Axelrod-Python/Axelrod

A repository used to reproduce Axelrod's tournament

Owen Campbell

@meatballs

I now can't even see the axelrod project if I log into Travis. But that link you gave works just fine. Very odd

Oops. Sorry @marcharper. I didn't mean to click the button to re-request a review from you!

Vince Knight

@drvinceknight

I now can't even see the axelrod project if I log into Travis. But that link you gave works just fine. Very odd

Weird...

We'll move to GitHub actions and it'll be nice and self contained... (famous last words...)

Marc

@marcharper

https://towardsdatascience.com/prison-escape-solving-prisoners-dilemma-with-machine-learning-c194600b0b71?gi=4987d3aa2d06

There's a github repository and a notebook here: https://github.com/gsurma/prison_escape/blob/master/PrisonEscape.ipynb

Vince Knight

@drvinceknight

Cool.

Owen Campbell

@meatballs

b 58

T.J. Gaffney

@gaffney2010

I accidentally Axelrod-Python/Axelrod#1301 while continuous-integration/appveyor/pr was still running. (I thought it would block.) How can I run this now to make sure that it passes?

Marc

@marcharper

I restarted the build in appveyor

T.J. Gaffney

@gaffney2010

Just for curiosity, how do you do that?

Marc

@marcharper

You can log into Appveyor with your github account

calumrowan

@calumrowan

Hi everyone, I'm interested in running an Axelrod-style tournament for the Iterated Optional Pridoner's Dilemma. I understand that for this, I would need to create a new, 5 payoff game (R,P,S,T,L); can anybody point me in the right direction for starting this? I have read through the documentation, but I don't believe it covers creating a new game.

calumrowan

@calumrowan

Would it be a case of writing a new case (and changing the ColOutcomeType function) to include this L payoff, or would the alteration be more complex than that?

Vince Knight

@drvinceknight

I'm afraid as is the strategies would not be able to play the Optional PD. The optional PD has 3 strategies for players: (Cooperate, Defect or Abstain) so not only would you need to modify the game but you'd also need to modify all the strategies as they all are programmed to play Cooperate or Defect given a history of play.

Vince Knight

@drvinceknight

FYI, I'll be putting a new release together tomorrow :+1:

calumrowan

@calumrowan

Ah, that makes sense. Thanks Vince!

Marc

@marcharper

@calumrowan we are looking into supporting more games in a future release, so feel free to open an issue with some relevant details

Vince Knight

@drvinceknight

Axelrod-Python/Axelrod#1309 (the move to GitHub actions) is ready for review now :) :+1:

Marc

@marcharper

Hey, we just passed the 5 year birthday of Axelrod: https://github.com/Axelrod-Python/Axelrod/graphs/contributors

Vince Knight

@drvinceknight

Wow!

Happy birthday :cake:

Vince Knight

@drvinceknight

https://www.aaai.org/Papers/AAAI/2020GB/AAAI-WangS.1218.pdf is a paper that uses the library

They use your Moran process plots in the paper @marcharper :)

(I'll open a PR adding it to the citations.)

Vince Knight

@drvinceknight

Version 4.9.0 has been released: https://pypi.org/project/Axelrod/

Vince Knight

@drvinceknight

4.9.1 has been released (it's a bug fix of 4.9.0).

Nikoleta Glynatsi

@Nikoleta-v3

The tournament results with the latest version 4.9.1 are now available: https://github.com/Axelrod-Python/tournament

Vince Knight

@drvinceknight

Thanks @Nikoleta-v3 :+1:

Marc

@marcharper

Interesting that OmegaTFT does so well now in the standard and noisy tournaments

Vince Knight

@drvinceknight

Yeah...

A neat project would be to visualise how strategies have performed through the versions of the library... I'll get my git repo scraping script back out and to put that together at some point...

sindriing

@sindriing

Hey everybody I’m new here! 👋
I’m working on a project where I want to apply payoff penalties to FSM agents in the IPD depending on which states they visit.

I’ve read through the docs and the code and I really want to make use of the Axelrod library for all the obvious reasons: reproducibility, good testing, contribute to an established library etc. I’m trying to figure out the best way to make the modifications I need while still staying compatible with the core library.

Currently, the scores are calculated by the Match class using the Interaction_Utils functions: compute_final_score(result, game), etc. The only input to the functions are results and game. My thought is to add a penalties array to the History class which would be used to store possible penalties that might be applied to a Player’s score. The penalties could be an extra optional parameter i.e. compute_final_score(result, game, penalties)

Any ideas for a better way to get this done? Anything I really need to watch out for? I’m quite new to the whole contributing to an established library thing so I just wanted to drop a message here before diving in 😀

Marc

@marcharper

Welcome @sindriing :) I recommend that you make your changes on a local branch, see how they work, and then consider opening a pull request to the axelrod library.

Is this new research or are you building on some established literature?

Also I'm not sure what you mean by states -- are they states of the FSM, or states of the history, e.g. results of the rounds of play like CC, CD, DC, and DD ?

@gaffney2010 has also been thinking about how to modify games in general for the next major release of the library, so it might be better to define a new game type rather than modify the existing IPD game

sindriing

@sindriing

Is this new research or are you building on some established literature?
This is an MSc project that I wanted to do. The gist is to research the value of information in the IPD by applying a cost to attaining information about the last player's move. Thus players can benefit from moving blind i.e. not checking what their opponent did.

Also I'm not sure what you mean by states...
Short: states in the FSM.
My thought is that the simplest way to come up with lots of strategies compatible with my model is to make random FSMs with two distinct state types: Blind states which make the same move regardless of the opponents move and Non-Blind states which decide on their move depending on the opponents last move. Each time a player enters a Non-Blind state they would have to pay a penalty.

...better to define a new game type rather than modify the existing IPD game
I'll keep that in mind. Thanks!

Zedtho

@Zedtho

Hey everyone, new here! I'm trying to test one of Axelrod's claims for a biology project at school. I started off programming something in C++ but just when I was almost done I found this, and I want to applaud your work here it's amazing! Anyway, Axelrod says that given a society of All-D, Tit-for-tat can invade it if clustering is high enough (i.e. Tit-for-tat interacts enough w/other tit-for-tat players, so the covariance factor is high) I understand there is already a w factor that one can set to give tournaments probabilistic ends, but as of yet I haven't found this question be directly answered in the documentation. It would also be great if one could both do a singular tournament, but also a moran process with this. Sorry if this is a dumb question, sadly my programming skills are reserved to C++ and Python is rather new to me

Zedtho

@Zedtho

I should add - I could of course also still investigate something else for my biology project if you have any great ideas! I really liked the documentation on different research topics, however, most big tournaments are a bit problematic considering I'd have to introduce a lot of strategies which takes up the word count sadly.

Vince Knight

@drvinceknight

Hi @Zedtho apologies for the delay in getting back to you.

There's documentation on Moran processes here: https://axelrod.readthedocs.io/en/stable/tutorials/getting_started/moran.html

Specifically Moran processes on graphs here: https://axelrod.readthedocs.io/en/stable/tutorials/further_topics/moran_processes_on_graphs.html#moran-process-on-graphs

I hope that's enough to nudge you in the right direction.

Zedtho

@Zedtho

Thank you very much for the answer! The Moran process on graphs specifically is useful, I must have missed that, thanks!

RomeroLaura

@RomeroLaura

Hi all, I was thinking of contributing to the project and after browsing the desired list, I glanced upon spiteful_cc from Mathieu2015. I do not think it has already been done but wanted to check to make sure! If it is available, I was also wondering what an appropriate alternate name would be, since _cc does not seem to be axelrod's style, or if that is even necessary at all? Thank you for your help!

Marc

@marcharper

Hi, when you write the class it can be named something like SpitefulCC. There's a "name" field where "spiteful_cc" would go

I don't believe spiteful_cc is in the library yet; we do have a SpitefulTitForTat though so take a look at it to see if it's truly different: https://github.com/Axelrod-Python/Axelrod/blob/8793e5838778a4db434e5cea1875d71434bc10ff/axelrod/strategies/titfortat.py

Also there are some others in that reference we don't have as well, so feel free to choose another

RomeroLaura

@RomeroLaura

Okay, thank you for your super quick and insightful comments! I think thats what I needed to know :) I shall get to it pronto

Where communities thrive

People

Repo info

Activity