The version 7 ano that you published Friday,1/11 was above 0 and not in
a short position and in fact version 7 had not been short since last July, early
Aug. This week, perhaps over last weekend the input numbers for 7 were overhauled
and it shows version 7 having been short in Oct. and going short in early Jan.,which
your numbers relate to. However,as of last weekend version 7 was holding a cash
position all this week.
Since version 7 is a developmental version, I don't show a table of actual
output. The values that you quote must have been taken off the graph that is
shown on the forecast page. Errors in estimating the values from the graph are
possible. Also, the learning algorithm will change previous values slightly to
improve forecasts going forward. Usually, the changes are small and unnoticeable.
This could be the source of the differences that you see. Since version 10 is
our standard model, previous values are fixed and not allowed to change.