Simplify API

[yhoiseth]

Inspired by #60 and #61, I’ve been thinking about how to make the API as simple as possible.

The simplest case is for binary (true/false) predictions. We could, for example, have something like this:

import predictionscorer

probabilities = [A, B, C]

binary = predictionscorer.Binary(probabilities)

print(binary.brier_scores) # (X, Y, Z)

Here, the user inputs the probabilities chosen for the true alternative as a sequence, and can access the scores.

[yhoiseth]

@loeiten What do you think?

[loeiten]

This looks promising to me :). I like the simplicity.

If I read this correct A, B and C will in this case be three different predictions. Either the same person on various days, or three different days.

That makes me think:

1. Should the Prediction class handle this out-of-the-box. I.e. that the Binary class is hidden for the end user?
2. Would it make sense to allow nested input? For example

import predictionscorer

george_on_different_days = [A, B, C]
kramer_on_different_days = [D, E]
probabilities = [george_on_different_days, kramer_on_different_days] 

binary = predictionscorer.Binary(probabilities)

print(binary.brier_scores) # ((X, Y, Z), (V, W))

or for non-binary prediction would it make sense to do something like this

import predictionscorer

george_on_day_1 = [a1, a2, a3]
george_on_day_2 = [b1, b2, b3]
george_on_day_3 = [c1, c2, c3]

kramer_on_day_1 = [d1, d2, d3]
kramer_on_day_4 = [e1, e2, e3]

george_on_different_days = [george_on_day_1, george_on_day_2, george_on_day_3]
kramer_on_different_days = [kramer_on_day_1, kramer_on_day_4]
probabilities = [george_on_different_days, kramer_on_different_days] 

prediction = predictionscorer.Prediction(probabilities)

print(prediction.brier_scores) # ((X, Y, Z), (V, W))

?