Pairwise Learning-to-Rank with Hungarian Assignment for NCAA Tournament Seed Prediction

A constrained ranking system that predicts exact 1-through-68 seed assignments for all NCAA March Madness tournament teams using pairwise comparison, ensemble blending, combinatorial optimization, and domain-specific post-processing.

Predicting the NCAA Selection Committee's seeding decisions is a challenging constrained ranking problem: each of 68 teams must receive a unique seed, rankings are relative rather than absolute, and the committee's decision process incorporates substantial subjective judgment. We present a four-stage pipeline that transforms this into a tractable learning-to-rank task. Stage 1 converts the problem from 68 point predictions into ~4,500 pairwise comparisons per season, increasing effective training data by 66×. Stage 2 blends three classifiers (Logistic Regression and XGBoost) with cross-validation-optimized weights. Stage 3 applies the Hungarian algorithm to enforce the one-team-per-seed constraint via globally optimal assignment. Stage 4 applies domain-specific zone corrections targeting systematic committee biases. On leave-one-season-out cross-validation across 5 seasons (340 teams), the model achieves 91.2% exact-match accuracy (83/91) with RMSE 0.392. On the held-out 2025–26 tournament, it correctly identifies 85.3% of the field (58/68), places 77.6% of predictions within ±2 seeds, and achieves RMSE 2.543 — within the pre-registered expected range of 2–3.

Keywords: learning-to-rank, Hungarian algorithm, constrained optimization, ensemble methods, sports analytics, NCAA basketball

                    ┌──────────────────────────────────┐
                    │     20 Raw Statistical Features   │
                    │  (NET, SOS, W-L, Quads, Conf.)   │
                    └──────────────┬───────────────────┘
                                   │
                    ┌──────────────▼───────────────────┐
                    │   Feature Engineering (→ 68 dims) │
                    │  Ratios · Composites · Context    │
                    └──────────────┬───────────────────┘
                                   │
               ┌───────────────────┼───────────────────┐
               ▼                   ▼                   ▼
    ┌─────────────────┐ ┌─────────────────┐ ┌─────────────────┐
    │  LR (C=5.0)     │ │  LR (C=0.5)     │ │  XGBoost        │
    │  68 feats       │ │  Top-25 feats   │ │  68 feats       │
    │  Adj-pairs ≤30  │ │  All pairs      │ │  All pairs      │
    │  Weight: 64%    │ │  Weight: 28%    │ │  Weight: 8%     │
    └────────┬────────┘ └────────┬────────┘ └────────┬────────┘
             └───────────────────┼───────────────────┘
                                 ▼
                    ┌──────────────────────────────────┐
                    │  Dual-Hungarian Ensemble          │
                    │  75% Pairwise + 25% Ridge(α=10)  │
                    │  ↓ Hungarian Assignment (p=0.15) │
                    │  Globally optimal 1-to-68 mapping │
                    └──────────────┬───────────────────┘
                                   │
                    ┌──────────────▼───────────────────┐
                    │  7 Zone Corrections + AQ↔AL Swap  │
                    │  Domain-specific post-processing  │
                    └──────────────┬───────────────────┘
                                   │
                    ┌──────────────▼───────────────────┐
                    │     Final Seed Assignments 1–68   │
                    └──────────────────────────────────┘

Instead of directly regressing seed values, we frame the problem as pairwise learning-to-rank: for each pair of teams $(i, j)$, we compute $\mathbf{x}_i - \mathbf{x}_j$ and train a classifier to predict $P(\text{seed}_i < \text{seed}_j)$. This converts 68 training examples per season into ~4,500 informative comparisons.

Three pairwise classifiers are blended with learned weights:

• Component 1 (64%): Logistic Regression (C=5.0), full 68 features, adjacent pairs only (gap ≤ 30)
• Component 2 (28%): Logistic Regression (C=0.5), top-25 features, all pairs
• Component 3 (8%): XGBoost (depth=4, 300 trees, lr=0.05), full features, all pairs

Pairwise scores yield a continuous ranking, but valid seeds require a discrete bijection from teams to ${1, \ldots, 68}$. We construct a cost matrix $C_{ij} = |s_i - j|^{0.15}$ where $s_i$ is team $i$'s raw score, and solve the assignment problem via the Hungarian algorithm (Kuhn, 1955). A dual ensemble blends the pairwise model (75%) with a Ridge regression committee model (25%) before final assignment.

Seven seed-range-specific correction rules address systematic committee biases (e.g., mid-major auto-qualifiers under-seeded, power-conference at-large teams over-seeded). Corrections only re-order teams within assigned seeds — they cannot introduce or remove assignments.

The gap between cross-validation RMSE (0.392) and held-out RMSE (2.543) — a 6.5× blowup — provides an empirical case study in overfitting with limited training data (N=340). Key findings:

├── README.md                             # This file
├── LICENSE                               # MIT License
├── CITATION.cff                          # Citation metadata
├── requirements.txt                      # Python dependencies
├── NCAA_vid.mp4                          # Project video presentation
│
├── ncaa_2026_model.py                    # Core model: features, training, inference
├── generate_kaggle_submission.py         # LOSO cross-validation & submission CSV
├── predict_2026.py                       # End-to-end 2025–26 prediction runner
│
├── data/
│   ├── NCAA_Seed_Training_Set2.0.csv     # 249 labeled teams (2020–2024)
│   ├── NCAA_Seed_Test_Set2.0.csv         # 91 labeled teams (held-out seasons)
│   ├── NCAA Statistics.xlsx              # Full D-I stats, 2025–26 season
│   └── NCAA_2026_Data.csv                # Processed 68-team model input
│
└── output/                               # Model predictions & submissions
    ├── submission_kaggle.csv
    └── 2026/                             # 2025–26 bracket predictions

git clone https://github.com/Om-singhaI/NCAA.git
cd NCAA
python -m venv .venv && source .venv/bin/activate
pip install -r requirements.txt

python generate_kaggle_submission.py

This runs leave-one-season-out cross-validation on all 340 labeled teams and outputs per-season accuracy, zone correction tables, and the competition submission CSV.

# 1. Place NCAA Statistics Excel in data/
# 2. Generate predictions:
python predict_2026.py
# → output/2026/seed_selections_2026.txt
# → output/2026/submission_2026.csv

Each team record contains 20 statistical features:

These 20 raw features are engineered into 68 model features across six categories: raw rankings, parsed win-loss metrics, quadrant quality scores, composite ratings, bid-type interactions, and historical context features.

This model evolved through 50 iterations, each targeting specific failure modes:

If you use this work in your research, please cite:

@software{singhal2026ncaa,
  author       = {Singhal, Om},
  title        = {Pairwise Learning-to-Rank with Hungarian Assignment for {NCAA} Tournament Seed Prediction},
  year         = {2026},
  url          = {https://github.com/Om-singhaI/NCAA},
  note         = {NCAA Final Four Analytics Challenge}
}

This project is licensed under the MIT License — see LICENSE for details.

• ESPN bracketology projections for 2025–26 field composition
• The Hungarian algorithm implementation via SciPy (linear_sum_assignment)