Factor Interaction with Size

[whotookmynickname]

Thank you for your work and your great community input! This is very much appreciated.

While this analysis does not look at factor specific interactions per se, ranks are derived from a comprehensive system inclusive of several factors types, including:

• Peer group technical momentum (without technical mean reversion, which is a variation that improves performance expectancies, especially of small and micro caps)
• Valuation (Absolute and Relative)
• Fundamentals (run of the mill “quality” factors)
• Street Expectations (value and sentiment; omitted for stocks with no coverage)
• Market inefficiency (i.e., relative analyst coverage, liquidity, volume, institutional coverage, and short interest)

If this is not too much additional work, it would be also very interesting to see the factor type interaction with size.
So how the 3D histogram would look like for each factor group (as mentioned by your, e.g. Street Expectations) on a standalone basis.

[primus]

whotookmynickname,

That’s also on my bucket list, but unfortunately it will have to wait until I have some additional free time.

//dpa

[ivillalongabarreiro]

Hi Primus,

When I click on your simple mean reversion system link, I get access denied, I got designer membership. would you share this with me?

Thanks,
Javier

[primus]

Javier,

I deleted it a while ago to make room for new things.

The essence of it was simply to create a standardized “Z” score which symbolizes how many sigmas a given move deviates from the expectation.

The trick is never how to capture a given move. We know what happened.

The trick is to how to setup an expectation and how to normalize the variation. The simplest setup is just to compare a move versus an historical average. It is only slightly more sophisticated to compare this difference over its historical standard deviation. E.g.,:

(Ln(Close(0)/Close(5)) - Ln(Close(0)/Close(255))/51) / LoopStdDev(“Ln(Close(CTR)/Close(CTR+5))”,51,0,5)

We can get fancier: e.g., do we only care how a stock moved versus its own history? Or do we care about its move relative to peers? If the latter, do we care about the size of the move in absolute or relative terms? Finally, are we concerned with correlation, cointegration, or both?

I hope this helps.

//dpa