You possess an invisible or hidden TrueSkill-Through-Time (TTT/TS2) metric for each individual player-vs-player (PvP) playlist mode referred to as your MMR (Match-Make Rating). The learning theory & formula behind TrueSkill v2 (TS2) has again helped give the gaming industry a great advancement toward understanding player skill to accurately predict match outcomes. It signals out particular individual performance based measurements which directly influence match time & outcome.
TS2 provides numerical indicators of player skill. A raw curve numeral & it’s conversion. The converted reflects the same numerals that the visual Competitive Skill Rank (CSR) uses.
MMR is how the Skill-Based Match-Making (SBMM) system & Team Picker are able to create balanced matches. To achieve optimal fairness it’s also highly dependent on being able to draw from a healthy player base population throughout the full skill spectrum.
It’s important for the MMing system to have restrictions to prevent wildly unbalanced matches that raise match, & post-match, quit rates while maintaining an eye on connection quality.
In H5 they’ve currently implemented a learning program (TrueMatch) to predict the best match options for the time of day and geographical location. Though it’s also been setup to ensure anyone can find a match, so it will allow lopsided matches to occur in order to support all searching players/parties.
Connection preference options: Focus, Balanced, & Expanded use to give the player some control over the available data centers. (ie. pop pool per potential ping quality). But, with TrueMatch they’ve been rendered superficial options.
The MM’r looks at skill parameters to help form teams. It’ll look at team-to-team (t2t) skill differences as well as individual-to-individual (i2i) skill differences; plus, any potential party boosts.
t2t is a priority factor for post-match adjustments while i2i is primarily helpful toward distributing the team’s skill adjustment credits individually post-match. More details below.
So the MM’r pulls a group of players together that it sees as being within the search’s acceptable parameters. The Team Picker strives for the smallest t2t gap it can generate.
Individuals can earn visible CSRs for Ranked playlists. To do so you must complete 10 placement matches. The system then assesses your CSR at some value below your current MMR. A ranked-in CSR has generally been between 100 or 200 points below a person’s MMR; however, there’s a rank-in cap that will cause higher skilled individuals to have a difference well beyond 100 or 200 points. The current rank-in difference is set at 100 with the CSR cap set at Diamond 3 (aka 1300) so anyone above 1400 will be set back further than 100 points.
Generally speaking, for those with a 1400 or below MMR after completing the 10 placement matches they should be able to converge their CSR to their MMR within 4 or 5 competitively balanced winning matches. CSR tends to float +/- 150 points (as the approximate max) around a player’s MMR & in-turn the potential luck of winning or losing streaks related to opponent difficulty.
Your visible CSR should provide a fairly decent estimate of your current MMR, but it’s unlikely to be a perfectly precise representation of your MMR at any given time. In ways, it’s actually a better overall gauge of your skill over time since it’s not as susceptible to dramatic oscillations & it’ll reflect party skill-boost influences too.
CSR not only helps reign in rank progress & gives a more controlled state to the visual reflection of our TS it also fulfills additional roles that MMR can’t provide.
CSRs can only adjust based on match outcomes which ensures that wins are always awarded & losses are always punished. Not to mention, CSRs, paired with seasonal resets, essentially act as a form of incentive to re-invest time into playlists in order to re-earn a visual indicator of your skill.
Re-earning your skill is important because it ensures players continue to prove to the system that they deserve their skill ratings (both CSR & MMR). Time away (not playing) does have a deteriorating effect on player skill.
CSR is designed to limit any potential inflation/deflation from a player’s MMR by self-correcting. This provides it with integrity & meaningfulness.
The CSR adjustments are based on a couple things: (1) match outcome via Win/Loss, & (2) a formula (see Joshua Menke on Twitter)*.
Members on a winning team will always gain CSR. Up to +30 for those under Onyx; otherwise, up to +10 for those within Onyx.
Members on a losing team will always lose CSR. Down to -30 for those under Onyx; otherwise, down to -10 for those within Onyx.
Comparing Pre-Match CSR to Pre-Match MMR means that an individual’s in-game performance won’t have an exact effect on their CSR adjustment. It’ll always lag by at least one match. In other words, an aspect of the Post-Match CSR adjustment will be reflective of a potential MMR skill change that occurred from your previous match or previous matches. This can lead to some confusion as players are unable to more directly relate their most recent individual performance to the adjusted amount.
The +/- 1 CSR number in particular, instead of a larger number, helps to significantly restrict CSR inflation/deflation relative to a player’s MMR because it requires sustaining a consistent 94% or greater win/loss ratio to maintain runaway inflation/deflation.
The TTT/TS2 system, which relates to MMR, is designed to ensure match outcome via wins & losses remains the crux of the skill system, but it also appropriately accounts for individual performance influence toward those outcomes.
TS2 is significantly more accurate at predicting match outcomes than TS1 (~70% vs ~58%) which inherently means it’s more capable of measuring & understanding a player’s individual skill.
The most influential individual component to winning has been determined to be earning kills; therefore, Kills-per-Minute (KPM), and to a much lesser degree Deaths-per-Minute (DPM), have a direct relation to your MMR & in-turn your visible CSR.
A winning team will always net a positive MMR. The amount is a combination of factors related toward match expectations. This doesn’t mean an individual will always receive a positive MMR adjustment from a win; though, their odds become considerably increased.
A losing team will always net a negative MMR. The amount is a combination of factors related toward match expectations. This doesn’t mean an individual will always receive a negative MMR adjustment from a loss; though, their odds become considerably increased.
Individual MMR adjustments (within a team) operate on a zero sum of the team’s net MMR adjustment. These individual adjustments remain relative to KPM & DPM expected ranges; including how those performances (relative to expectations) stack up between one-another (ie team members) because it’s a zero sum adjustment.
KPM & DPM expectations possess various weighted ranges based on the mode & historic certainty. They’re constantly self-adjusting.
As mentioned previously it’s possible for an individual’s MMR to independently increase or decrease in opposition to the team’s net adjustment, but the zero sum aspect means that their teammates (per their particular individual adjustments) are required to offset that individual’s MMR adjustment such that the team’s net adjustment is realized.
Severe outlier performances will get flagged by the system. Nefarious actions can lead to a ban.
Quitting has some direct influences too.
*not for FFA & multi-team