Market Replay vs RealTime

[makhmout]

Hi all!

I have created strategy that shows good results in 6 month market replay. It is based on Renko charts.Yesturday I have lauched it in real-time, but results were opposite. I tried to run market replay on yesturday's data and have got positive results. I absolutely don't understand it. As I think, market replay just runs all deals that have took place on market in past and it can't be different from yesturday's real-time. But in real-time I got -$5000, in the same day's market replay I got +$5000. Such difference can't be explained as the result of slippage. It seems that it driven by order placement and execution. Because of a lot cancelled orders in real-time, which I can't understand yet. Can anybody explain me the reason that explains such great difference between market replay and real-time deals?

P.S. I clearly understand that the backtest of renko graphs is fully differ from the real-time one, because of it tested my strategy on market replay ony.

Regards, Makhmout

[NinjaTrader_Bertrand]

Hi Makhmout, what market / instrument and datafeeds do you run this on please? What exact Renko type do you use here? The default NinjaTrader one?

Thanks,

[makhmout]

Hi Bertrand,

I'm using 10 renko default NT charts and CQG datafeed.
GC 06-12

[NinjaTrader_Bertrand]

Thanks, how active would that strategy be typically? Are we talking a few intraday swing trades here or an aggressive scalp type model?

[makhmout]

My strategy generates deals in every renko bar.
It looks similar to agressive scalping.
I think we can find problem only by viewing every deal
in markey replay vs every deal in real-time.
So, can you explain me, does any difference
in order placement and execution in market replay and
real-time trading?

Can problem be generated by many copies of strategy,running in
real-time? I run few versions of one strategy with different parameters.
On other side, I run the same set of strategies in market replay,
and results are ok.

I fully don't understand.

[NinjaTrader_Bertrand]

They should be very close from my experience - you would need to compare the individual executions seen for live simulation vs replay to see where the difference would be for your strategy.

Do you test on self recorded CQG replay data or would you use our servers (which holds recorded ZenFire replay data)?

Could it be perhaps a different set of commissions entered that would show on the very high amount of trades done on a daily basis for your script?

All the best,

[billr]

In a related question, will stops/profit targets/limits etc. execute the same way in market replay vs. live trading? (I'm speaking mechanically, ignoring any broker order slippage/discrepancies)
Thanks.

[NinjaTrader_Bertrand]

That's basically correct Bill, same orders types / events are used.

[billr]

Thanks Bertrand, but to be more specific, the recorded data feed is fed to NT in the same way as live data would be and therefore none of the logic errors you might find in backtesting via Strat Analyzer occur (especially in Renko). Correct?

[NinjaTrader_Bertrand]

Correct, Market Replay Renko strategy testing will be more accurate than purely backtesting in the SA with the primary series only.