[Draft] 349 feat simulated annealing

[amarshah1999]

[Draft] Introduces the simulated annealing project as outlined here #349

ToDo List:

• Replace Tester file with actual docs and a sample notebook
• Add unit test
• Create a visualisation function to show the changes that the algorithm makes to player positions (and then undo my changes in visualize.py)

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 99.28%. Comparing base (e4ce2b3) to head (b53cc9f).
⚠️ Report is 89 commits behind head on develop.

Additional details and impacted files

@@             Coverage Diff             @@
##           develop     #380      +/-   ##
===========================================
+ Coverage    99.22%   99.28%   +0.05%     
===========================================
  Files           49       67      +18     
  Lines         3736     5441    +1705     
===========================================
+ Hits          3707     5402    +1695     
- Misses          29       39      +10     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[Alek050]

Hi @amarshah1999,

I really like the implementation, it is clean and structured. I was able to play with it and it worked really nicely. I think this is a great starting point and now we have to make it (1) more customisable and intuitive and (2) production ready. Let me give a few examples of point 1 I think would benefit:

• Add a random_state variable
• Add a verbosity to track progress
• fix overflow warnings (especially in match.exp calls)
• Add a patience so the early stopping is customizable as well
• More complicated but a deepcopy of a Game is probably the least efficient way to make the visualisation, so we have to think about if we can make this better.

Again, I really like it. It also confirmed what I thought would be the case: not the optimization algorithm but the creation of the right objective function is probably the bottleneck of these use cases, but implementing this will give the community the tools to work on that.

I also think that it would be great to be more elaborate on different situations, including/excluding players to change position, etc.. And ideally I would not want the skillcorner but some databally inherintly supported exmaple data (e.g. docs/saved_games/synced_game) to showcase since the integration of skillcorner, kloppy, and databallpy is not really the subject of this project.

Let me know what you think and if you want me to help you out anywhere with that!

[amarshah1999]

• Add a random_state variable
• Add a verbosity to track progress
• fix overflow warnings (especially in match.exp calls) - could you elaborate on this? I could not figure out how to repro this error. I found a runtime error which happens when calling tti on a stationary player, this might be the same thing you were seeing?
• Add a patience so the early stopping is customizable as well
• More complicated but a deepcopy of a Game is probably the least efficient way to make the visualisation, so we have to think about if we can make this better. agree! let me think on this some more
• Replace skillcorner data with default synced_game
• Filtering players to include - I added this to the docs, it was already available just hidden.

Will remove the warnings from the docs later on. Just have it like that because my local setup is overly verbose.

[Alek050]

@amarshah1999,

I mean this part, when new_score - best_score / T is greater than 709 or smaller than -709 it will result in a overflow warning/error. Just clip it with np.clip((new_score - best_score) / T, -700, 700) or a similar solution.