Reflections on Ryze and Winrate Data
On Boards, you'll see a lot of discussions about champion winrate data, and how much it really means. I myself used to put quite a bit of stock in winrate data until I saw multiple examples of champions whose win rates clearly didn't sync up with their strength when played at a high level. For instance, new Kassadin's winrate has remained consistently low (43%) despite his proving to be a pretty good pick at the pro level. I still think that winrate data is valuable, especially when combined with considerations such as how it varies with player level, and what sort of champion kits tend to lend themselves to high winrates. However, I think that new Ryze offers a particularly stark example of the level of care with which winrate data really needs to be treated. So, a brief history:
Patch 5.8 (4/30): Ryze rework is released. Ryze's winrate remains consistently low (~41% overall, rising (heh) to perhaps 43% in Platinum/Diamond).
Patch 5.9 (5/19): Ryze gets a bugfix, but also has a potentially powerful interaction with Muramana removed. His winrate continues to be low until...
5/28: A small, mysterious bump in Ryze's popularity appears. It is imperceptible when looking at Bronze/Silver data but grows increasingly large with player level. For a time, Ryze enjoys very high winrates from Platinum on up (perhaps 53-54% overall).
Patch 5.10 (6/4): Buffs to Ryze's passive, Q, and R come through, officially breaking Ryze in half (as anyone who's watched LCS knows). Ryze experiences a massive popularity surge (noticeable at all levels but largest in Diamond, smaller in Master+). At the same time, Ryze's winrate at all levels drops to below 49%.
(These numbers are drawn from lolking.net; see here .)
Lessons to be learned from this:
It is fairly well-understood that changes in winrates are frequently inversely correlated with changes in popularity, as a rise tends to mean that more inexperienced players are picking up a champion, and vice-versa. However, small rises in popularity can also mean that experienced players are slowly catching on that a champion is sleeper OP, and it's particularly worth notice when popularity and winrate start trending in the same direction.
It's possible for a champion's winrate to be lower than their actual power (given competent but not necessarily perfect play) for a surprisingly long period of time. There was an entire month where Ryze's winrate barely hovered above the 40% line that normally indicates a garbage champion. However, the period from 5/28-6/4 indicates that Ryze was actually at least fine, if not pretty good, before 5.10. Presumably, it just took a lot of people that long to learn the best build, how to play around his passive and chain his abilities, etc. It may also be that the Muramana build was actually just a distraction from the actual best way to play Ryze.
A few closing thoughts on using winrates: I think the data is more meaningful when a champion has gone untouched for at least a few months, so that their strengths and weaknesses are well-understood. Also, if a champion's numbers are tuned in a way that doesn't really change their playstyle, the winrate numbers are probably more reliable. In general, though, all these numbers need to be interpreted in context and with caution. And honestly, the folks at Riot, who have access to a treasure trove of data that we Boards lurkers don't, are in a much better position to make these judgment calls than we are.
Even though they kinda screwed the pooch with Ryze =P
