A debate scoring rubric for practice: five criteria and a worked example
Use this original 100-point worksheet to review clarity, reasoning, evidence, responsiveness and listening. It is a practice aid, separate from Debagle’s scoring system and official tournament rules.
4 min readThe debagle team
Choose the purpose of the score
A practice rubric should help a speaker revise something specific. Before assigning points, agree on the question and the skill being assessed. Do not score an open conversation as if both people had promised to defend opposite sides.
The worksheet below is an original informal exercise. It is not Debagle’s production scoring formula, a validated assessment instrument or a tournament rulebook. Use it with a human review partner and keep notes explaining each rating.
Five criteria, twenty points each
Swipe across to compare all columns →
| Criterion | What to observe | What to avoid rewarding |
|---|---|---|
| Clarity | A specific claim, scope and proposed action | Confidence with no identifiable proposal |
| Reasoning | An explained connection between reasons and conclusion | Repeating the conclusion |
| Evidence | Relevant sources, examples and honest uncertainty | An uncited number said forcefully |
| Responsiveness | Direct engagement with the strongest objection | Changing the subject |
| Listening and comparison | Accurate paraphrase, concessions and tradeoff comparison | Interruptions or a distorted opposing view |
Use anchored levels rather than an unexplained number
For each criterion, use the same rough anchors: 0 when the feature is absent; 5 when it is mostly unclear; 10 when it is present with a significant gap; 15 when it is generally effective; and 20 when it is strong and specific throughout. Intermediate scores should come with a sentence explaining the distinction.
A quiet speaker can reason well. A fluent speaker can evade the objection. Keep presentation preferences separate from the criteria unless delivery is explicitly the task. If a recording failed, mark the missing passage as unassessable rather than scoring the absence as poor reasoning.
A worked hypothetical example
Suppose A proposes later Friday library hours and B raises staffing costs. A acknowledges the cost and proposes a funded three-month trial, but gives no estimate of the budget.
Swipe across to compare all columns →
| Criterion | Illustrative points | Reason |
|---|---|---|
| Clarity | 15 | Trial and time frame are clear; funding source is missing |
| Reasoning | 15 | Explains why evening access matters |
| Evidence | 10 | No local demand or cost data yet |
| Responsiveness | 15 | Answers staffing concern with a trial but leaves funding open |
| Listening and comparison | 15 | Acknowledges tradeoff; alternatives need more comparison |
| Total | 70 / 100 | An example of applying this worksheet, not a Debagle result |
Turn the score into a revision
The example’s next task is not “sound more confident.” It is to estimate the trial cost, identify evidence of unmet demand and compare the trial with another use of the same budget. A score that cannot suggest a revision is doing little work.
Have both speakers review the same passage independently. Discuss the biggest difference in their ratings. The disagreement may reveal a different interpretation of the motion, not just a harsher reviewer.
Read platform scores on their own terms
Different systems may use different scales and criteria. Do not convert this worksheet’s seventy into a Debagle result or an Elo rating. Read the platform’s scoring explanation and inspect the feedback for the actual exchange.
AI interpretation can be mistaken even when the scoring math is fixed for its inputs. External factual questions still need source checking. A debate review does not certify that the winning policy will work.
Sources: Debagle: the scoring system · Zheng et al.: Judging LLM-as-a-Judge with MT-Bench and Chatbot Arena
Questions people ask
Is this Debagle’s scoring formula?
No. It is an original practice worksheet. Debagle’s scoring system has its own explanation and scale.
Should every criterion be equally weighted?
Equal weights make this introductory exercise easy to inspect. Change weights only when both participants agree on the purpose before the round.
What if reviewers disagree?
Compare the passages and reasoning behind their ratings. A useful review makes its interpretation visible rather than claiming the number settles the dispute.