What's Up, Wratings?

1. Why don't you use Classic Elo, Glicko, or WHR?

While foundational to competitive rankings, systems like Classic Elo, Glicko, and WHR (Whole-History Rating) struggle fundamentally with the "pond problem"—how to handle mathematically disconnected groups of players. We found it impossible to solve this issue without reworking of the core of those systems. We migrated to OpenSkill because its Bayesian approach, combined with graph theory, naturally handles these disconnected networks. We discovered along the way that it also offers vastly superior predictive accuracy.

2. Why am I not in Pond 1?

Ponds are disjoint groups of players, a measure of mathematical connectivity, not skill. If you are not in Pond 1, it simply means that neither you, nor anyone you have played against, has played against a competitor in Pond 1 during the current time window. To merge into Pond 1, someone from your local pond needs to play a match against someone in the Pond 1 ecosystem, creating a mathematical bridge.

3. Why isn't my local tournament showing up in the calculations?

To maintain the integrity of the graph and filter out incomplete bracket data, Wratings processes tournaments with a minimum of 8 entrants by default. The Tournaments filter on the leaderboard also lets you restrict the view to larger events (64+ entrants), which focuses the rankings on results from major competitions. Additionally, because we use a sliding time window (e.g., a 100-Day Season), tournaments older than the current window are not included in the active calculation. Finally, if a Tournament Organizer has not fully finalized the bracket on Start.gg, our system will wait up to 14 days for the official results to lock in before processing it.

4. Will you support games other than Super Smash Bros.?

Yes. Our architecture is designed to be game-agnostic. We just integrated Super Smash Bros. Melee, with other competitive titles to follow based on data availability and community interest.

5. Why do some players appear as (id: 123456) instead of a name?

These are UGTs (Unidentified Gaming Tags). They are real competitors whose mathematical record exists in our database, but whose profile information is missing. This usually happens for one of two reasons:

• Anonymized Accounts: A player competed in a tournament, but later deleted their Start.gg account or invoked privacy rights to have their data scrubbed. Start.gg removes their name, but leaves their unique ID in the bracket.
• Unregistered Entrants: A player registered for a local tournament in person with cash. The Tournament Organizer manually added them to the bracket without linking a Start.gg account, leaving their profile blank.

Why are they on the leaderboard?

Mathematical Purity. If a UGT defeats a highly-ranked player, that upset must be mathematically accounted for. If we were to delete UGTs from the system, we would create "phantom nodes" in the graph, causing inexplicable rating drops for the players they defeated. Wratings is an objective, empirical engine. If the math dictates that a UGT is a 2000-rated threat, the ecosystem deserves to know that threat exists, even if we don't know their name.

Do You Know This Person? If you know who an ID belongs to, drop us a line at thisiswhatsup@wratings.gg!

6. How do I interpret the leaderboard controls?

The leaderboard dropdowns fall into two distinct mathematical categories. Understanding the difference is key to reading the results correctly:

• Engine Controls (Game, Connection, & Time Window): Each combination of these settings queries an entirely separate, independently calculated leaderboard. Changing the game, connection (Offline/Online) or the time window is equivalent to switching to a completely different power ranking.
• View Filters (Region, Tournaments, & Pond): These controls simply narrow down which players are displayed. They do not alter the underlying math or ratings—they only present a focused subset of the current calculation.

In short: adjusting the engine controls asks the system for a new ranking. Adjusting the view filters asks the system to highlight specific players within the current ranking.

7. How do I read the three tiers on the PvP Probabilities page?

The PvP dashboard provides three distinct levels of analysis, each pulling from a different layer of our data architecture:

• 1. Wrating Expectation: This is a simple, player-independent calculator. It uses a standard Elo logistic curve to predict a winner based only on the difference between two final Wrating numbers.
• 2. OpenSkill Projection: This is the true engine. It is player-dependent and context-dependent. It looks under the hood at the specific players' raw skill (μ) and uncertainty (σ) within the currently selected Time Window and Pond to calculate the true Gaussian probability.
• 3. Recorded Head-to-Head: This is a historical archive query. It counts the actual, physical sets played between these two specific competitors within the selected Time Window.

8. How do I read the odds on the PvP Probabilities page?

Both American Odds and Decimal Odds are displayed, side-by-side (e.g., +217 | 3.16x).

• American Odds (+ / -): Centered around a $100 baseline. A minus (-) indicates the favorite, showing how much you must bet to win $100 (e.g., -217 means bet $217 to win $100). A plus (+) indicates the underdog, showing how much you win if you bet $100 (e.g., +217 means a $100 bet wins $217).
• Decimal Odds (x): The global standard. It represents the total payout (initial stake + profit) for every $1 wagered. For example, 3.16x means a $1 bet returns $3.16 total.

9. What does "Last 40 Days" mean on the Evolution Graphs?

The dropdown menu shows two different timeframes to help you understand exactly what data you are looking at:

• "Last 40 Days" (The Calculation Window): This is the rule used for the math. It means every single dot on the graph is calculated using only the matches from the 40 days leading up to that specific date. The visible timeline on the graph grows organically as the system accumulates daily snapshots — it will lengthen automatically over time.

Why toggle between different windows?

Changing the calculation window allows you to analyze a player's development from different perspectives.

• Shorter windows (e.g., 40 days) are highly reactive. They highlight current momentum, hot streaks, and recent meta shifts, while providing a longer visible timeline on the graph.
• Longer windows (e.g., 100 days) are highly stable. They reveal a player's true baseline and consistency, filtering out the noise of a single bad tournament, though they yield a shorter visible timeline.

By comparing a player's short-term momentum against their long-term baseline, you can spot emerging trends, identify when a competitor is peaking, and make much more informed predictions about future matchups.

10. What does the "Updated" timestamp mean? Is my tournament included?

The "Updated" timestamp on the leaderboard represents the exact moment our data pipeline ended its daily scrape. It typically starts at 6:00 UTC. If your tournament finished and was fully finalized by the TO on Start.gg before our start time, it is mathematically locked in and included in the current calculation. If the bracket was left unfinalized, or if it is a multi-day major, our self-healing pipeline will continuously audit the tournament for up to 14 days and automatically include it the moment the final results are published.

11. Why does a player's graph sometimes show a sharp drop right before the line ends?

This is called Window Amnesia, and it is a feature of True Math, not a bug.

Wratings uses a rolling time window. As a player's most recent tournaments age out of the back of the window one by one, OpenSkill has fewer and fewer matches to work with. With less data, the algorithm's uncertainty about the player's true skill level grows — and a higher uncertainty mathematically pulls the rating estimate downward toward the prior. When the player's last match finally exits the window entirely, the line ends.

This means a rating drop near the end of a graph line almost always signals that the player has recently become inactive in that window, not that they have gotten worse. Their rating on a longer window — which still contains their full match history — will reflect their true standing.

We deliberately do not smooth, interpolate, or carry forward the last known value. Every dot on the graph is the exact output of the math engine for that day. No arbitrary choices, no hidden corrections.