[Feature Proposal] Explainable Score Analysis System

### Is your feature request related to a problem?

Yes. Currently, users receive a final score/rating at the conclusion of a debate, but there is a lack of transparency regarding the "why" behind the numbers. This "Black Box" scoring makes it difficult for debaters to identify specific areas for improvement, such as whether they struggled with logical consistency, oratorical delivery, or rebuttal effectiveness.

---

### Description of the Proposed Feature

The Explainable Score Analysis (ESA) system aims to decompose the aggregate score into granular, interpretable components.
This feature will provide users with a detailed post-match report that balances quantitative data with qualitative insights.

---

### Key Components:

- Categorical Breakdown: Scores divided into specific pillars (e.g., Argument Strength, Evidence Quality, Rebuttal Efficiency, and Delivery).
- Sentiment/Tone Analysis: Feedback on whether the user's tone was appropriate for the debate context.
- Critical "Turning Points": Identifying specific exchanges in the transcript where the user gained or lost the lead.
- Actionable Coaching: 2–3 specific "Next Steps" generated by the AI based on the weaknesses identified.

---

### User Impact

- Educational Growth: Users can target specific skills (e.g., "I need to work on my rebuttals").
- Trust & Transparency: Users are less likely to feel "cheated" by an AI judge if they can see the logic behind the deduction.
- Engagement: Detailed feedback loops encourage users to return and try to improve specific sub-scores.

---

### Proposed Implementation Ideas

- Backend: Update the LLM prompt instructions to return a structured JSON object containing both the total score and a breakdown by criteria.
- Frontend: Introduce a "Score Summary Card" and a "Deep Dive" tab in the post-debate UI.

Start by locating the LLM prompt instructions and the post-debate UI described in the proposal. Define how the structured score breakdown, sentiment and turning-point feedback, and coaching steps should be represented, then determine how the Score Summary Card and Deep Dive tab will display them; done means the backend and frontend support the complete explainable report.

Written by the indexing model from the issue text.