[WIP] Implements ELO rating for sunspotter

[Raahul-Singh]

This is the python implementation of the ELO score.
I need to figure out how to test this. Running it on the entire dataset will take a long time, and we cannot verify it otherwise.
Probably need to create some dummy data.
Finally, this may have some minute bugs that may have escaped manual testing. These will be sorted after we figure out testing.

Fixes: #7

[pep8speaks]

Hello @Raahul-Singh! Thanks for updating this PR.

• In the file pythia/cleaning/elo.py:

Line 54:101: E501 line too long (130 > 100 characters)
Line 66:101: E501 line too long (123 > 100 characters)
Line 141:101: E501 line too long (121 > 100 characters)
Line 143:101: E501 line too long (121 > 100 characters)
Line 160:101: E501 line too long (108 > 100 characters)
Line 161:40: E127 continuation line over-indented for visual indent
Line 170:101: E501 line too long (123 > 100 characters)
Line 188:101: E501 line too long (113 > 100 characters)

• In the file pythia/cleaning/tests/test_elo.py:

Line 15:45: E251 unexpected spaces around keyword / parameter equals
Line 46:101: E501 line too long (103 > 100 characters)
Line 47:101: E501 line too long (118 > 100 characters)

Comment last updated at 2020-06-30 10:51:35 UTC

[dpshelio]

Very good @Raahul-Singh I'm very happy with the quality shown in here! Well done. I've made some small comments (many are more of a conversation with myself 😅), you don't need to agree with them all.

Finally, maybe we want to keep this more general, then we may break this into ELO only and ELO for sunspotter.

[Raahul-Singh]

Rankings initially:

          image_id  score  k_value  count  std_dev last_scores
image_id                                                      
1                1   1400       32      0       32        1400
2                2   1400       32      0       32        1400
3                3   1400       32      0       32        1400
4                4   1400       32      0       32        1400
5                5   1400       32      0       32        1400
6                6   1400       32      0       32        1400
7                7   1400       32      0       32        1400
8                8   1400       32      0       32        1400
9                9   1400       32      0       32        1400
10              10   1400       32      0       32        1400

The matches are

    Player 1 id      Player 2 id     does player 1 win? (1: Yes, 0: No)
           1           6                           1
           2           7                           0
           3           8                           0
           4           9                           1
           5          10                           1

Rankings after each player have had one match:

      image_id   score  k_value  count  std_dev    last_scores
image_id                                                          
1                1  1416.0     16.0    1.0      8.0  1400.0,1416.0
2                2  1384.0     16.0    1.0      8.0  1400.0,1384.0
3                3  1384.0     16.0    1.0      8.0  1400.0,1384.0
4                4  1416.0     16.0    1.0      8.0  1400.0,1416.0
5                5  1416.0     16.0    1.0      8.0  1400.0,1416.0
6                6  1384.0     16.0    1.0      8.0  1400.0,1384.0
7                7  1416.0     16.0    1.0      8.0  1400.0,1416.0
8                8  1416.0     16.0    1.0      8.0  1400.0,1416.0
9                9  1384.0     16.0    1.0      8.0  1400.0,1384.0
10              10  1384.0     16.0    1.0      8.0  1400.0,1384.0

[Raahul-Singh]

@dpshelio if this looks good, I'll run it on the 14 years dataset and leave it to run overnight.

[drsophiemurray]

Nice tidy code @Raahul-Singh! Only a couple comments from me as I think @dpshelio covered most of it 👍

[dpshelio]

It seems good to me, so you can go ahead and merge it. You can create issues for the test data or names of variables to solve them later.