Machine Learning / Quantitative Research Consultant for Esports (LoL LCK) Betting Model

We’re building an esports betting/prediction system for League of Legends (LCK) and we’re looking for a senior Machine Learning / Quantitative Research consultant to help us validate and improve our model before scaling to production. Our current pipeline trains on LCK-only historical match data (recent years, with time-decay weighting) and outputs win probabilities and expected value (EV/ROI). We need an expert who can audit the methodology, eliminate leakage, improve probability calibration, and help us implement evaluation that matches real betting conditions. This is a consulting role: you’ll work closely with our technical team, review code/data approach, propose fixes, and help implement the highest-impact improvements. What you’ll do Audit the ML pipeline for data leakage, label leakage, and look-ahead bias (especially in rolling stats/ELO/time-based features). Redesign evaluation to be production-realistic: time-based splits / walk-forward validation group splits by series/match (avoid map-level leakage) proper reporting of accuracy + log loss / Brier score / calibration Calibrate probabilities (e.g., Platt scaling / isotonic regression) and recommend confidence/uncertainty handling. Review and improve EV / ROI calculations and ensure consistent definitions (EV per bet vs ROI on risk, no-vig odds handling, vig/hold modeling). Help define bet selection rules and backtest methodology: edge thresholds, stake sizing (flat risk vs “to win” vs fractional Kelly), drawdown controls. Recommend feature and modeling improvements aligned with LoL reality: recency weighting/patch awareness matchup/counter and synergy features (duos/combos) player–champion proficiency signals series-state handling for BO series formats Provide a clear “production readiness checklist” and a plan for ongoing monitoring (drift, calibration, stability). Deliverables A written audit report identifying issues, risks, and prioritized fixes. A revised evaluation/backtest framework with reproducible methodology. Calibration results + recommended probability output format. Recommendations (and optionally implementation support) for model and feature improvements. A “go/no-go” assessment for launch. Required experience 5+ years in ML, quantitative research, or applied statistics (ideally in sports betting, trading, or forecasting). Deep familiarity with: leakage prevention, time-series / non-i.i.d. validation probability calibration and proper scoring rules (log loss, Brier, reliability curves) backtesting pitfalls (selection bias, survivorship bias, data snooping) Strong Python skills (pandas, numpy, scikit-learn; bonus for PyTorch/XGBoost/LightGBM). Comfortable reviewing code and giving actionable engineering guidance. Nice to have Experience with sports betting markets, odds → implied probability conversions, vig removal, CLV tracking. Knowledge of esports / League of Legends (draft, patch effects, meta shifts). Experience designing end-to-end ML systems in production (monitoring, drift, versioning). Project details Type: Consulting + code review + implementation support Data: LCK match-level + map-level historical dataset Output: win probability + EV/ROI suggestions per map/match Start: ASAP Duration: 2–6 weeks initial engagement (possible ongoing advisory) To apply, please include A brief summary of relevant ML + quantitative research experience (betting/trading/forecasting is a plus). Examples of prior work around calibration, time-based validation, or backtesting. Your recommended first steps to audit a model that shows unusually high accuracy/ROI (how you’d detect leakage quickly).