AverageRatingGain

Understanding Average Rating Gain

dsstats is using an ELO based rating algorithm.

The difference in the ratings between two teams serves as a predictor of the outcome of a match (expectation to win). A player's Elo rating is represented by a number which may change depending on the outcome of rated games played. After every game, the winning players takes points from the losing ones. If the higher-rated team wins, then only a few rating points will be taken from the lower-rated team. However, if the lower-rated team scores an upset win, many rating points will be transferred.

Now, considering the Average Rating Gain of Commanders played by the players in these matches:

• This metric focuses on the performance of individual commanders chosen by players in 3 vs. 3 scenarios.

• The Average Rating Gain offers insights into how each commander influences a match's outcome compared to expectations.

• If a commander consistently leads their team to victories when they were not expected to win (e.g., against higher-rated opponents), the Average Rating Gain for that commander will be notably positive.

• Conversely, if a commander struggles to secure victories against opponents they were expected to defeat (e.g., lower-rated opponents), the Average Rating Gain for that commander may show minor rating losses.

In summary, the Elo-based rating system assesses player performance in 3 vs. 3 matches based on rating differentials and adjusts ratings accordingly. The Average Rating Gain of Commanders complements this by offering a commander-specific perspective on how their choices impact match outcomes, factoring in the expectations for each match.

Why it Trumps Winrate

Respects Player Skill:

Winrate alone doesn't consider the proficiency of players. Average Rating Gain factors in the relative skill levels of participants, offering a more balanced evaluation.

Handles Duplicates Gracefully:

In cases where multiple players choose the same commander in a single match, winrate can become skewed. Average Rating Gain accounts for this scenario more effectively.

Scenario:

• Team 1 has a player using the commander Dehaka.
• Team 2 also has a player using Dehaka.
• Winrate for Dehaka: 50% (because each team has a Dehaka).

Using Winrate:

• Winrate would indicate that Dehaka's performance is average, with a 50% win rate.
• However, it doesn't consider the specific context of the match or the relative skill levels of the players.

Using Average Rating Gain:

• Average Rating Gain takes into account the strength of each team and the individual players' skill levels.
• If Team 1 has higher-rated players using Dehaka, and the system expects Team 1 to win (let's say an expectation to win > 50%), the Average Rating Gain for Dehaka on Team 1 would be positive. This is because Dehaka contributed to an outcome that exceeded expectations.
• Conversely, if Team 2 has stronger players using Dehaka, and the system expects Team 2 to win (expectation to win < 50%), the Average Rating Gain for Dehaka on Team 2 would also be positive because Dehaka helped achieve an outcome that defied expectations.
• In both cases, Average Rating Gain provides a more accurate assessment of Dehaka's impact. It recognizes that Dehaka's contribution can lead to positive rating gains when expectations are exceeded and negative rating gains when expectations are not met.

This nuanced approach ensures that commanders are evaluated not just based on whether they win or lose but on how they influence the match's outcome relative to the expectations set by the skill levels of the players involved. It results in a more precise and informative metric for evaluating commander performance in multiplayer games.

Mitigates 3-Stack Influence:

Winrate can be disproportionately affected when highly skilled players "3-stack" and continually select certain commanders. Average Rating Gain is less susceptible to this manipulation.

Scenario:

• Two teams are playing a Commander (3 vs 3) match.
• On Team 1, there's a "3-stack" group of highly skilled players who consistently choose the commander Raynor.
• Team 2 comprises individually matched players of varying skill levels.

Using Winrate:

• Winrate would show that Raynor has a very high win rate because the "3-stack" group on Team 1 consistently wins when playing Raynor.
• However, this doesn't accurately represent Raynor's performance for individual players or in other contexts.

Using Average Rating Gain:

• Average Rating Gain considers the performance of each player individually, regardless of whether they are part of a "3-stack" group or not.
• If the highly skilled "3-stack" players on Team 1 are consistently winning with Raynor, their individual Average Rating Gain might be close to zero or even slightly negative because they are expected to win.
• On Team 2, the individually matched players using Raynor might have a wider range of Average Rating Gains, with some experiencing significant gains when they manage to defeat the "3-stack" group, as their performance exceeds expectations.
• The overall Average Rating Gain for Raynor would provide a more balanced assessment, reflecting the varying performance outcomes and mitigating the skewing effect of the "3-stack" group. In this scenario, Average Rating Gain recognizes that the commander's performance can differ based on the skill level and composition of the teams. It prevents the "3-stack" group's influence from artificially inflating the commander's winrate, allowing for a more accurate evaluation of Raynor's effectiveness across different match contexts and player combinations.

Ultimately, Average Rating Gain provides a fairer and more granular assessment of commanders, ensuring that their performance is not disproportionately influenced by certain player behaviors, such as "3-stacking."

Quantifying the Gain:

It provides quantifiable insights into rating adjustments. Winning against a strong opponent might yield only a small rating increase, while a victory against evenly matched opponents can result in a significant gain.

By computing the average rating gain for players using a specific commander, we gain a more nuanced understanding of that commander's effectiveness. This approach minimizes the influence of 3-stack players and the impact of duplicate commanders in a match.

Explaining Negative Averages

The reason the average of all average rating gains typically falls below zero is because new players, who begin with a default rating of 1000.0, often experience rating losses initially. Consequently, the average across all commanders tends to hover around -0.8. Any average rating gain above this threshold signifies that the commander performs better than the average, making it a valuable indicator.

Identifying Good and Bad Matchups

One of the most valuable applications of Average Rating Gain is in assessing commander matchups. By grouping commanders based on their opponents (the opposing players in the same lineup) and calculating the average rating gain for each matchup, we can discern which matchups favor a specific commander.

For instance, when considering Mengsk, the average rating gain versus Stettmann stands at a positive +2.77, indicating a favorable matchup. Conversely, facing Tychus yields an average rating loss of -3.63. The overall average across all matchups for Mengsk sits at -1.06. This data empowers players with valuable insights into commander strengths and weaknesses when pitted against different opponents.