One Notebook, Grant that Luca posted to share his study on technical indicators may be of interest to you?

I found the study fascinating, eg. IchimokuKinkoHyo, FastStochasticOscillator, AnnualizedVolatility - a head start for you to add RSI/short_term_reversal, momentum and the Q Risk Model factors?

There is also a study by Luca on the cumulative returns of the risk factors in 11 sectors and 5 styles in the Q Risk Model.

Hope this helps. Keen to see your findings.

Thanks. Will take a look as time allows. If you learn anything more about the risk model factors, please let me know. I guess they are built-in Pipeline factors, but were released under quantopian.pipeline.experimental. They are the same as in quantopian.pipeline.factors described here, I suppose.

Thanks Luca, when I try to run it, I get:

In [4]: 

    prices_cache, factor, factor_data = \

    run_tear_sheet( factor       = -Momentum,

                    factor_name  = factor_name,

                    start_date   = start_date,

                    end_date     = end_date,

                    top_liquid   = top_liquid,

                    show_sector_plots = show_sector_plots,

                    avgretplot   = avgretplot,               

                    periods      = periods,

                    quantiles    = quantiles,

                    bins         = bins,

                    filter_zscore = filter_zscore,

                    long_short   = long_short,

                    group_neutral = group_neutral,

                    prices_cache = prices_cache)


TypeErrorTraceback (most recent call last)  
<ipython-input-4-43b7d72b8ee4> in <module>()  
----> 1 prices_cache, factor, factor_data =     run_tear_sheet( factor       = -Momentum,  
      2                     factor_name  = factor_name,  
      3                     start_date   = start_date,  
      4                     end_date     = end_date,  
      5                     top_liquid   = top_liquid,

TypeError: bad operand type for unary -: 'ABCMeta'

Here is an updated NB

Here's the whole enchilada from Luca--thanks! Interpretations? My working hypothesis is that the Quantopian style risk factors are weakly predictive, at best, and just noise, worst-case. As "common" factors, they would have had the life sucked out of them a long time ago. But I could be wrong.

EDIT - TO DO: Note, this should be run as far back as the data will support. The time frame is too limited, at this point.

@ Luca -

I'm trying:

start_date  = '2006-01-01'  
end_date    = '2018-06-01'  

Seems to be hanging, but maybe I just need to be patient. How many years would you expect it to handle?

No limits Grant, the line

daily_factor = run_pipeline(pipe, start_date=start_date, end_date=end_date, chunksize=250)  

set the chunk size to 250 days so it shouldn't run out of memory. I successfully ran it many times with 15 years of data

Thanks Luca. It does run, although slowly. Q engineering needs to shovel a bit more coal into the firebox.

By the way, will a notebook keep running in the background? Or do I need to keep it open in the browser?

Finally got it to run. I used:

start_date  = '2016-03-08'  
end_date    = '2018-06-04'  

These dates cover the current contest.

At a high level, is there a conventional pass-fail for assessing factors? For example, a 95% confidence level that a given factor is predictive?

Basic fund template alphalens