Port logic from standalone to catch cases when no live trading data is available about pyfolio

[twiecki]

Yes, the reason is that data_live is empty because the algo_created_date < last_returns_dt. I don't know why that would be the case though.

[justinlent]

ah good idea. yeah, this is quite commonly the case when the test harness backtest that was run does not include the algo_created_date in the backtest dates (this will be the case with very recent contest entries)

[twiecki]

Still odd that it would work in standalone but not quantrisk?

[justinlent]

in standalone, i added in logic to catch this case and then work around it. i should port over this code to the new nb to be robust to this condition possibly

[twiecki]

Great!

[justinlent]

@twiecki do you think we need this for the Public Beta? I believe it is only necessary logic when using one of our Q backtests.

Actually, I guess from the perspective of the bigger picture, I wonder if we need to just start maintaining 2 tearsheets from the beginning/now. e.g. tearsheet_public and tearsheet_internal, where tearsheet_public will the be tearsheet code that we release in the public opensource repo, and tearsheet_internal ends up eventually being in our internal (private) branch of the repo (or simply a whole new Quantopian private repo). Maybe you have some thoughts about organizing this as well?

[twiecki]

@justinlent Agreed. I removed the milestone. However, I think we really need to get rid of the multiple versions floating around as soon as we can or we'll continuously have to port NBs and will be out of sync. I actually thought that's what 8b8c0fc was doing?

[justinlent]

yep,i definitely should have caught this edge case in that commit.

[twiecki]

@justinlent I moved your change over to here now: d4eff10