You just beat someone 200 points above you and gained... 23 points. But yesterday you beat someone at your level and gained 16. Why the difference? And what would have happened if the K-factor was 48 instead of 32?
These are the kinds of questions an Elo calculator answers instantly. Instead of guessing, you plug in the numbers and see exactly what happens. Our Elo Rating Calculator lets you experiment with ratings, match outcomes, and every parameter that affects the result — including advanced settings like loss dampening, rating capping, and free-for-all distribution.
This guide walks through the math behind the calculator, explains each parameter, and shows you how to use the results to configure your own competitive community.
How Elo Calculations Work
Every Elo calculation boils down to two steps: figure out who was expected to win, then adjust ratings based on what actually happened.
Step 1: Expected Score
The expected score tells you the probability of each player winning based on their current ratings. The formula is:
Expected Score = 1 / (1 + 10^((opponent_rating - your_rating) / influence_range))
The influence_range (default 400) controls how much a rating gap affects the probability. With the standard value of 400, a 200-point advantage gives roughly a 76% expected win rate. A 400-point gap gives roughly 91%.
Here's a quick reference table:
| Rating Difference | Favorite's Win Probability |
|---|---|
| 0 | 50% |
| 100 | 64% |
| 200 | 76% |
| 300 | 85% |
| 400 | 91% |
| 500 | 95% |
Both players' expected scores always add up to 1. If you're expected to win 76% of the time, your opponent is expected to win 24%.
Step 2: Rating Change
After the match, each player's rating adjusts based on the difference between the actual result and the expected result:
Rating Change = K × (Actual Score - Expected Score)
- Actual Score: 1 for a win, 0 for a loss, 0.5 for a draw
- K: the K-factor (default 32), which scales the magnitude of every rating change
Let's work through a real example.
Worked Example: Underdog Wins
Player A is rated 1200. Player B is rated 1450. Player A wins.
Step 1 — Expected scores:
- Player A's expected score:
1 / (1 + 10^((1450 - 1200) / 400))=1 / (1 + 10^0.625)= 0.191 (19.1%) - Player B's expected score:
1 - 0.191= 0.809 (80.9%)
Step 2 — Rating changes (K = 32):
- Player A (won):
32 × (1 - 0.191)= +25.9 → New rating: 1226 - Player B (lost):
32 × (0 - 0.809)= -25.9 → New rating: 1424
The underdog gained 26 points instead of the 16 they'd gain against an equal opponent. That's the Elo system working as designed — bigger rewards for beating stronger opponents.
If Player B had won instead (the expected outcome), they'd gain only 32 × (1 - 0.809) = +6.1 points. Expected wins produce small gains.
Worked Example: Even Match
When two players have identical ratings, the math is simple:
- Expected score for both: 0.5 (50/50)
- Winner gains:
32 × (1 - 0.5)= +16 - Loser drops:
32 × (0 - 0.5)= -16
This is the baseline. Every other scenario is a variation on this — stronger favorites gain less, bigger upsets gain more.
Understanding K-Factor
K-factor is the single most important parameter to get right for your community. It controls the maximum number of rating points a player can gain or lose from a single match.
With K = 32 (the default), the maximum possible gain is 32 points (beating someone with an infinitely higher rating). In practice, most gains fall between 8 and 26 points.
| K-Factor | Max Change | Best For |
|---|---|---|
| 16 | ±16 | Large communities with frequent matches. Ratings are stable and take many games to shift. |
| 24 | ±24 | Active communities where you want steady progression without wild swings. |
| 32 | ±32 | The standard default. Works for most communities. |
| 48 | ±48 | Small groups where players don't play often. Ratings converge faster. |
| 64 | ±64 | Very small groups or casual settings. Every match has a big impact. |
How to choose: If players in your community complain that their rating doesn't move enough, increase K. If they complain about losing too many points from a single bad game, decrease K. Start with 32 and adjust from there.
Try different K-factor values in the calculator to see the difference in practice.
For the numbers behind the choice, see Elo K-Factor Curves — simulated convergence and stability figures for each K value, why picking one number for everyone forces a tradeoff, and how a provisional K curve avoids it.
Advanced Parameters
The standard Elo formula covers most use cases, but real competitive communities often need more control. Here's what each advanced parameter does and when to use it.
Influence Range
Default: 400
The influence range controls how steep the probability curve is. It determines the rating difference at which the favorite's expected win rate reaches ~91%.
- Lower values (200–300): Small rating gaps produce more lopsided expectations. A 200-point difference might give the favorite a 91% expected win rate instead of 76%.
- Higher values (500–800): The system treats rating gaps as less meaningful. Even large differences produce closer-to-50/50 expectations.
Most communities should leave this at 400. Only adjust it if your rating scale is compressed (low influence range) or if you want upset wins to be worth less (high influence range).
Curve Factor
Default: 10
The curve factor is the base of the exponent in the expected score formula. The standard Elo system uses 10 (as in 10^(difference/400)). Changing this adjusts the shape of the probability curve.
This is a rarely-used parameter. The default of 10 is the standard mathematical foundation of Elo and there's almost never a reason to change it.
Loss Dampening
Default: 0%
Loss dampening reduces the points a player loses after a defeat by a percentage, while keeping win rewards unchanged.
| Dampen % | Effect on a -16 Loss |
|---|---|
| 0% | -16 (no change) |
| 25% | -12 |
| 50% | -8 |
| 75% | -4 |
When to use it: Communities where player retention matters more than strict mathematical accuracy. If newer players are quitting because losses feel too punishing, 15–25% dampening softens the blow without making wins feel meaningless.
Trade-off: Loss dampening causes rating inflation over time, since more points enter the system than leave it. Ratings will gradually drift upward. This is usually fine for casual communities but matters for competitive leagues where rating represents absolute skill.
Rating Capping
Default: Off
Rating capping prevents extreme rating changes when the skill gap is very large. Without capping, a 1800-rated player beating a 900-rated player gains almost nothing (+1–2 points), and the 900-rated player loses almost nothing. Capping sets minimum gains and maximum losses so that every match feels meaningful.
The capping parameters are:
| Parameter | What It Controls |
|---|---|
| Cap Range | Rating difference threshold where capping activates |
| Favorite Win Impact | Minimum points the favorite gains for winning |
| Favorite Loss Impact | Maximum points the favorite loses for losing |
| Underdog Win Impact | Maximum points the underdog gains for winning |
| Underdog Loss Impact | Minimum points the underdog loses for losing |
When to use it: Communities with wide skill gaps where high-rated players feel like beating lower-rated opponents is a waste of time. Capping guarantees a minimum reward.
Max Advantage
Default: Unlimited
Max advantage caps the effective rating difference used in the expected score calculation. If set to 600, a 1200 vs 1900 matchup is treated as 1200 vs 1800 for calculation purposes.
This prevents extreme probability calculations when mismatches happen and works well alongside rating capping.
Free-for-All (FFA) Calculations
Standard Elo is designed for two-player matches, but many games involve 3 or more players competing simultaneously — battle royales, racing, FFA shooters, and party games.
The Elo Calculator supports FFA mode with up to 24 players. Here's how it works.
How FFA Elo Works
In a free-for-all match, each player is compared against every other player in a series of pairwise matchups. If you place 1st out of 6 players, the system treats it as if you beat all 5 opponents individually. If you place 3rd, you beat players 4th–6th and lost to players 1st–2nd.
The rating changes from each pairwise matchup are summed together.
FFA Distribution
Default: On
Without distribution, FFA matches produce much larger rating swings than 1v1 matches because each player has multiple opponents. A 6-player FFA generates 5 pairwise comparisons per player.
When FFA distribution is enabled, the total rating change is divided by the number of opponents. This keeps FFA rating changes on the same scale as 1v1 matches:
| 6-Player FFA (K=32) | Distribution Off | Distribution On |
|---|---|---|
| 1st place (5 wins) | ~+80 | ~+16 |
| Last place (5 losses) | ~-80 | ~-16 |
Recommendation: Keep distribution on unless you want FFA matches to have a much bigger impact on ratings than 1v1 matches.
FFA Example
Five players enter an FFA match:
| Player | Rating | Placement |
|---|---|---|
| Alice | 1400 | 1st |
| Bob | 1350 | 2nd |
| Carol | 1200 | 3rd |
| Dave | 1250 | 4th |
| Eve | 1100 | 5th |
Alice (1st) is compared pairwise against Bob, Carol, Dave, and Eve — winning all four. Bob (2nd) beat Carol, Dave, and Eve but lost to Alice. And so on.
With distribution on and K=32, the results are comparable to 1v1 swings: Alice might gain +12, Eve might lose -14. The exact numbers depend on the rating differences in each pairwise matchup.
Plug this scenario into the calculator's FFA mode to see the exact breakdown for each pairwise comparison.
What's a Good Elo Rating?
This depends entirely on your community, but here's a general distribution for a system starting at 1200:
| Rating Range | Tier | Approximate Percentile |
|---|---|---|
| Below 1000 | Beginner | Bottom 15% |
| 1000–1150 | Below Average | 15–35% |
| 1150–1250 | Average | 35–65% |
| 1250–1400 | Above Average | 65–85% |
| 1400–1600 | Strong | 85–95% |
| 1600+ | Elite | Top 5% |
In active communities, ratings typically settle into a bell curve centered around the starting value. Most players cluster within 200 points of the default. The extremes (below 900 or above 1700) contain only the most active players who have played enough matches for their rating to diverge significantly.
How many games until my rating stabilizes? Roughly 20–30 matches. During placement matches (usually the first 10), ratings move more aggressively to quickly sort players. After that, each match has a more moderate effect. The leaderboard setup guide covers how to configure placement matches.
From Calculator to Real Matches
The calculator is useful for understanding the math and experimenting with parameters. But the real value comes when you apply those settings to a live leaderboard.
If you've found a K-factor and parameter set that feels right in the calculator, you can apply those exact settings to your Discord server's leaderboard using the /leaderboard_config set command. The calculator even generates the command for you — look for the Discord command button after running a calculation.
Here's what the workflow looks like:
- Experiment in the Elo Calculator with different settings
- Find parameters that produce rating changes you're happy with
- Copy the command and run it in your Discord server
- All future matches on that leaderboard will use your tuned settings
If you haven't set up a leaderboard yet, start with the setup guide. It takes about 2 minutes to get your first match recorded.
FAQ
How is Elo calculated?
Elo uses two steps: first, it calculates the expected probability of each player winning based on the rating difference. Then it adjusts both ratings based on the actual result. The winner gains points and the loser drops points, with the amount determined by how surprising the result was. Use the Elo Calculator to see the exact math with your own numbers.
What K-factor should I use?
Start with 32, the standard default. If your community plays a lot of matches and you want stable ratings, lower it to 24 or 16. If matches are rare and you want each one to matter more, raise it to 48 or 64. The right K-factor depends on how many matches your players typically play per week — Elo K-Factor Curves has the convergence numbers for each value.
Can Elo work for free-for-all games?
Yes. FFA Elo works by comparing each player against every other player in pairwise matchups based on finishing position. With FFA distribution enabled, the rating changes are normalized to stay on the same scale as 1v1 matches. The calculator supports FFA with up to 24 players.
What's the difference between Elo and MMR?
Elo is a specific mathematical formula for calculating skill ratings. MMR (Matchmaking Rating) is a generic term for any hidden rating used to match players. Many games use MMR systems that are based on or inspired by Elo, but the visible rank (Gold, Diamond, etc.) is often loosely connected to the underlying number. Read more in our Elo rating explainer.
Does loss dampening cause rating inflation?
Yes, gradually. Loss dampening means more rating points enter the system (from wins) than leave it (from dampened losses). Over time, the average rating in your community will drift upward. For casual communities this is usually fine — it makes the experience more rewarding. For strict competitive leagues, keep dampening at 0% or use a low value (10–15%).
How does Team Up handle Elo differently from chess?
Team Up extends standard Elo for gaming scenarios that chess doesn't have: team matches (using average team rating), free-for-all with up to 24 players, multiple rating types per leaderboard, character-specific ratings, and advanced parameters like loss dampening and rating capping. The core formula is the same — the extensions handle the complexity of modern competitive gaming.
Further reading:
- Elo Rating Calculator — experiment with ratings and parameters
- What Is Elo Rating? — the fundamentals of how Elo works
- How to Set Up Elo Leaderboards — get started in 2 minutes
- Best Discord Bots for Elo & Matchmaking — compare the top options
- Configuration Guide — all leaderboard settings explained