Methodology

Our work is open. Anyone can reproduce it from the same inputs. Nothing about the method is secret.

School & program rankings (live now)

These lists are taken directly from published standings and results of the organizations named on each page: college sweepstakes and points reports (NPDA, NDT) and high school state results (Texas UIL, Iowa IHSSA). We keep each source as its own list and do not blend different circuits into a single score, because their rules differ. For each list we show what the source published (placement, points, honors), the source link, and the date we retrieved it. We record school names and placements only; student names are not collected.

• Where a source is old (for example the 2017-18 NDT report), the list is marked historical.
• Where a source does not rank (Texas UIL publishes placings, not an overall rank), we show placings, not a made-up rank.
• If a document cannot be read reliably, we skip it rather than guess.
• We read each site's robots.txt, space our requests, identify ourselves, and stop if a site shows an anti-bot challenge.

Individual-entry rankings (as results are submitted)

The model below powers entry-level rankings once coaches and tournament hosts submit round results (submit). No entry rankings are shown until real, permitted data exists.

What is rated

An entry is one debater or team in one event and level (for example, High School Public Forum) as published by the tournament: school plus the entry code the source shows (usually initials). We do not collect or display full student names, contact details, or judge identities. School and state pages aggregate entries.

The model

We use Glicko (Glickman, 1999), a head-to-head rating system like chess Elo that also tracks uncertainty. Every entry starts at 1500 with high uncertainty. After each tournament, each entry moves up for wins and down for losses, by more when the result was surprising given both entries' ratings. Uncertainty shrinks as rounds accumulate and grows again during inactivity.

• Tournament strength: rounds count 1.0 normally, 1.25 (octafinal-bid), 1.5 (quarterfinal-bid), 1.75 (semifinal-bid) or 2.0 (final-bid) times, reflecting stronger fields.
• Elimination rounds count 15% more than preliminary rounds.
• Window: the last 365 days. Older results age out.
• Listing rule: at least 12 rounds across 3 tournaments. Fewer than that and an entry is not shown, because early ratings are noisy.
• Range: the shown range is rating plus or minus two standard deviations. Entries whose ranges overlap are effectively tied.
• Schools: average of a school's top 3 listed entries in an event; overall school rating averages its events. States: average of the top 5 schools.
• No byes or forfeits: only decided head-to-head rounds count. Speaker points, judge preferences, and ballots are not used.

Known limits

• Only tournaments whose results we are permitted to use are included. Missing tournaments mean missing evidence; a missing entry is not a judgment.
• Teams that mostly debate within one region are connected to the wider pool only through shared tournaments; their ratings are less certain.
• Partner changes and school transfers can split an entry's history.
• Congressional debate and individual-speech events are not head-to-head and are not rated here.

Sources and permission

Each round in the ranking carries the page it came from and the basis on which we may use it (for example written permission from the data owner, host-published, or coach-submitted). Entry pages list those sources. We respect robots.txt and rate limits and do not collect pages a site tells automated tools to avoid.

Permission types: written permission from the data owner, host-published, coach-submitted, or robots-allowed public pages.

AI explanations

The plain-language explanation and improvement tips on an entry page are generated from that entry's own numbers. An optional AI model can rephrase them; it receives only those numbers (never free text from visitors), is instructed to give debate coaching only, and can be wrong.

Not affiliated with NSDA or Tabroom. Model version: glicko1-weighted-v1.