Limit order fill accuracy in replay

[dirkdiggler]

Hello,
Is there anything that can be done to a strategy that will enhance the accuracy of market replay with respect to limit orders? I have my strategy with limit entries that submit orders on a 1 Tick data series. Take NQ for example, let's say I run my strategy live today, it may show 10 trades filled via limit order entries in a live account. If I re-run that same strategy over the same CQG data in market replay, there is usually a significant amount more entries made, sometimes triple the # of trades. In a live market these trades were not filled. My VPS server latency to CQG is very low. In general terms is there anything that can be done to have market replay show more realistic limit entry fills, such as adding a 1 Tick Bid and 1 Tick Ask series? Or any other strategy configurations to make replay more realistic?

[NinjaTrader_Jesse]

Hello dirkdiggler,

To better understand what is happening you would need to add prints to the script to ensure that the entry conditions are becoming true identically as it did in realtime. Without doing that kind of comparison we wouldn't be able to say if its due to fill accuracy. You already submit orders to a 1 tick series so that would be the most granular you can use for order submission.

[dirkdiggler]

Prints are added in already, the entry conditions align, it’s just a matter of replay showing the order as filled versus in real time we are just not getting the fills as often. I am also a vendor and the strategy is coded very well, we’re just trying to understand the replay logic a bit more as to why there are so many more fills in replay versus live. Does replay account for a certain network latency between order submission and being received by broker?

[NinjaTrader_Jesse]

Hello dirkdiggler,

The playback account uses the same general settings as a Sim101 account for fill processing. There is some simulated delay however it will not specifically match internet connection -> broker directly for a variety of factors.

You may want to try recording your live data as playback, if you are downloading it from the historical data server that may be another source that produces differences.