Backtesting with tick data vs. live trading results

Johannes Katzer
916
Johannes Katzer  

Hello,

is there any plausible explanation why live trading results (and backtests using MQL data and generated ticks) can be much better than any simulation with real tick data (with commissions, variable spread, stop level = 0, different data sources, etc.)?

Since the tick data from TDS can be modified in almost any way, like fixed or lower spread or the like, I wonder why MQL data backtests can still be that different? Can anybody help me out?

Thanks

JK

Lorentzos Roussos
24310
Lorentzos Roussos  
Johannes Katzer:

Hello,

is there any plausible explanation why live trading results (and backtests using MQL data and generated ticks) can be much better than any simulation with real tick data (with commissions, variable spread, stop level = 0, different data sources, etc.)?

Since the tick data from TDS can be modified in almost any way, like fixed or lower spread or the like, I wonder why MQL data backtests can still be that different? Can anybody help me out?

Thanks

JK

if you are talking about MT4 (since you mentioned TDS) ,the generation process -if i recall correctly- goes like this : 

if it was a bullish bar : generate ticks between open-low-high-close 

if it was a bearish bar : generate ticks betwee open-high-low-close

Thats why you see J backtests some times ,in Mt4

Johannes Katzer
916
Johannes Katzer  
But how do you explain the difference to live trading then? Is there any plausible explanation why live trading could be much better than a real tick backtest?
How to quickly develop and debug a trading strategy in MetaTrader 5
How to quickly develop and debug a trading strategy in MetaTrader 5
  • www.mql5.com
Scalping automatic systems are rightfully regarded the pinnacle of algorithmic trading, but at the same time their code is the most difficult to write. In this article we will show how to build strategies based on analysis of incoming ticks using the built-in debugging tools and visual testing. Developing rules for entry and exit often require...
Lorentzos Roussos
24310
Lorentzos Roussos  
Johannes Katzer:
But how do you explain the difference to live trading then? Is there any plausible explanation why live trading could be much better than a real tick backtest?

From MetaTrader 5's real tick test ,or from TDS real tick test ? 

Enrique Dangeroux
735
Enrique Dangeroux  

First you have to know the tester quirks very well and adjust for it if necessary depending on what you try to do. 

If you want matching results always use "every tick based on real ticks" mode, any other mode will be further away from reality to nothing related to reality. Use the same datafeed you are going to trade on. And you have to know exactly what datafeed your broker has from market watch as this can be different from the feed for history. This can happen with brokers with different account types. Also MT5 and MT4 feeds from same broker can be different. 

The higher the resolution your strategy, the more important above will be. So for tick strategies, it is absolutely critical. If your strategy runs on higher timeframes, it becomes less critical but does have an impact.

Never assume anything, always test and verify everything.

Lorentzos Roussos
24310
Lorentzos Roussos  

I assume an :

  • actually lower than test slippage
  • actually lower than test latency 
  • actually lower than test spreads 

could do that

Johannes Katzer
916
Johannes Katzer  
You don´t get my point yet. How can live trading results be much better than backtests running with real ticks and optionable commissions, real spread, etc.? How can this be? I am not talking about a specific tick data source. Meanwhile I tested all available sources from TDS. If anyone of you also has TDS, please contact me in PM to discuss this. Thanks a lot.
Lorentzos Roussos
24310
Lorentzos Roussos  
Johannes Katzer:
You don´t get my point yet. How can live trading results be much better than backtests running with real ticks and optionable commissions, real spread, etc.? How can this be? I am not talking about a specific tick data source. Meanwhile I tested all available sources from TDS. If anyone of you also has TDS, please contact me in PM to discuss this. Thanks a lot.

Simple , dont use TDS , use MT5 

As enrique said , the feed must be identical .So is there a way for you to record ticks and then pass them on TDS ? (i was off topic ,way off sorry)
Robert Jagger
786
Robert Jagger  
Johannes Katzer:
You don´t get my point yet. How can live trading results be much better than backtests running with real ticks and optionable commissions, real spread, etc.? How can this be? I am not talking about a specific tick data source. Meanwhile I tested all available sources from TDS. If anyone of you also has TDS, please contact me in PM to discuss this. Thanks a lot.


Don't put too much trust in the term 'real ticks', since it is the coding of the EA and not the settings or data in the strategy tester that matters most. 

This video covers this topic well. https://www.youtube.com/watch?v=dGFphm2bFbs

11.1) How to Ensure your Backtest Results are Indicative of Future Live Trading Performance
11.1) How to Ensure your Backtest Results are Indicative of Future Live Trading Performance
  • 2020.08.20
  • www.youtube.com
This video uncovers common issues with the use of price data in the backtesting process, and provides solutions - a ‘must-watch’ for any algorithmic trader. ...
Johannes Katzer
916
Johannes Katzer  

What can MT5 that TDS cannot? On TDS you can test even the most unrealistic conditions. For example I tried without spread, without commission, without slippage and I tried this for several tick data sources. The results are still bad every time. But in live trading the results are not that bad, so I wonder what a plausible explantion could be like?

There is no MT5 version of the robots I am talking about. But since they are from the MQL market place I would be happy to discuss this in PM with anybody interested. Thanks a lot for all your comments so far.

Enrique Dangeroux
735
Enrique Dangeroux  
Robert Jagger:


Don't put too much trust in the term 'real ticks', since it is the coding of the EA and not the settings or data in the strategy tester that matters most. 

This video covers this topic well. https://www.youtube.com/watch?v=dGFphm2bFbs

Ensuring your backtest results are indicative of future results is a different topic.