This PR adds the Momentum strategy

[dongwonmoon]

Summary: This pull request introduces a new strategy, Momentum, inspired by the concept of Gradual and the Momentum optimizer used in deep learning. The strategy models the dynamics of trust evolution in a repeated game setting, with momentum reflecting how trust shifts based on previous interactions. The momentum formula used in this strategy is derived from the Momentum optimizer in deep learning, which adjusts the velocity of a model's parameters based on gradients and previous updates. Similarly, the Momentum strategy adapts based on the opponent's previous actions.

Key Features:

Momentum Formula: The strategy uses an alpha parameter to weight the influence of the previous momentum value and the current opponent's action (cooperate or defect). This formula is inspired by the Momentum optimizer in deep learning, which accelerates convergence by adjusting updates with a combination of previous gradients and the current gradient.
Threshold: A threshold value determines the decision to cooperate or defect. If the momentum is above the threshold, the strategy cooperates; otherwise, it defects.
Initial Momentum: When the history is empty (first round), the momentum is set to 1.0, and cooperation is guaranteed.
Implementation Details:

The strategy uses an alpha (momentum decay factor) and a threshold to determine the behavior.
The momentum is updated after each round based on the opponent's previous action, using the same principle as the Momentum optimizer.
The strategy's behavior evolves over time, with a more cooperative stance when momentum is high and a defection tendency when momentum is low.

[drvinceknight]

Thanks for this @dongwonmoon.

The CI is failing with:

ImportError: cannot import name 'DataFrame' from 'dask.dataframe.core' 

I have been able to recreate this locally by rebuilding the tox environments:

$ python -m tox -r

The error is not caused by anything you're adding, it's something to do with an upstream update to task (I think!).

I have fixed the error by removing this line in test_resultset.py:

from dask.dataframe.core import DataFrame

I also noted when running things locally that you're going to need to run black on the two new files. If you need a hand with any of this let me know.

[dongwonmoon]

Thank you for the help with resolving the previous CI issue. However, a new issue has emerged.
It seems to be related to invalid escape sequences in the strings.

How sould i go about fixing this? @drvinceknight

[Nikoleta-v3]

Hey 👋🏻 @dongwonmoon,

This isn’t exactly where the tests are failing. These are just warnings that invoke a warning message but don’t raise an error.

The tests are actually failing because of some other code that was recently added to the codebase. More specifically:

axelrod/strategies/frequency_analyzer.py:65: error: Need type annotation for "frequency_table" (hint: "frequency_table: dict[<type>, <type>] = ...")  [var-annotated]
Found 1 error in 1 file (checked 1 source file)

In another strategy that was introduced recently, one of the attributes’ types is not specified, and this is causing the MyPy tests to fail. Could you tweak line 65 in strategies/frequency_analyzer.py (see: link) to:

self.frequency_table: dict = dict()

The tests should pass after that 😅 (I’m about 90% sure). I assume it’ll be okay with the rest of the team if you include this fix in this pull request.

[dongwonmoon]

Thank you for the detailed explanation. @Nikoleta-v3

This is my first PR, so I wanted to check—am I allowed to fix the error myself, or should I leave it for the maintainers to handle?

[drvinceknight]

This is my first PR, so I wanted to check—am I allowed to fix the error myself, or should I leave it for the maintainers to handle?

I can do it if you'd like me to do it BUT usually what happens is you would make the change on the same branch you are already have (just adding a new commit). Then when you push it here it will just update.

Let me know if I can help with anything.

[dongwonmoon]

Tests have been completed. Please let me know if there's anything else i should do.

[drvinceknight]

This looks good to me, I've made some minor stylistic suggestions as well as a request for a bit more documentation.

[dongwonmoon]

I tried running black and isort locally, but nothing changed.
Can you give me some hint about this problem?

[drvinceknight]

Can you give me some hint about this problem?

This could be an upstream update of black:

  ===== 5190 passed, 6 skipped, 1 xfailed, 78 warnings in 1163.11s (0:19:23) =====
  py311: commands[1]> python -m black -l 80 . --check
  would reformat /home/runner/work/Axelrod/Axelrod/axelrod/classifier.py
  would reformat /home/runner/work/Axelrod/Axelrod/axelrod/makes_use_of.py
  would reformat /home/runner/work/Axelrod/Axelrod/axelrod/strategies/memorytwo.py
  would reformat /home/runner/work/Axelrod/Axelrod/axelrod/tests/strategies/test_gambler.py

The CI installed black==25.1.0 could you check if that's different to the version you're running?

[dongwonmoon]

Thank you. Version was different.
It passed the test!

[Nikoleta-v3]

Include {self.momentum} in the name as well for completeness.

[Nikoleta-v3]

Thank you, @dongwonmoon, again for your PR. One small comment: since the player depends on alpha, threshold, and momentum, all three values can be included in the name.

Everything else looks great!

[Nikoleta-v3]

Your def test_repr(self): is failing because the name was changed 😄

[dongwonmoon]

Thank you. I applied that.
Please check and let me know if there is anything else I need to do