Method to calculate the most accurate method of prediction

[joshAU]

Hi, this is my first post, and I'm entirely self taught re statistics, so go easy on me.

I want to calculate which of several prediction methods generates the most accurate predictions for a given actual result.

eg. 1 have a table with 3 columns...
Col1 = actual results
Col2 = Predictions using method 1
Col3 = Predictions using method 2

ie..

Speed --- Prediction1 --- Prediction2
100--------99---------------103
102--------110--------------94
110--------112--------------110
108--------107--------------99
...and so on

How do I determine which method is the most accurate?

I searched these forums using the terms prediction and accuracy and came across a few posts, but most dealt with calculating the accuracy of one prediction to the actual result, not multiple predictions.
Furthermore, the posts I saw referred to using correlation to estimate accuracy.

To me, while using correlation will certainly reveal which method is the most correlated, it does not necessarily show which is the most accurate... by which I guess I define as the prediction where the mean of those predictions is the closest to the actual mean of the results, and where the standard deviation is the smallest.

While using correlation will help, it doesnt necessarily seem, to me at least, to help me fully.

Ie.. its possible that the correlation between the actual and one of the predicted columnns may be perfect, but the actual prediction is always double, eg, or some other figure that is correlated...

speed--------predicted1
100----------200
101----------202
99-----------198

if you see what I mean.

I guess I'm having a bit of difficulty verbalising my question, so if you need further info or have any questions, please let me know.

Thank you for any help/advice.
joshAU

[Link]

There are (at least) two good ways to do it. Either 1) calculate the MSE and compare between the models, or 2) compare the R^2 values (assuming you're using regression).