Dear Kates, thank you very much for taking the time to write me...

Dear Kates, thank you very much for taking the time to write me such a smarmy response given you are obviously a "professional trader".
Let's get onto your graph and what it is actually saying:
1. You've got data plotted at set intervals (known by you) plotted as candlesticks.
2. There is a "high cut" regression which so smooths the data that 95% of data over the last couple of months fall below this line.
3. There is a "low cut" regression which so smooths the data that >60% of data over the last couple of months fall above this line.
4. The "range" between the 2 regressions is 40c.
5. The "range" is >100% of the current stock price which is fantastic if you wish to represent and incorporate all "previous" price points but hopeless as a predictive (forward casting) tool but it's not all bad because rest assured stock prices almost never increase by >100% in a single session and if this unlikely event does occur - rest assured you will always be correct with you're regression if you wait "x-days" to plot you're data as it will always be incorporated into a "post predictive" regression model.

I personally not being a "Professional Trader" wouldn't know what they are using to "predict or forecast" future stock prices but as someone who has used statistics quite often can tell you that there is nothing "predictive" in the regressions you've plotted.

When analysing data I find it best to find the (equation, function, curve) that "best" represents the real data points. As I am limited to "unprofessional charting applications and statistical tools" I try to keep it simple - Hence I find the 5 day MA most closely approximates the data.

As for a predictive model (I do not have one) however if someone were using such a tool I would expect it to be based on the following:
1. "Forecast data" be subject to the same level of "smoothing or averaging" as the function that best represents the data.
2. "Forecast data" be represented as Confidence Intervals.
3. "Forecast data" have a tight range to be as useful as possible for it's intended purposes- prediction.