Testing price predictions

[insighter]

Technical analysis should lead to a prediction of future price development. To get more calibrated and accurate over time it's helpful to get your predictions scored and your reasoning documented so that you can learn without fooling yourself. A friend and I have built a forecasting tool that let's you do exactly this. Here's my 80% confidence interval for the Tesla stock as an example: https://www.empiricast.com/time-series/35

Do you guys think it's useful to make price predictions in this way?

 

[MasterOfCoin]

No, I don't.

Predictions get less and less accurate and so less and less useful the farther ahead you cast them.

As that's a direct consequence of any dynamic system with more than one variable, it makes predictions near to useless.

The theory and equations can be 100% perfectly right. The prediction will not be.

You can never define the conditions in a variable system accurately enough to make reliable predictions.

Will it rain in Paris next Monday afternoon ?

 

[insighter]

No, I don't.

Predictions get less and less accurate and so less and less useful the farther ahead you cast them.

As that's a direct consequence of any dynamic system with more than one variable, it makes predictions near to useless.

The theory and equations can be 100% perfectly right. The prediction will not be.

You can never define the conditions in a variable system accurately enough to make reliable predictions.

Will it rain in Paris next Monday afternoon ?

Making accurate predictions is for sure difficult, but when making investments what you do is really to make a prediction of future price developments. If you believed that the price was going down you would wait to invest. Over time when getting your predictions scored you would get feedback on how accurate you are, on what kind of asset classes you are good/bad at forecasting and see how calibrated you are (with a 80% confidence interval, does 8/10 outcomes fall within your stated interval)