- Joined
- Jun 8, 2023
- Messages
- 29
Hey Everyone!
I know I've talked a lot about World War Matchmaking here on the forums, but that has always been in answer to your questions. Today I thought it would be fun to share what I've been working on with @RAH-66 from our Design team and with our Analytics team in order to improve matchmaking before Tournament of Glory releases!
You may have noticed recently that you haven't seen as many bad matches or had to wait as long in the worst cases as you used to. This is the result of us increasing the granularity of our matchmaking times, value spreads (collectively called "windows"), and the amount of time you sit in matchmaking before you get a forced match. These changes went out on September 30th, and we are already seeing some interesting results!
The most noticeable changes are the alliances on the extreme ends of the spectrum in power, skill, and variance. Here we have seen a ~33% reduction in the ceiling of our matchmaking system's inefficiency across the 10v10, 15v15, and 20v20 queues. This means that the alliances who used to have to wait the longest now are waiting only 66% of the time that they used to (4 hours less)! The median alliance in these queues won't notice too much of a difference, but here we still see about a ~5% drop in matchmaking inefficiency, which equates to 36 minutes less time spent in matchmaking!
Alongside the faster matchmaking times in those queues, we've also noticed an increase in our Match Quality past 1 standard deviation, which ultimately just means the matches are about ~5% better for you guys!
Now, the 30v30 queue is a bit of a different story. Some of you noticed how that queue has always felt a bit different in both the matchmaking times and the quality of matches. Admittedly, this queue is harder to work with because of the smaller number of alliances that interact with it, but the alliances that DO interact with it are very dedicated and share their passion for a truly competitive experience.
Since we have less data to work with, we can't analyze 30v30s in the same way that we analyze the other groups. What we have done here instead is to increase matchmaking times a bit in order to give the system more breathing room to find the best possible match for you, and not force match alliances so much. As a result of that, we've noticed the matchmaking system focusing a bit more on alliance power levels in an attempt to find good matches. We've also seen a massive reduction in forced wars that effectively have a random match quality associated with them. We are still analyzing this, but for now, this solution seems solid!
Overall, we are quite happy with these first steps and plan on letting things run as-is while we finish up Tournament of Glory. When we release this glorious feature to you all, we are going to have to do another pass at the matchmaking windows due to how we are playing with some of the numbers around glory itself.
Happy Warring!
-Your friendly neighborhood server dev, Daedalus
I know I've talked a lot about World War Matchmaking here on the forums, but that has always been in answer to your questions. Today I thought it would be fun to share what I've been working on with @RAH-66 from our Design team and with our Analytics team in order to improve matchmaking before Tournament of Glory releases!
You may have noticed recently that you haven't seen as many bad matches or had to wait as long in the worst cases as you used to. This is the result of us increasing the granularity of our matchmaking times, value spreads (collectively called "windows"), and the amount of time you sit in matchmaking before you get a forced match. These changes went out on September 30th, and we are already seeing some interesting results!
The most noticeable changes are the alliances on the extreme ends of the spectrum in power, skill, and variance. Here we have seen a ~33% reduction in the ceiling of our matchmaking system's inefficiency across the 10v10, 15v15, and 20v20 queues. This means that the alliances who used to have to wait the longest now are waiting only 66% of the time that they used to (4 hours less)! The median alliance in these queues won't notice too much of a difference, but here we still see about a ~5% drop in matchmaking inefficiency, which equates to 36 minutes less time spent in matchmaking!
Alongside the faster matchmaking times in those queues, we've also noticed an increase in our Match Quality past 1 standard deviation, which ultimately just means the matches are about ~5% better for you guys!
Now, the 30v30 queue is a bit of a different story. Some of you noticed how that queue has always felt a bit different in both the matchmaking times and the quality of matches. Admittedly, this queue is harder to work with because of the smaller number of alliances that interact with it, but the alliances that DO interact with it are very dedicated and share their passion for a truly competitive experience.
Since we have less data to work with, we can't analyze 30v30s in the same way that we analyze the other groups. What we have done here instead is to increase matchmaking times a bit in order to give the system more breathing room to find the best possible match for you, and not force match alliances so much. As a result of that, we've noticed the matchmaking system focusing a bit more on alliance power levels in an attempt to find good matches. We've also seen a massive reduction in forced wars that effectively have a random match quality associated with them. We are still analyzing this, but for now, this solution seems solid!
Overall, we are quite happy with these first steps and plan on letting things run as-is while we finish up Tournament of Glory. When we release this glorious feature to you all, we are going to have to do another pass at the matchmaking windows due to how we are playing with some of the numbers around glory itself.
Happy Warring!
-Your friendly neighborhood server dev, Daedalus
