ParabolicSAR descrepencies

[astrolobe]

I have a strategy that has a ParabolicSAR indicator added to it from within the strategy (orange dots in attached image). I have also manually added another ParabolicSAR indicator to the same chart that has the strategy. This ParabolicSAR indicator has the exact same parameters as the one added in the strategy (blue dots in attached image).

Question: Why do they not plot exactly the same?

Sometimes they are exactly the same, other times the blue accelerates faster, and other times the orange accelerates faster.

[NinjaTrader_Bertrand]

astrolobe, which market and timeframe do you run those comparisons on?

This would not only be seen at the start of your charts, so leaving init effects aside, correct?

[astrolobe]

I am running it on EURUSD 30 minute timeframe using Replay data from NT.
The sample image is after the Replay data has been running for 4 months.

[NinjaTrader_Bertrand]

Thanks, can so far not reproduce the issues here - which PSAR settings were you using? COBC set to false or true in replay?

[astrolobe]

COBC = false for both indicators
ParabolicSAR(0.05,0.2,0.05)

[NinjaTrader_Bertrand]

Thanks astrolobe, with my replay data and exactly replicating those settings from the indicator and strategy I could so far not hit it here - would you mind contacting me directly at support with your replay data for the 'issue' periods zipped up perhaps? I will then give your exact setup a run on our end.

[zacwhitesi]

Same Issue

Hi Bertrand,
What was the result of your findings? It would be nice for the rest of us, who have the same issue, to know.
I found a similar issue using Market Replay data with one PSAR using Calc on Bar Close = False and the other PSAR with CoBC = True. PSAR using the default periods. Here is a screenshot... http://screencast.com/t/7mbdcR1qpJ

Thanks,
Zac

[Harry]

There are about 5 to 10 different versions of the PSAR across different charting packages. The PSAR that comes with NinjaTrader is mostly correct.

1) There is a minor bug in the NinjaTrader Parabolic, as it does not cope correctly with wide ranging bars that breach both sides of the stop. In this case the Parabolic should revert, but the NinjaTrader PSAR does not revert. This is a rare case that does not happen very often.

2) The NinjaTrader PSAR is not CPU efficient as it recalculates the stop with every incoming tick when set to COBC = false. This has no impact on the result.

3) It needs to be discussed whether a touch of the stop line shall revert the position or whether price needs to break the stop. This question can only be debated after a decision has been made whether the PSAR shall be rounded to the full tick. Best have both options in any implementation. In case that you round the result for FOREX instruments will also depend on the "Quote currencies" settings used for your data feed.

More details are here:

https://www.bigmiketrading.com/ninjatrader/31700-bugs-ninjatrader-parabolic-code.html

[NinjaTrader_Bertrand]

Hi Bertrand,
What was the result of your findings? It would be nice for the rest of us, who have the same issue, to know.
I found a similar issue using Market Replay data with one PSAR using Calc on Bar Close = False and the other PSAR with CoBC = True. PSAR using the default periods. Here is a screenshot... http://screencast.com/t/7mbdcR1qpJ

Thanks,
Zac

Thanks for checking in Zac, could you please let me know your CL chart settings needed to reproduce?

[NinjaTrader_Bertrand]

Thanks Harry for summing those points up - they are definitely noted on in our Suggestion and Feedback trackings for the product and will be considered going forward.

[zacwhitesi]

Thanks for checking in Zac, could you please let me know your CL chart settings needed to reproduce?

That is an 8 brick Renko in the screenshot.