ModalTFT, ModalDefector strategies - Ben McGuffog

[benjjo]

Hi, I've added two strategies that are interesting (well at least I think). Here are their descritpions:

ModalTFT
A player starts by cooperating and then analyses the history of the opponent. If the opponent Cooperated in the
last round, they are returned with a Cooperation. If the opponent chose to Defect in the previous round,
then this strategy will return with the mode of the previous opponent responses.

ModalDefector
A player starts by Defecting and then analyses the history of the opponent. If the opponent Cooperated in the
last round, they are returned with a Defection. If the opponent chose to Defect in the previous round,
then this strategy will return with the mode of the previous opponent responses.

Interestingly, when pitted against a few of the classics from the orignal Axelrod tournement, they do quite well. When tested agianst all of the strategies in this repository, they rank 103 and 200 respectively.

I have never written code for other people in my life. This would be the first. My tests, while simple, seem to pass though.
This is also my first pull request ever so I'm probably doing a number on it.
I have tried searching through your code to see if these strategies already exist and I haven't found any like them. It may be that I've simply missread the code and im wasting your time right now, for which I appologise in advance.
It would be cool to have the contributors badge on my account though.

[marcharper]

Hi, there's one doc test failing:

----------------------------------------------------------------------
File "/home/runner/work/Axelrod/Axelrod/./docs/index.rst", line 55, in index.rst
Failed example:
    len(axl.strategies)
Expected:
    240
Got:
    242


----------------------------------------------------------------------
Ran 65 tests in 76.362s

FAILED (failures=1)

It's because you added two new strategies.

Also, are you sure there's no similar strategies in the library? Many are TFT with a tweak like what you've done.

[benjjo]

Thanks Marc, I'll go through them all again. I'm a little more familiar with the whole package now.

…

On Sat, 2 Mar 2024, 10:07 Marc, ***@***.***> wrote: Hi, there's one doc test failing: ---------------------------------------------------------------------- File "/home/runner/work/Axelrod/Axelrod/./docs/index.rst", line 55, in index.rst Failed example: len(axl.strategies) Expected: 240 Got: 242 ---------------------------------------------------------------------- Ran 65 tests in 76.362s FAILED (failures=1) It's because you added two new strategies. Also, are you sure there's no similar strategy in the library? Many are TFT with a tweak like what you've done. — Reply to this email directly, view it on GitHub <#1433 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ABJUH5LT5R4ROWBJXMQ76F3YWEC23AVCNFSM6AAAAABC5KWMOGVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTSNZUGA2TEMJTGY> . You are receiving this because you authored the thread.Message ID: ***@***.***>

[marcharper]

Left some feedback, should you find that these are new strategies.

[marcharper]

We'll want to keep this as is for others.

[marcharper]

Rather than use the statistics library, the history class already keeps a running count of C and D, so you can do something like if opponent.history.cooperations > opponent.history.defections return C else return D. That saves a library import and is more efficient (mode is going to be O(n^2) over the entire game, this way is linear).

[marcharper]

Same comment re: statistics.mode

[marcharper]

Try to add a few more tests that single out the specific behaviors like the first move and that the mode works for both C and D. If possible try to use a real strategy like Alternator rather than mock players, but mocks are ok if it's hard to target a particular case.

[Nikoleta-v3]

@benjjo thank you very much for your contribution! I agree with marcharper that you should add a few more tests for the strategies.

I am assuming you designed these strategies? If that's the case, could you add a line in the docstring like this:

     Names: 
  
     - Modal TFT: Original name by benjjo or Ben McGuffog

Here's an example:

Axelrod/axelrod/strategies/qlearner.py

Lines 12 to 22 in fa74862

	class RiskyQLearner(Player):
	"""A player who learns the best strategies through the q-learning
	algorithm.
	This Q learner is quick to come to conclusions and doesn't care about the
	future.
	Names:
	- Risky Q Learner: Original name by Geraint Palmer
	"""

[Nikoleta-v3]

I believe this comment/line is from our contributing guidelines. You can remove it.

[Nikoleta-v3]

Same here.