in the printout on page 78 (figure 1). It displays the daily price statistics and system parameter values. The complete report also identifies rollovers, lock-limit days and trend reversals as they occurred. Finally, this report displays the trading signals that were triggered. However, because it does not display the actual executions of those signals, the report is the same for all four tests.
The trading signals report (figure 2) details the actual executions for each of the four tests. This report shows the specific contracts traded, execution dates, whether entry and exit signals were to buy or sell and execution prices. It displays profit or loss on closed trades, number of consecutive losing trades, maximum dollars of consecutive losses and cumulative net equity. For each test, it also shows any open positions outstanding at the end of the trading history, with the unrealized profit or loss displayed.
Rollovers, lock-limit situations and trend reversals also were incorporated into the report. When these occurred, signals would be reported consistently for the corresponding trading day, execution time and price.
The summary report (figure 3) provides a comprehensive summary of key performance measures for each of the four tests. Because the trading system or logic block was the same for all four tests and the system parameters were kept constant except for execution timing, all differences in trading performance can be attributed solely to execution timing.
An analysis of the summary report reveals that the enter on open/exit on close combination of execution timing performed best for T-bills during this period. The enter on close/exit on close performed better than the enter on open/exit on open. The enter on close/exit on open performed worst.
The performance table (figure 4) reveals there were significant differences in performance between the best and worst execution timing combinations. Multi-year rollover testing, using a logic block and system parameters which generate more signals, or testing on a more volatile commodity would cause these differences to be even more pronounced.
Because of the complexity of forecasting trading signals prior to their occurrence and of optimizing execution timing on a commodity specific basis, it would be time consuming and somewhat difficult to incorporate execution timing into the architecture of manually operated systems.
However, in the case of computerized systems, there is no technical reason why the next generation of microcomputer trading system software cannot incorporate execution timing into its architecture and be capable of forecasting trading signals prior to their occurrence. Also, computerized history testers should be able to test and optimize execution timing as a parameter value along with the other mathematical parameters that comprise the system.
With these additional system capabilities, futures traders would be able to customize timing of entry and exit executions on a commodity specific basis, independent of the other system parameters that trigger trading signals. By doing so, substantially greater profitability as well as trading flexibility could be realized.
Louis Mendelsohn president of Market Technologies Corporation, a software development firm in Tampa, Fla., “ProfitAnalyst” and “ProfitTaker” are trademarks of the firm.