Cash flow

In the theory, the last part, we talked about loans/deposits and its change over some fixed trades pace. We got R * R', which we show had a peak as the window size went from one to larger.  This peakness property is due to the assumption of sphere packing.

When flow from loans to deposits is maximum bit error is minimum. When RR' is maximum., flow between loans to deposits should be maximum. The matching algorithm has found window lengths in each sequence such that the two are most closely congruent, gets us sparse repeating fraction, I think.

All this means is that loans mostly divide deposits will the smallest residual. The pit boss can put together a group of deposits to match most loans, fairly, like a consortium. In the structured queues,there should be bisn, sort of one bin of depositors per one bin of loaners when the two generator are normalized.

This analysis says something about a coin exchange, they have to have long windows. huge flowm to keep the bit error proportionally small. The coin exchange pit boss is not going to carry measurement error between two economies. This is a repeat subject, coin exchange tend to be large window high flow rate when they work. The exchange should be quick on changing the fee to force it accumulated matching error as small as possible.n It ends up running two coherent one coherent queues, managed by exchange fee. He uses these fees to keep window sizes large enough on either side.