TW Matchmaking and "Sandbagging"
Hi all, what follows is my analysis of what's happening in the TW metagame and matchmaking system, as someone that spends a decent amount of time doing data analysis. For context: I'm currently in a guild where we've won 47 of our last 48 TWs from the 160 million up to the 200 million bracket as we've grown. Basically, we're a very TW focused guild that has lots of (mostly f2p/dolphin-type) people that are dedicated to keeping up with the pvp metagame.
Overall, based on the posts on this forum, I see an interesting ranking of guilds from the perspective of TW:
1. Guilds that have 50 active members that play for fun, don't always keep up with the meta, and don't have every top end team
2. Guilds that have 35-45 active members that play for fun, don't always keep up with the meta, and don't have every top end team
3. Guilds that have 35-45 active members that try to keep up with the meta and have pretty strong pvp rosters
4. Guilds that have 50 active members that try to keep up with the meta and have pretty strong pvp rosters
Note that 2 will usually beat 1, because if you're not following the meta, having bigger GP numbers per player probably means having more teams at higher gear levels, even though your guild is outnumbered, which gives you the advantage.
3 will usually beat 2 because they have more meta teams (and 1 because they're basically just strictly stronger 2s)
4 will usually beat 3. We love it when we fight the "sandbagger" guild that has 42 members and 8 alts, because even though it means they have higher GP per player than us, it means we have more meta teams. For example, they *can only* have 42 darth revans, while we have 49, and the same is true for every meta team down the line. That's a huge advantage in our favor. 4 also usually beats 1 and 2 because we have better meta teams *and* more of them. Our only tough matchups are when we get paired against other type-4 guilds, and we actually love that, because it makes for interesting gameplay.
Now here is where CG's matchmaking seems to go wrong: Their matchmaking seems to prevent matchups of 1 vs 4 and 2 vs 3 even given similar GP numbers, because they do some sort of meta-team-analysis. However, it seems to readily allow for 1 vs 2 and 3 vs 4 (For perspective, I'm in a type 4 guild and I'd say 90% of what we fight are type 3 guilds), and the forum posts on here complaining seem to often be type 1 guilds that get paired against type 2 guilds.
My theory is that without careful data analysis on CG's part, these pairs of mismatches could actually make their algorithm look fair! Let's say they are aware that some guilds complain when they get matched against guilds with different numbers of active members. They might go look at the data and say "no wait! The guild with fewer active people seems to win about 50% of the time so it's fine," but in reality, what they're seeing is that in the 1v2 matchups the less-people guild always wins and in the 3v4 matchups the less-people guild always loses, and those offset eachother.
Therefore, my proposal @CG is to try to pair guilds only against other guilds with +/- 2 active members. For the most part it should be possible (even if it makes the matchmaking "scores" behind the scenes look slightly more mismatched), and probably avoids some of these "TW Cheating/Sandbagging" mismatches that people complain about.
(Edited for typos)
Overall, based on the posts on this forum, I see an interesting ranking of guilds from the perspective of TW:
1. Guilds that have 50 active members that play for fun, don't always keep up with the meta, and don't have every top end team
2. Guilds that have 35-45 active members that play for fun, don't always keep up with the meta, and don't have every top end team
3. Guilds that have 35-45 active members that try to keep up with the meta and have pretty strong pvp rosters
4. Guilds that have 50 active members that try to keep up with the meta and have pretty strong pvp rosters
Note that 2 will usually beat 1, because if you're not following the meta, having bigger GP numbers per player probably means having more teams at higher gear levels, even though your guild is outnumbered, which gives you the advantage.
3 will usually beat 2 because they have more meta teams (and 1 because they're basically just strictly stronger 2s)
4 will usually beat 3. We love it when we fight the "sandbagger" guild that has 42 members and 8 alts, because even though it means they have higher GP per player than us, it means we have more meta teams. For example, they *can only* have 42 darth revans, while we have 49, and the same is true for every meta team down the line. That's a huge advantage in our favor. 4 also usually beats 1 and 2 because we have better meta teams *and* more of them. Our only tough matchups are when we get paired against other type-4 guilds, and we actually love that, because it makes for interesting gameplay.
Now here is where CG's matchmaking seems to go wrong: Their matchmaking seems to prevent matchups of 1 vs 4 and 2 vs 3 even given similar GP numbers, because they do some sort of meta-team-analysis. However, it seems to readily allow for 1 vs 2 and 3 vs 4 (For perspective, I'm in a type 4 guild and I'd say 90% of what we fight are type 3 guilds), and the forum posts on here complaining seem to often be type 1 guilds that get paired against type 2 guilds.
My theory is that without careful data analysis on CG's part, these pairs of mismatches could actually make their algorithm look fair! Let's say they are aware that some guilds complain when they get matched against guilds with different numbers of active members. They might go look at the data and say "no wait! The guild with fewer active people seems to win about 50% of the time so it's fine," but in reality, what they're seeing is that in the 1v2 matchups the less-people guild always wins and in the 3v4 matchups the less-people guild always loses, and those offset eachother.
Therefore, my proposal @CG is to try to pair guilds only against other guilds with +/- 2 active members. For the most part it should be possible (even if it makes the matchmaking "scores" behind the scenes look slightly more mismatched), and probably avoids some of these "TW Cheating/Sandbagging" mismatches that people complain about.
(Edited for typos)
