BHedge

Scientist & Engineer (NLP, stats, some CV). I am here for fintwit. Europoor.

[BHedge @BahluiHedge]

@calvinfroedge @AndreasSteno Also imo you are right @calvinfroedge. Low variance and slightly higher mean is exactly what is needed to make existing things cheaper and better.

[BHedge @BahluiHedge]

@calvinfroedge @AndreasSteno Averages matter (no pygmy math/phys geniuses) but variance is the actual driver for extreme outcomes.

Assumes Lynn and Murray were right not Hsu.

[BHedge @BahluiHedge]

@bryan_johnson All governments.

[BHedge @BahluiHedge]

correlation based on non-stationary data yields a random variable. Corr is well defined but sample corr does not converge to it. You diff(1) to try to make it stationary.
@INArteCarloDoss @nickgiva1 @CumLordeAwards Is Steno a “guru“? Why isn't anyone saying anything?

[Andreas Steno Larsen @AndreasSteno]

@GordonJohnson19 @GLJ_Research @crossbordercap Correlations will have to be measured in non-stationary time series (i.e. changes) and then you see that central bank liquidity is only a fraction of the total explanation

This is statistics 101 mate

[BHedge @BahluiHedge]

@WifeyAlpha You can add reversal to trend momo by fitting a 90-deg rotated S-like sigmoid on the output. This will revert the pnl allocation to zero when trend momo gets extreme. You knew that already, it works don't doubt it.

[BHedge @BahluiHedge]

@kerberos007 Clustering in what feature space?

[BHedge @BahluiHedge]

@adrianscottcom Gm