First results

This would be the Yield Curve of the complete S&P 100 for a single trading day.  Actually it is the symbol length from the encoder, intrepreted as signal power plotted against the typical period between observations, or one/transaction rate. Prices were weighted by volume, and the resolution reduced. Bandwidth normalized to one, trading efficiency equal log(2).

No conclusion here, except to illustrate the point, encoding is a model of a production system. We can see that grouping the members of the S&P 100 according to information content implies fewer groups of traders along the X axis are needed, we get a good fit of the trades into a restricted channel with less work.

Once I implement trading strategies limited to a graph algebra, then we can see if trading adapts fast enough to control for changing information content; that is, does generalied yield curve get inverted. I also noticed some fund managers adapting their strategies to a volatility curve, based on the same formula I suspect.

What would this tell us about optimum government? A sudden chunk of Congressional inventory results in government traders crowding around Washington DC to break the inventory into managable cargo sizes.  We get the same effect with the Fed and the QE(n) game.

The other question that might come up is why is the numerator a little different than in the generalized formula.  The generalized formula is  simply the optimum transaction set and its own quantization error,  a Fibonacci series measuring the trading efficiency value.  The numerator changes when dealing with real data, it goes from log(Wi), to the actual cargo sizes of the economic work.