I’ve been thinking about one of the largest changes of the preseason: removal of summoner names. Scouting your team with op.gg or something similar was a great way to know when to use your dodge. With that gone, what now?
I propose looking at not just individual winrates on champions, but on how your team works together as a whole. There are a few websites that look into this, but as far as I know, not many are looking at the entire two team’s synergies.
What do I mean by synergies? I’m taking each pair of champions and looking at what their winrate is together. By doing this for each pair of champions on each team, I can have a better guess of how well each team works together.
TL;DR I made a small program to figure this out. Go here: Releases · amunchet/DodgeOrNot (github.com) and download the first one (since I don’t have a Windows developer key, your anti-virus should complain. More in the questions below). If you have any issues or suggestions (or kindly want to leave me a star), go to the Github and let me have it: https://github.com/amunchet/DodgeOrNot
Some more in-depth details follow.
Why are you doing this? I’m a one trick Viktor player. I think team comps are especially important, especially given my champion’s limitations (for example, Viktor needs a good CC engage to really be effective). Knowing how well the team fits around me is really important in determining whether or not the lobby has good odds of success.
Does this work? I’ve personally been finding that team differentials of greater than 2% (i.e. our team has a 52% win rate based on comp, theirs has a 49%) are significant. Within 1% or so becomes more of a coinflip. Now will a Challenger Teemo main absolutely dumpster me, no matter what? Of course. I’m just looking for an odds advantage over a large number of games.
How do I deal with tiers? The data source I’m using goes from bottom to top - I’m personally around Plat, so it’s equally useless to me how well Challengers use a champion as Iron. What I’m looking to understand is generally how well do the champions fit together across as wide a population as possible.
Can’t you see the summoner ids in client’s API? Yes, but I expect Riot to remove that before the season starts. Also, it somewhat goes against their philosophy to continue using that, so I’m avoiding it.
Isn’t this doing exactly what Riot doesn’t want you to do? This is a more philosophical question, but I believe that dodging because your team comp doesn’t look as strong as the enemy is a much better solution than dodging because that top Teemo is on a 20 game losing streak. The former is much less of personal targeting, so I think it’s more in line with Riot wanting to eliminate the “meta” game of scouting players, etc. in Solo Queue.
Is this program safe? While I don’t have a Windows developer key (yes, your anti-virus will complain about it), I’m using Github actions to build the executables directly. The source code is very basic Python, so you can read through it.
How much of this code was AI generated? I used ChatGPT for a fair bit and it did quite well. The logo was also generated by Midjourney. Very impressive. I managed to put this together way faster than I would have without the AI assists.
That’s about all I can think of - I’d love to hear your feedback either here or on Github. Is this super dumb? Should I just get gud and carry? Let me know!
- Start up the Riot Client and League of Legends
- Download the latest release of DodgeOrNot (say yes to all the scary anti-virus messages)
- Start up the DodgeOrNot
- Get into a ranked lobby and see whether or not you should dodge!
pytest --cov --cov-fail-under=95 tests/ --cov-report=term-missing
- 0.1: Pre-release, testing concepts