For the first time today I estimated a model with weekly data. Again, I was stunned by the results - the high correlation between the forecasts and the actual change in the stochastic oscillator in the following week. The R-Squared (a measure of the goodness of fit of the model) between the one step ahead forecasts and the actual data is above 0.9 (very high). This is the sort of number that got my girlfriend to say "whoa!" when I told her about it :) Anyway, to cut to the chase the model predicts the stochastic will increase next week. Therefore, one should be long. I intend to use this model when the daily model is ambiguous as seemed to be at the moment. Well actually my model has clearly been long the last two days but I was stubborn and thinking that the potential turning point on Tuesday after Labor Day would actually lead to a turn to the downside. A day or two down is possible before the market continues to the upside.
Of course all this goes against the fundamentals as I see them. But so be it. The bond market is forecasting recession, but the stockmarket is beginning to change its mind.
Here is a chart that points out the analogous point last November and a guess of the likely price action going forward:
No comments:
Post a Comment