Posted By: Roy   Date: Thu Nov 5, 2009

Title: cash 11/2

Message:
  If you had been using the new library before 11/2, would you have gone to cash sooner than 11/2?
  Roy

Response:
  In this last library update, I replaced about one-half of the 275 networks used to calculate the forecast (ANO). This updated library would have gone to cash several times in the last three months but it did not stay there more than a day or two. As you recall, the previous library held a long position continuously during this time. A more significant system change was made in early August. I modified the algorithm that weights each network in the calculation of the ANO. This change corrected the short trading position in July.

  I think the real question is how well does the performance data on the forecast page reflect the current model? The answer is probably not very well. The current model would produce a return of about 80 percent over the last 165 days while the value in the performance table on the forecast page currently reads 18 percent. Reporting past performance presents an issue when the system changes. I want to keep the performance data blind to show actual returns, but it may be a poor indicator of how well the model will perform in the future.