Ninjatrader realistic fills?

[nourozi]

Hi there, I would like to test my automated strategy in live sim with Ninjatrader. I am just wondering how accurate the live sim order engine is in Ninjatrader? Particularly, with limit orders.

Specifically I am interested in finding out what % of my limit orders are likely to fill on the touch, without penetrating.

Ninjatrader describes their own simulation engine as follows:

"The simulation engine is not a simple algorithm that fills your order once the market trades at your order price. The engine uses a scientific approach to determine fill probability by including a number of variables including: ask/bid volume, trade volume, time (to simulate order queue position), and random time delays for switching between order states."

Will live sim give me a good idea?

Thanks, Erfan.

[NinjaTrader_Bertrand]

Hi Erfan, yes I think the simulation engine will be ideal for this kind of testing. Especially with tougher settings delay wise and Enforce immediate fills unchecked (while potentially partials checked depending on your size) you should get a realistic perspective - keeping in mind it would be an approximation still.

[nourozi]

Hi Erfan, yes I think the simulation engine will be ideal for this kind of testing. Especially with tougher settings delay wise and Enforce immediate fills unchecked (while potentially partials checked depending on your size) you should get a realistic perspective - keeping in mind it would be an approximation still.

Is the simulation engine particularly suited to one market, for example the ES? Or, does it simulate all markets equally well? Also what % of accuracy can I expect from the NT simulator?

For example, can I expect my live results to be within say 10% accuracy of NTs simulator?

Basically, what do you define as an approximation. What level of approximation as this is important before deciding to go live with an automated strategy.

Thanks.

[nourozi]

Also, on live sim, do limit orders need to be penetrated in order to be filled, or does NT do a good job of simulating limits fills even on the touch?

[NinjaTrader_Bertrand]

nourozi, limit orders can be filled on touch and penetration as well, with Enforce Immediate fills unchecked this will be factored in - otherwise you can just simulate being basically always in pole position in the order q.

It's just an approximation still albeit from our and our customer experiences we believe a realistic one - an accuracy could not be given for those scenarios since realtime dynamic market behavior simply cannot be predicted, also not including the impact on other market participants your own orders could have.

[nourozi]

nourozi, limit orders can be filled on touch and penetration as well, with Enforce Immediate fills unchecked this will be factored in - otherwise you can just simulate being basically always in pole position in the order q.

It's just an approximation still albeit from our and our customer experiences we believe a realistic one - an accuracy could not be given for those scenarios since realtime dynamic market behavior simply cannot be predicted, also not including the impact on other market participants your own orders could have.

Does Market Replay work the same way as Live Simulation? Which one is more accurate for testing an automated strategy which uses limit orders?

Can you explain the differences?

Thanks,

Erfan.

[NinjaTrader_Bertrand]

Market Replay works off recorded live data from your data source or our servers, this will provide you with an ability to test your strategies using intrabar information and preserving exact event sequences for the Level 1 and Level 2 Market Data streams - the trading / fills would also be determined by your simulation engines, so filling behavior would be very comparable withstanding any small differences in data you might experience across feeds.

[nourozi]

Market Replay works off recorded live data from your data source or our servers, this will provide you with an ability to test your strategies using intrabar information and preserving exact event sequences for the Level 1 and Level 2 Market Data streams - the trading / fills would also be determined by your simulation engines, so filling behavior would be very comparable withstanding any small differences in data you might experience across feeds.

So are you saying that Live Simulation would give almost the same results as Market Replay with Level 1 and 2 data?

[NinjaTrader_Bertrand]

Correct I would expect them to get very close. The usual recommended cycle would be -

Backtest > Market Replay > Live Sim > Live.

[nourozi]

Correct I would expect them to get very close. The usual recommended cycle would be -

Backtest > Market Replay > Live Sim > Live.

OK, thank you for your help

[nourozi]

Hi, I was just testing live simulation with limit orders on GC gold futures and am constantly getting filled with positive slippage... Is the ninjatrader live simulation suited more towards highly liquid markets such as the ES?

Or should I consider these fills as a 'realistic perspective' and 'approximation'?

Thanks,

Erfan.

[NinjaTrader_Bertrand]

Positive slippage can happen, but should not be constantly happening from my understanding - what exact order type do you use?

[nourozi]

Positive slippage can happen, but should not be constantly happening from my understanding - what exact order type do you use?

Not sure if it is exactly 'positive slippage' but I place a limit order. Then once I am filled, I am instantly in profit by a few ticks. Maybe it is because of the benefit of using limit orders on markets with a large spread?

[NinjaTrader_Bertrand]

Correct, that would not really be called slippage - slippage would be I place a market order and expect a fill at 1250 and then I'm in 2 ticks later only, so I got filled with 2 ticks of slippage.

[nourozi]

Correct, that would not really be called slippage - slippage would be I place a market order and expect a fill at 1250 and then I'm in 2 ticks later only, so I got filled with 2 ticks of slippage.

Ok, so sounds like nothing unusual is happening...