Write a new documentation page with branch info (#1367)

* Write a new documentation page with branch info

This is related to #1352.

Once this PR is merged we should:

- Change github default branch to `dev`
- Delete `master`
- Confirm that read the docs is looking at `release`

Am I missing anything?

We should also remove the release process information from the wiki
(assuming we're happy with what I've written here).

Finally, once #1360 is done we should make sure we update the docs with
the relevant information.

* Remove ambiguous `very`.

* Remove hypothesis version specification in docs.

This is actually no longer correct since #1288

* Test properties not affected by floating point error

This build found a particular failing example of
`TestTournament.test_seeding_equality` https://github.com/Axelrod-Python/Axelrod/pull/1368/checks?check_run_id=975415322

Upon closer investigation it looks like that was not due to seeding but
due to the floating point error of some calculations made by the result
set.

I investigated using:

```
import axelrod as axl
import numpy as np

seed = 2
repetitions = 10
rng = axl.RandomGenerator(seed=seed)
players = [axl.Random(rng.random()) for _ in range(8)]
tournament1 = axl.Tournament(
    players=players,
    turns=10,
    repetitions=repetitions,
    seed=seed
)
tournament2 = axl.Tournament(
    players=players,
    turns=10,
    repetitions=repetitions,
    seed=seed
)
for _ in range(4):
    results1 = tournament1.play(processes=2, progress_bar=False)
    results2 = tournament2.play(processes=2, progress_bar=False)
    assert results1.wins == results2.wins
    assert results1.match_lengths == results2.match_lengths
    assert results1.scores == results2.scores
    assert np.allclose(results1.normalised_scores, results2.normalised_scores)
    assert np.allclose(results1.ranking, results2.ranking)
    assert results1.ranked_names == results2.ranked_names
    assert results1.payoffs == results2.payoffs
    assert results1.payoff_matrix == results2.payoff_matrix
    assert np.allclose(results1.payoff_stddevs, results2.payoff_stddevs)
    assert results1.score_diffs == results2.score_diffs
    assert results1.payoff_diffs_means == results2.payoff_diffs_means
    assert results1.cooperation == results2.cooperation
    assert results1.normalised_cooperation == results2.normalised_cooperation
    assert results1.vengeful_cooperation == results2.vengeful_cooperation
    assert results1.cooperating_rating == results2.cooperating_rating
    assert results1.good_partner_matrix == results2.good_partner_matrix
    assert results1.good_partner_rating == results2.good_partner_rating
    assert np.allclose(results1.eigenmoses_rating, results2.eigenmoses_rating)
    assert np.allclose(results1.eigenjesus_rating, results2.eigenjesus_rating)
```

Note I'm using `np.isclose` for some properties.

In this commit:

- I add the specific seed for which the error was found as a hypothesis
  example (`seed=2`).
- Replace the `results1 == results2` check with just a check of some
  properties (from which the others are essentially calculated).
- Added `progress_bar=False`

* Add instructions for using venv.

* s/requirements/requirements.txt

* Spell requirements correctly..

axelrod/tests/unit/test_tournament.py

@@ -750,13 +750,19 @@ def test_write_to_csv_without_results(self):
         self.assertTrue(df.equals(expected_df))
 
     @given(seed=integers(min_value=1, max_value=4294967295))
+    @example(seed=2)
     @settings(max_examples=5, deadline=None)
     def test_seeding_equality(self, seed):
         """Tests that a tournament with a given seed will return the
         same results each time. This specifically checks when running using
         multiple cores so as to confirm that
         https://github.com/Axelrod-Python/Axelrod/issues/1277
-        is fixed."""
+        is fixed.
+
+        Note that the final asserts test only specific properties of the results
+        sets and not the entire result sets as some floating point errors can
+        emerge.
+        """
         rng = axl.RandomGenerator(seed=seed)
         players = [axl.Random(rng.random()) for _ in range(8)]
         tournament1 = axl.Tournament(
@@ -776,9 +782,12 @@ def test_seeding_equality(self, seed):
             seed=seed,
         )
         for _ in range(4):
-            results1 = tournament1.play(processes=2)
-            results2 = tournament2.play(processes=2)
-            self.assertEqual(results1.ranked_names, results2.ranked_names)
+            results1 = tournament1.play(processes=2, progress_bar=False)
+            results2 = tournament2.play(processes=2, progress_bar=False)
+            self.assertEqual(results1.wins, results2.wins)
+            self.assertEqual(results1.match_lengths, results2.match_lengths)
+            self.assertEqual(results1.scores, results2.scores)
+            self.assertEqual(results1.cooperation, results2.cooperation)
 
     def test_seeding_inequality(self):
         players = [axl.Random(0.4), axl.Random(0.6)]

docs/tutorials/contributing/guidelines.rst

@@ -23,7 +23,7 @@ The project follows the following guidelines:
    from `Chris Beams <https://chris.beams.io/posts/git-commit/>`_
 5. Testing: the project uses the `unittest
    <https://docs.python.org/2/library/unittest.html>`_ library and has a nice
-   testing suite that makes some things very easy to write tests for. Please try
+   testing suite that makes some things easy to write tests for. Please try
    to increase the test coverage on pull requests.
 6. Merging pull-requests: We require two of the (currently three) core-team
    maintainers to merge. Opening a PR for early

docs/tutorials/contributing/index.rst

@@ -10,6 +10,7 @@ Contents:
    :maxdepth: 2
 
    guidelines.rst
+   setting_up_the_environment.rst
    strategy/index.rst
    library/index.rst
    running_tests.rst

docs/tutorials/contributing/running_tests.rst

@@ -4,10 +4,6 @@ Running tests
 Basic test runners
 ------------------
 
-Before running tests, you should have hypothesis 3.2 installed::
-
-    $ pip install hypothesis==3.2
-
 The project has an extensive test suite which is run each time a new
 contribution is made to the repository.  If you want to check that all the tests
 pass before you submit a pull request you can run the tests yourself::
@@ -73,12 +69,3 @@ You can also run the type checker on a given file. For example, to run the type
 checker on the Grudger strategy::
 
     $ mypy --ignore-missing-imports --follow-imports skip axelrod/strategies/grudger.py
-
-
-Continuous integration
-======================
-
-This project is being taken care of by `travis-ci
-<https://travis-ci.org/>`_, so all tests will be run automatically when opening
-a pull request.  You can see the latest build status `here
-<https://travis-ci.org/Axelrod-Python/Axelrod>`_.

docs/tutorials/contributing/setting_up_the_environment.rst

@@ -0,0 +1,38 @@
+Setting up the environment
+==========================
+
+Installing all dependencies
+---------------------------
+
+All dependencies can be installed by running::
+
+  $ pip install -r requirements.txt
+
+It is recommended to do this using a virtual environment tool of your choice.
+
+For example, when using the virtual environment library :code:`venv`::
+
+  $ python -m venv axelrod_development
+  $ source axelrod_development/bin/activate
+  $ pip install -r requirements.txt
+
+The git workflow
+----------------
+
+There are two important branches in this repository:
+
+- :code:`dev`: The most up to date branch with no failing tests.
+  This is the default branch on github.
+- :code:`release`: The latest release.
+
+When working on a new contribution branch from the latest :code:`dev` branch and
+open a Pull Request on github from your branch to the :code:`dev` branch.
+
+The procedure for a new release (this is carried out by one of core maintainers):
+
+1. Create a Pull Request from :code:`dev` to :code:`release` which should
+   include an update to :code:`axelrod/version.py` and :code:`CHANGES.md`
+2. Create a git tag.
+3. Push to github.
+4. Create a release on github.
+5. Push to PyPi: :code:`python setup.py sdist bdist_wheel upload`