Due to the ever-present 'debate' about riot buffing a champion at the same time as they release a skin for them I decided to gather a bit of data on this topic to see if stats back this theory up.
DISCLAIMER: I don't claim that this data is 100% accurate, the buff/nerf-classifier only achieved a testing accuracy of ~60%. This was my first project like this since I'm pretty new to programming and Data Science therefore my method was probably suboptimal. But if you have more experience and some advice for me, feel free to comment.
First I classified each change to a champion since Patch 4.8 as either a buff or a nerf using Machine Learning(60% Accuracy):
Total changes considered: 1286
After that I filtered out the changes to a champion, that got a new skin or chroma in the same patch, the patch before or the following patch. Then I looked at how much percent these changes make up of their respective category (buff or nerf):
As you can see their is no notable difference between the two categories, champions that got nerfed actually were more likely to get a skin at the same time. So there seems to be no correlation between buffs and skin releases for a champion.
As a source I used the official patch notes. I started after Patch 4.7 because Riot then began announcing the upcoming skins in their patch notes.
I scraped the changes from the website and created a text file with the summary and text of each change and named it after the campion and patch (eg. Patch 4.8-Blitzcrank). I also created a text file for all the skins in each patch, the patch before and the following patch.
To know if each change was either a buff or nerf I decided to try out Machine Learning, more exact Natural Language Processing and Text Classification.
Basically, the program looks at each change, splits the text into words and then decides based on these words if the change is a buff or nerf.
To train the algorithm, which means to teach the program which words belong to which category, I had to label 240 changes manually as either a buff or a nerf. The program then looks at the words and learns which words indicate a nerf and which words indicate a buff (For example the word 'decreased.' is 13x more likely to come up in a nerf than in a buff in my training data).
When testing the algorithm it reached a not-so-great (but acceptable for my expectations) accuracy of ~60%.
I compiled all of this information into a table which contains each change (e.g. Champion: LeBlanc, Patch: Patch 4.10, Buff/Nerf: nerf, Skin: No) and now only had to visualize the data (which proved to be harder than I thought).
Possible errors in the data: A small number of patches were not included due to a different formatting of the patch notes which messed with my program. The algorithm only reached an accuracy of about ~60% which could definitely be improved. And because of my programming inexperience and typical spaghetti code there are probably more bugs that I didnt't even notice.
As I said in the beginning if you have any questions or want to give me advice, feel free to comment.
Source: Original link
© Post "History of buffs and nerfs and their relation to skin releases [OC]" for game League of Legends.
Top-10 Best Video Games of 2018 So Far
2018 has been a stellar year for video game fans, and there's still more to come. The list for the Best Games of So Far!
Top-10 Most Anticipated Video Games of 2019
With 2018 bringing such incredible titles to gaming, it's no wonder everyone's already looking forward to 2019's offerings. All the best new games slated for a 2019 release, fans all over the world want to dive into these anticipated games!