He’s usually precise but less often accurate . A regular problem in forecasting is generating precision far outside the sphere of confidence. Frequently there are those who display ranges, which is better than false precision.
The master problem in forecasting is using com with wide confidence ranges, but showing on a fixed value. Every time that is repeated the expected values further lose confidence.
Long ago I was tasked to evaluate an acquisition candidate providing a five year forecast. Following the project the acquisition was made. Five years later I was strongly complimented for my stellar forecasting talents, since the bottom lines happened to be within 3% of my forecasts. I was astonished and thrilled ! Then I made a fatal mistake. I reviewed my model against actual results. Nearly all components were off by 50% or more for every year. By gigantic but fortuitous coincident, the bottom lines were right on.
Ever since then I have remained acutely aware of the ‘last significant digit’. As a result my own forecasts all consider ranges, never absolutes. Some, like
@The Accountant deal with this with multiple cases, which is the optimal way for general audiences. For internal use I use ranges for each variable, then attempt to adjust others for dependencies. That is very hard to display. Securities analysts rarely seem to do any significant analysis, just compare with purported competitors.
Woth Tesla the complexity is rapidly rising and accurate data is becoming more difficult to acquire. Traditional data sources are of diminishing value. Frankly, the value of TMC is rising quickly, just as the volume of irrelevancies is growing.
Bluntly our mods are doing a stellar job trying to keep us focused on investment evaluation. They are to be applauded and,perhaps, pitied! Luckily some of our mods live in such remote areas that being our disciplinarians might serve as comic relief. I don’t know, but I thank them!
