Easy to do. The currency banker watches the variance in incoming deposits and incoming loans. It wants the differences in variances to fall in a band. When the band limit is exceeded, the currency banker pays off the bets. Mathematically, the currency banker can simply add or subtract to each member bank deposits and loans. That assumes transaction costs are zero.
The variance band is the currency float, the amount of currency at risk which is used for price discovery. The band is pre-announced. But actual rates and terms are discovered after the fact, pay off is for events that have been uncovered. The currency banker is carving out a rectangle where it operates, at the short end of the yield curve. As the variance band is violated, the currency banker tales gains and losses. But there is no time in the model.
Transaction costs? Azure or ethereum, either one gets you close to zero. The control bot? Its a spreadsheet function. I think Microsoft has one working.
What spreadsheet function?
IO thonk of it as an imprecise Huggmsn encoder which creates a 'block chain' that is a decoding graph, it generates a representative sample of the past transactions. The rransactions are the arrival of deposits and loans. The Huffman decoding graph has to change configuration, slightly. It is a finite block chain of the member bank activities. on configuration change,n the winning member banks are those who acted close to a representative sample, they produced the most insider information.
No comments:
Post a Comment