Market Watch: The primary indicator that Cook uses is the “Cook Cumulative Tick,” a proprietary measure he created in 1986 that uses the NYSE Tick in conjunction with stock prices. His indicator alerted him to the 1987, 2000, and 2007 crashes. The indicator also helped to identify the beginning of a bull market in the first quarter of April 2009, when the CCT unexpectedly went up, turning Cook into a bull.
What is a Cumulative Tick?
I am sure he is measuring market entropy, or its complement, market redundancy. He wants to know if the combination of ways the market delivers information is balanced with the combination of trades. When they are out of balance, the 'temperature' of the market will correct, up or down, to bring them into balance. This is Plank's curve type of analysis, measuring degrees of freedom in a Shannon optimized system. Looking at Fibonacci relationships does a bit of this, it looks for redundant estimates of value which can be eliminated, or decorrelated. ZeroHedge does a lot of this, looking for redundant correlations.
How would I do entropy analysis?
Pick an index, say the SP500, and run a windowed Huffman encoder over the thing. Vary the window size and the number of 'bits' until all the encoded trades are closes to each other in terms of -pLog(p), p being the probability of that trade out of the total trades in the window. Then discount stock which make the resulting -pLog(p) curve out of balance from a standard Plank, reward stocks which make it look like a standard Plank. In other words, make the decoding tree look like a minimum NlogN spanning tree. When that happens, the market is delivering the maximum value with the minimum number of trades, that is minimal redundancy.
No comments:
Post a Comment