Tester mode

New comment
 
When using default settings modelling, every tick gives amazing results, but that doesn't exist in real market. Using for test the every tick based real tick give real sight about results. Most EA sold or free works on default set, but in real, it's wasting time and money.
 
While inputs of the EA are one thing, the main thing to blame is the strategy. Some strategies don't hold up. Some systems are not refined enough - victim to noise, victim to fake breakouts and stop hunting, bad entry (too early or too late), and not efficient at closing trades at statistically significant prices. Step away from code, and really blueprint the strategy.
 

You’re absolutely right. That’s why all of my backtests are performed using “Every Tick Based on Real Ticks”, not the default “Every Tick” model.

The default model can produce overly optimistic results because it simulates ticks instead of using actual historical market data. In contrast, Real Ticks include the real bid/ask prices, spreads, and tick-by-tick market movements, making the results much closer to live trading conditions.

I believe an EA should be evaluated under the most realistic conditions possible. While no backtest can guarantee future performance, testing on Every Tick Based on Real Ticks provides a much more reliable benchmark than simplified modeling.

That’s the standard I use when developing and publishing my EA.


 

A rare exception to real ticks as the only universal testing method is the case of very unique custom charts and very unique EA's. For example, if you have a Renko brick chart where all of the bricks are the same size, and an EA coded to analyze the Renko brick open and close prices but trade the underlying tradable symbol; then testing on O[HL]C prices becomes fairly universal... even multiple broker-dealer feeds. You will however, have to calculate your broker-dealer's average spread and enter it into the Tester. To be clear, this the nature of testing any EA on OHLC data.

As an exception to that exception, if you introduce brick wicks/shadows into the strategy, the noise of standard timeframe bars is imported into the otherwise smooth bricks. In this case, outlying ticks determine the high and low prices so your Renko chart must include ticks, which brings us full circle right back to testing on real ticks.

 
It actually depends on what type of EA you are using/creating/buying. Not all EAs need tick data. For example, I hate waiting 30 minutes for a backtest to run, so I make EAs that work perfectly even with 1 minute OHLC modelling. This way, calculating the right parameters is much more efficent. All my published EAs are like this. I'm working on a few tick per tick EAs, but none of them has better performances than the ones I create with my efficent method. Especially on years of data.
 
Daniele Fughelli #:
It actually depends on what type of EA you are using/creating/buying. Not all EAs need tick data. For example, I hate waiting 30 minutes for a backtest to run, so I make EAs that work perfectly even with 1 minute OHLC modelling. This way, calculating the right parameters is much more efficent. All my published EAs are like this. I'm working on a few tick per tick EAs, but none of them has better performances than the ones I create with my efficent method. Especially on years of data.
And... That higher efficiency is exponentially amplified during optimization. It can make the difference between needing a network/agent farm versus a standalone pc.
 

Swing strategies like the ones I use can be tested on anything: ticks, real ticks, M1 OHLC... They have rare entries and huge profits, and 1k points (due to spread or M1 delays) more or less by the end of the day isn't a big deal.

If you want to test scalping or, especially, high-frequency trading, you need to account for far more factors than just using real ticks in your backtest.

 
Alexey Dudin #:

Swing strategies like the ones I use can be tested on anything: ticks, real ticks, M1 OHLC... They have rare entries and huge profits...

Be careful. Moderators may not like when a Seller mentions "huge profits" due to the rule against self-promotion.
New comment