R squared or coefficients for prediction

[xeon123]

I created a little model that predicts the time a task will take to execute. Than, I calculate the real time that the task took to execute. With the predict and real value I have the error of my prediction. I want to use that error to predict the next execution (the next execution is a similar task has the one I ran). I was looking to R-squared or coefficients to help me. Does anyone has an opinion about this. What is the best use I can do to the error of my prediction to help to predict again?

For now I am using Adjusted R^2, but I don't know other methods.

 

[Mark44]

I would use the mean squared error (see https://en.wikipedia.org/wiki/Mean_squared_error). The formula is $$MSE = \frac 1 n \sum_{i = 1}^n(Y_i - \hat{Y_i})^2$$
Here $Y_i$ is the actual time of the i-th task, and $\hat{Y_i}$ is the predicted time of the same task.