Live Sim and Replay discrepancies

[Downhillryder]

Hi

I am having problems getting consistent backtest results. I have collected 3 sets of trade data whilst testing my strategy. I'll use just the first trade from each set to illustrate the problem :

Trade 1 collected from Strategy running on Gold Live Zen Feed on Sim01 -

Entry Short 1574.2 Exit 1573.6

Trade 1 collected from Strategy running on Gold Replay data from ZenFire on Replay01 for same period as above -

Entry Short 1574.7 Exit 1573.5

Trade 1 collected from Strategy running on Gold Sim01 after shutdown and restart for the same period as above -

Entry Short 1573.9 Exit 1573.2

All timestamps show execution in the same second.

I have attatched spreadsheets for that days trading of my strategy. The strategy is running on "GC 0413" contract , 233 Volume charts and (as per suggestions on this forum to aid backtesting) dispatches Entry/Exit trades using a supplementry 144 tick instrument.

Needless to say the discrepancies represent a large difference between profit and loss calculations. I have attached the 3 spreadsheets.

Any ideas/help in getting consistent results would be appreciated.

regards dave

[NinjaTrader_PatrickH]

Hello Dave,

Thank you for your post.

There can be discrepancies between real-time and historical data, and data can be different after reloading historical data.

For information on why data can look different after reloading historical data please visit the following link: http://www.ninjatrader.com/support/h..._are_built.htm

For information on Discrepancies Between Real-Time and Historical Data please visit the following link: http://www.ninjatrader.com/support/h...ime_vs_bac.htm

Please let me know if you have any questions.

[Downhillryder]

Thnx Patrick.

I have read those links , but have just reread them in case there was something new.

So in summary , if my Strategy scalps for 4/5 ticks the only way I'm going to know it
is profitable is to run it live ? (Because the stats I gathered show upto 7/8 ticks margin
of error).

Is there nothing that can improve the accuracy ?

regards dave

[NinjaTrader_PatrickH]

Hello Dave,,

Thank you for your response.

As detailed in the links you reviewed you should expect to see differences in the results from real-time versus historical. Running in real-time is the best way to see the results of the strategy as it would be ran in real-time, you can use the SIM101 account to place the trades to in simulation during your tests.

Please let me know if I may be of further assistance.

[Downhillryder]

Hi Patrick

So are you saying the "Live Sim" trades (first attached doc) are likely to be the most accurate ?

thnx

[NinjaTrader_Matthew]

Yes, these should be more realistic, but keep in mind there is no way to completely simulated market dynamics. The accuracy will be relative to the best of our abilities.

[Downhillryder]

Ok. Thanks for your help.

regards dave