Someone please explain this:

Games Other Halo
1

So I started a new account last night to see how I would rank starting over. Check this out: https://www.halowaypoint.com/en-us/games/halo-5-guardians/xbox-one/service-records/arena/players/dixsyn%20cider. (I can’t lazy link, but the profile name is Dixsyn Cider). Can someone please explain to me how I ranked Diamond 6 on my main account with more losses and lower K/D, while the new account ranks Platinum 5? The only thing I can think of is the fact that I didn’t play against enough higher ranked players, but that would mean the system is completely luck based. A true working rank system would put you against higher ranked players when it realized how much better you were doing than the other lower ranks, but it never did that. Maybe it’s because the low population won’t allow the system to work correctly and instead of finding a “close” match, it just finds you any. In H3, I remember the search would go through 3 steps where the first step looked for the closest possible match, the second widened it to about 5 levels higher or lower, and the last search would be for any. With the high population, you would usually find a match in the first or second step of the search and you would always be within a few levels of the players you were going against. I just don’t understand why this isn’t the way it’s being done here. Maybe I’m wrong and if so someone please explain to me why I ranked where I did.

2

How a developer decides to implement Xbox One’s Live Smart Match system into their game’s matchmaking has pretty much everything to do with who we’re getting matched up with and how long it takes to find matches.

  1. Xbox Live assigns a player with a rudimentary skill-metric when playing a new game based on thier profile’s (gamertag) performance history within all the online matchmaking games they’ve ever played in. If a player creates a new account it’ll start them with a rudimentary skill-metric just below the game’s average mean.

  2. This rudimentary metric has a wide standard deviation to allow for quick skill-based adjustments under the new game, but the more time a player spends with the new game under their gamertag the more confident (aka narrow) that standard deviation becomes.

  3. The skill-based matchmaking on the Xbox One is also balanced against factors such as a player’s Reputation, Age, and Language in order to provide an optimal experience for a player beyond what a simple skill metric may alone provide (aka the new Smart Match system that’s built off the back of Microsoft’s TrueSkill system). How effective this aspect is working within the matchmaking system across Xbox Live I couldn’t say as there’s little information put out on it.

  4. Team-based gameplay requires the system’s algorithm to estimate an average skill-metric beta for the teams based on the skill-chain (worst to best player); optimizing a match-up around these averages is how the system attempts to provide a “probable” competitive match. Unfortunately, what appears to be a competitive match “on paper” won’t always be competitively balanced because the sum of individual skill-metrics don’t account for the special dynamics that can occur within a team-based game that often make the sum greater or lesser than the parts; plus, people tend to have outlier performances here or there which can significantly alter outcomes from expectations.

  5. Expected outcomes cause smaller updates to a player’s skill-based metric because the algorithm already had a good guess of that player’s skill; unexpected outcomes (upsets) will cause larger updates to make the algorithm more likely to predict the outcome in the future.

  6. A player’s personal performance influences a match’s outcome and a match’s outcomes affects their skill-metric in reference to those they’re playing with and against; these metrics are continuously growing their metric’s level of confidence (aka the standard deviation). In earning a visual playlist CSR in Halo 5 the system will analyze an individual’s finishing placement within the match’s skill-chain (where an individual is expected to finish amongst the group) for the first 10 matches. Once a CSR is earned it will only get adjusted based solely on the team’s result (if playing a team-based gamemode) verses the player’s individual result.

  7. 343i have setup the playlists CSRs to have particular metric ranges for the divisions and sub-ranking tiers; some appear to be more or less restrictive. A correction came to the SWAT and FFA playlists, but I’d argue that both still need further refinement (just my opinion).

  8. Seasonal resets to specific playlist CSRs will allow individuals to be re-measured but with more confident data thanks to most players having more background experience with the game; the new 10 matches used to earn a playlist CSR still has all of a player’s existing game experience (which constantly grows their skill metric’s confidence level) to draw from in pairing players and teams.

  9. 343i have also setup specific parameters for the search process in order to allow the system to try pairing the most ideal matches immediately, but over time that requirement gets loosened in order to match players within reasonable times; nevertheless, I’ve heard many complaints from players overseas that don’t think the system loosens them up enough because they’re stuck searching for extreme lengths of time. The parameters revolve not only around SBMM (skill-based matchmaking) but also connection quality and I believe it’s that aspect that 343i are holding a tighter requirement to regardless to search times which delays matchmaking for many overseas; all-in-all, 343i don’t want people to have poor experiences per the connection, but denying them the experience all-together is just another form of a bad experience.