I am presently carefully monitoring my skilled advisor’s efficiency on gold. One of many monitoring steps is evaluating trades on the sign account with the outcomes obtained within the technique tester.
I attempt to write Knowledgeable Advisors that solely act when a brand new bar opens. That is as a result of lack of high-quality tick knowledge and restricted machine assets for testing. So, the ends in the technique tester and on actual knowledge ought to match. However sadly, this isn’t the case.
Take a look at outcomes for the primary 10 days of February
Over the identical 10 days, the skilled earned 70% on an actual account . It is a checklist of transactions.
General, the outcomes are similar. In each circumstances, the skilled advisor demonstrated large income with minimal drawdowns.
However there are variations! The actual account has extra trades and, on common, decrease income. Let’s examine trades on the chart from the tester and the true account.
Tester
Actual rating
Causes for this habits
1) The distinction within the variety of trades happens as a result of, within the “at opening costs” testing mode, the skilled advisor is late studying a couple of commerce closed by stop-loss or take-profit. Consequently, it does not open a brand new commerce utilizing the identical system. On an actual account, the commerce is closed throughout the bar, and when a brand new bar opens, the skilled advisor can open a brand new place.
2) Revenue on an actual account is decrease as a consequence of slippage, opening delays and different requotes.
Outcomes and conclusions
Having performed a comparative evaluation of the advisor’s work within the “testing at opening costs” mode (Open Costs) and on an actual account, we will draw the next key conclusions.
1. The order closing part is the testing bottleneck
The principle purpose for the discrepancy within the variety of transactions is the completely different logic of occasion processing. closing a place .
-
Within the tester: The sign concerning the activation of Cease Loss or Take Revenue is distributed to the advisor Strictly on the opening of the following bar . Due to this, the chance to open a brand new commerce on the identical ranges is missed—one time interval is “dropped” from buying and selling.
-
In actual life: The deal is closed throughout the bar. The system instantly releases the margin and locks within the outcome. By the point new bar The advisor is not burdened by the earlier place and has the precise to open a brand new transaction instantly after receiving a tick/bar.
Conclusion: The “Open Costs” testing mode underestimates the variety of accomplished trades by artificially creating “useless zones” after protecting orders are triggered.
2. The character of decrease returns on actual cash
The distinction in last revenue is due solely to the market microstructure, which can’t be simulated in a tester with out tick knowledge:
-
Slippage: Execution at a worth worse than the asking worth, particularly in periods of excessive volatility.
-
Execution delays: The worth at which the system sees a sign to open and the worth at which the order is definitely stuffed are completely different values.
-
Re-quotas: Lack of time and potential revenue when asking for a worth once more.
Conclusion: The tester idealizes the entry, assuming the order will at all times be opened on the present market request worth. An actual account pays a liquidity tax.
3. Elementary limitation of the methodology
Regardless of the mathematical discrepancy (variety of transactions + share of revenue), the system habits sample is preserved :
Because of this the advisor’s code is appropriate and the decision-making logic works accurately. Deviations are quantitative , not qualitative character.
