Recent change in fundamental data after 28th Oct 2023

Based on the provided information, it's not possible to identify any bugs in the code. The code seems to be logically correct and should work as expected. However, the change in backtest results could be due to updates or changes in the fundamental data.

QuantConnect sources its fundamental data from Morningstar, and it's possible that Morningstar updated or corrected their data after your initial backtest. This could result in different ROA values and hence different universe selection results.

Unfortunately, without logging the data or having access to the historical data used in the initial backtest, it's hard to pinpoint the exact cause of the discrepancy.

For future reference, it might be helpful to log the data used in your universe selection to help debug such issues. You can use the self.Debug() function to log data in QuantConnect.

For example, you could add the following line to your SelectFine function to log the ROA values:

self.Debug(f"{x.Symbol.Value}: {x.OperationRatios.ROA.ThreeMonths}")

This will log the symbol and its ROA value for each security in the fine universe.

Please note that logging can slow down your backtest, especially if you're logging a large amount of data. It's generally recommended to only use logging for debugging purposes and to remove or comment out logging statements once you're done debugging.

If you continue to experience issues or discrepancies in your backtest results, I would recommend reaching out to the QuantConnect support team for further assistance.

Hi Nguyen Huu Quan , 

We verified the data is correct and present. In reviewing your code, it looks like you've got a top 100 filter([:100]), and likely it's not being included in the top 100 anymore as other (now delisted) assets are being included. The number of assets covered in the data doubled from 5K→11K to include many delisted companies. 

If you shuffle the filters, you can probably make it appear if that entity is necessary.

Best

Jared

Many thanks for the super helpful insight Jared Broad This makes sense now. So actually so far my algo kind of had a look-ahead bias with the (now) delisted assets being excluded at various points in time of the algo. 

Let me try to play around with the filter to see if I can recover some of the old selections.

Best,

Quan