RailTC Prediction Methodology

1. How RailTC Predicts Today

RailTC is a hybrid confirmation engine, not a heuristics-only calculator and not an unaccountable black box.

• Trained models on verified outcomes: Models are fit on large volumes of real journeys whose final chart status (CNF / RAC / WL) was later verified.
• Transparent baseline scorer: A readable multi-signal baseline (quota, class, time-to-chart, route patterns) supports explainability and fallback.
• Public accountability: Aggregate accuracy is published on the Accuracy dashboard. Overall accuracy uses scored Safe + Risky bands; Medium is not scored.

Earlier public notes that described RailTC as "rules-only" or "heuristics-only" are outdated. The current architecture write-up is on Project Architecture. Machine-readable facts: /llms.txt.

2. Inputs We Consider

RailTC combines current ticket state with route-pattern context. We do not rely on a single factor like only WL number.

• Current status and waitlist progression context (WL/RAC/CNF transition behavior).
• Quota behavior differences (for example GNWL vs RLWL style movement).
• Class-level cancellation tendencies and seat-release patterns.
• Time left to chart preparation, because volatility increases near charting.
• Route- and train-specific historical outcome baselines from verified records.

3. Scoring Framework

We convert the inputs into a probability score and then map it into user-friendly confidence bands. The mapped band is designed for decision support, not guaranteed travel authority.

4. Why Medium Is Not Scored As Correct/Incorrect

Medium predictions represent uncertainty buckets where final outcomes can still swing due to late cancellations or charting dynamics. To avoid inflating headline accuracy, RailTC treats Medium as "Not Scored" while still verifying and displaying those journeys. This is an accounting choice for honesty, not a claim that Medium is permanently unpredictable. Improving Medium coverage is an active modeling focus.

You can review this policy on the Accuracy page where scored outcomes are separated from uncertain cases.

5. Verification And Feedback Loop

• Prediction is generated for active PNR state.
• After chart preparation, final outcome is captured when available.
• Outcome is matched against prediction banding policy.
• Aggregate performance is published publicly on the accuracy dashboard.
• Thresholds are tuned conservatively to reduce false confidence.

6. What We Do Not Use

• No scraping or publishing of sensitive passenger personal details for prediction copy.
• No paid "guarantee" logic claiming final chart control.
• No hidden upsell dependency to unlock core risk labeling.

7. Practical Limits

Final chart authority remains with Indian Railways. Sudden operational interventions, quota shifts, or exceptional-day demand can still produce outlier results. RailTC is best used as a planning signal, not as a legal travel confirmation document.

8. How To Use This In Real Decisions

• If score is high, continue monitoring but keep normal travel plan.
• If score is medium, delay irreversible expenses and monitor chart updates.
• If score is risky, keep a backup only when missing the trip would be costly — weak bands are not automatic cancellations.
• For booking-stage decisions, use Pre-Booking Analysis.