It is better to select all ticks for backtesting

The abbreviation of the currency pair and the name in brackets are the formal names.

The first six characters of the abbreviation are the same across all pairs, but some brokers may append characters after that.

In the case of OANDA, it is -a01. If the EA creator sets the currency pair name to Symbol(), NULL, or "" universally, there is no problem, but when specifying a currency pair, this can cause the EA not to run.

 

Period

The timeframe and period of the settings.

The period is the actual period, and the text in parentheses is the configured period. If there are no data on holidays, the start or end date, the configured period and the actual period may diverge.

The length of the period is one of the indicators of reliability. However, excessively long periods are also problematic.

Data older than five years is less realistic.

More than 3,000 trades with few trades performed are considered reliable for a year-long backtest.

※ In simulation, when trading with a width of 50 pips, if you perform more than 3,000 trades, it settles to a constant width. In incomplete periods like 11 months or 13 months, there is a possibility of champion data.

 

Model

This is the tick model for backtesting.

 

All Ticks

Generate pseudo-ticks and verify.

From historical data, about 30–60 ticks are generated. Backtesting takes longer, but accuracy increases.

 

Control Points

Validation uses the next lower timeframe. For a 1-minute chart, about 10 ticks are generated. It is lighter than all ticks, but its accuracy is about eighty percent of all-ticks.

 

Open Price

Validation is performed only on four-carets (4-candle) charts. Very fast. The accuracy is about sixty percent of all ticks.

Recommended for hourly charts or for EAs that capture the timing of timeframe changes.

Choose backtests that are produced with all ticks.

For EAs without a stop loss where the profit factor (PF) exceeds 1.9, all-tick data is more appropriate.

Because the more granular the ticks, the higher the reproducibility of breaches.

Even if you backtest for a long time with all ticks, it may not perform well in reality.

Do past data not apply to the future? Are pseudo-ticks unrealistic?

I do not think so.

There are questions about pseudo-ticks during sharp market moves, but they yield results that are very close to real operation.

One of the biggest problems is time lag.

Backtests compute inside the computer, so there is no time lag, but in real operation, there is a lag of about 70 ms to several hundred ms for round-trip execution.

Also, some brokers can cause slippage to occur reliably when slippage is set.

Set slippage to zero and consider the broker's real spread plus about 4 as close to actual operation.

If you have an EA sample, or when you can backtest, try with a slightly higher spread to confirm it can consistently profit.

 

Parameters

Parameters are the items labeled as expert settings in a backtest.

There are two main sides to the settings.

• Things the user wants to configure
• Things that are unnecessary

Being able to configure settings such as the MAGIC number, Lot size, and Comments adds convenience for the user.

However, it is safer to avoid those that include conditions like Period and StopLoss and TakeProfit.

So you won’t end up with an EA that you can’t optimally configure after obtaining it.Choose an EA with minimal unnecessary settings.

Next time: about the profit factor and its pros and cons