Rating System

Frontier tracks competitive skill using a per-ladder rating system. Two algorithms are available: Glicko-2 (the default) and Showdown K-Scaling. Each ladder can use a different one.

Glicko-2

Developed by Mark Glickman (2013). The algorithm tracks three values per player per ladder:

Value	What it represents	Default
Rating (R)	Skill estimate on a 0—9999 scale.	1500
Rating Deviation (RD)	Uncertainty in the rating. Lower = more confident.	350
Volatility (sigma)	Expected fluctuation in performance.	0.06

How Rating Changes Work

After each match, the algorithm compares the expected outcome (based on both players’ ratings and RDs) against the actual result (1.0 for a win, 0.0 for a loss). Then it updates all three values for both players.

Bigger rating swings happen when:

• The player’s RD is high (uncertain rating — the system is still learning).
• The opponent’s RD is low (confident opponent — beating a known quantity means more).
• The result is unexpected (upset victories move the needle fast).

Smaller rating swings happen when:

• The player’s RD is low (well-established rating).
• The opponent’s RD is high (uncertain opponent — less informative match).
• The result is expected.

RD and Inactivity

RD drifts upward when a player stops playing. The system gets less confident about a dormant player’s skill over time. When they return, their rating moves more aggressively until RD settles back down.

With regular play, RD drops toward rdFloor (default 50), signaling a well-established rating.

Configuration

rating {
    algorithm = GLICKO2
    startRating = 1500.0
    startRD = 350.0
    startVolatility = 0.06
    tau = 0.5
    rdFloor = 50.0
    minRating = 0.0
    maxRating = 9999.0
}

Field	Type	Default	Description
startRating	Double	1500.0	Initial rating for new players.
startRD	Double	350.0	Initial rating deviation.
startVolatility	Double	0.06	Initial volatility.
tau	Double	0.5	Controls how quickly volatility changes. Higher = more reactive to upsets. Recommended range: 0.3—1.2.
rdFloor	Double	50.0	Minimum RD. Prevents ratings from becoming “too certain.”
minRating	Double	0.0	Floor for the rating value itself.
maxRating	Double	9999.0	Ceiling for the rating value.

Tau explained

The tau parameter is a system constant that governs volatility updates. A low tau (0.3) makes the system slow to react — good for large player pools where upsets are statistical noise. A high tau (1.2) makes it react fast — better for small servers where every match is meaningful. The default of 0.5 is a solid middle ground.

Showdown K-Scaling

A simpler alternative inspired by Pokemon Showdown’s ladder. Elo-style rating with a K-factor that scales based on games played.

How It Works

1. New players start with a high K-factor (kMax = 50), so their rating moves quickly.
2. As they play more games, K decreases linearly toward kMin = 20 over kScalingGames = 20 matches.
3. Established players have smaller rating swings, creating more stable leaderboards.

The K-factor formula:

K = kMax - (kMax - kMin) * min(gamesPlayed, kScalingGames) / kScalingGames

For a player with 10 games played (defaults): K = 50 - (50 - 20) * 10/20 = 35.

Configuration

rating {
    algorithm = SHOWDOWN_K_SCALING
    startRating = 1500.0
    startRD = 350.0
    kMax = 50.0
    kMin = 20.0
    kScalingGames = 20
    rdFloor = 50.0
    minRating = 0.0
    maxRating = 9999.0
}

Field	Type	Default	Description
kMax	Double	50.0	K-factor for brand-new players. Large swings early on.
kMin	Double	20.0	K-factor floor for experienced players.
kScalingGames	Int	20	Number of games over which K scales from max to min.

Anti-Boost Protection

Prevents rating farming by repeatedly matching the same opponent. Rating changes are multiplied down for repeat matchups within a time window.

antiBoost {
    windowHours = 24
    match1Multiplier = 1.0
    match2Multiplier = 0.5
    match3PlusMultiplier = 0.1
    blockSameIp = false
}

Field	Type	Default	Description
windowHours	Int	24	Time window for tracking repeated matchups.
match1Multiplier	Double	1.0	Rating change multiplier for the first match against an opponent. Full value.
match2Multiplier	Double	0.5	Second match in the window. Half value.
match3PlusMultiplier	Double	0.1	Third and subsequent matches. 10% value.
blockSameIp	Boolean	false	Block matchmaking entirely between players on the same IP.

The multiplier applies to the rating change, not the rating itself:

actualChange = calculatedChange * antiBoostMultiplier
newRating = oldRating + actualChange

Small servers

On a server with a tight-knit community where the same players match often, consider raising match2Multiplier to 0.75 and match3PlusMultiplier to 0.25. Too aggressive a decay makes ratings stagnate.

Player Stats Tracked

For each player on each ladder, Frontier tracks:

Stat	Description
Rating	Current rating on the ladder.
Rating Deviation	Current RD (uncertainty).
Volatility	Current volatility (Glicko-2 only).
Highest Rating	Peak rating ever achieved.
Wins / Losses	Total win and loss counts.
Current Win Streak	Consecutive wins (resets on loss).
Current Loss Streak	Consecutive losses (resets on win).
Best Win Streak	Longest win streak ever achieved.
Last Match Timestamp	When the player last played on this ladder.

Rating Confidence Labels

Frontier displays a confidence label in GUIs based on the player’s current RD:

Label	RD Range	Meaning
Stable	RD ≤ 50	Rating is well-established. This player has played enough recent matches for a reliable estimate.
Uncertain	RD 50—100	Rating is moderately certain. Could shift noticeably with a few matches.
Volatile	RD > 100	Rating is highly uncertain or provisional. Expect large swings.

These labels are configurable in messages.conf under rdDescriptions.