Strategy analyzer issue

[tradejockey]

Please see attached file. entry price and exit price dont match the time (forexample: 10/2 11am - price was between 2748 to 51 but strat analyzer is showing something else 2740 entry 2747 exit both at same time 11am). Used seconds as interval.

[NinjaTrader_Josh]

tradejockey,

I suggest you open up the Chart tab and see how it stacks. Remember that orders don't have to fill on the bar as it is possible to fill outside the bars in backtesting.

[tradejockey]

Originally Posted by NinjaTrader_Josh

Remember that orders don't have to fill on the bar as it is possible to fill outside the bars in backtesting.

I fail to understand how this would be considered an acceptable behaviour. If the instrument has not traded anywhere close to that price in bars surrounding the time mentioned, then wouldnt that be a red flag? Not sure where NT's QA team stands on this.

What we need is a robust Backtesting module that we can believe in. It doesnt have to be accurate but should be somewhere in the ball park. Close enough is good enough. But it should inspire some sort of confidence. Granted past performance is not indicative of future success..

What is of concern is that lately I've seen statements such as "live sim > replay > backtest". So, does it mean we have to run scripts in live sim for years to find out if it works? I understand all the nuances of ticks coming in late or that replay has 0 delay etc. But if both replay and backtest use same historical data, why would replay be considered better than bt? If anything I would like to see that order reversed i.e BT > live sim / replay

[NinjaTrader_Ray]

Originally Posted by tradejockey

I fail to understand how this would be considered an acceptable behaviour. If the instrument has not traded anywhere close to that price in bars surrounding the time mentioned, then wouldnt that be a red flag? Not sure where NT's QA team stands on this.

What we need is a robust Backtesting module that we can believe in. It doesnt have to be accurate but should be somewhere in the ball park. Close enough is good enough. But it should inspire some sort of confidence. Granted past performance is not indicative of future success..

What is of concern is that lately I've seen statements such as "live sim > replay > backtest". So, does it mean we have to run scripts in live sim for years to find out if it works? I understand all the nuances of ticks coming in late or that replay has 0 delay etc. But if both replay and backtest use same historical data, why would replay be considered better than bt? If anything I would like to see that order reversed i.e BT > live sim / replay

It is acceptable since in the cases where it would fill off the bar, it should only happen as you would place (one example) buy limit order at the close of the bar for a price higher than the high of the next bar. In real-time, you would have been filled at that price (intraday anways) and thus you would get a worse fill and have more conservative performance results. If you don't agree with that logic, use the "Liberal" fill type which if memory serves me correctly you will get filled at the more optimal price on the bar.