Ryno's Edge — NRL Edition

NRL Predictive Intelligence

Welcome to Ryno's Edge

The ultimate command center for rugby league analytics. Combining advanced machine-learning ensembles, real-time ELO ratings, and AI-parsed media sentiment, Ryno's Edge gives you professional-grade data projections to outsmart the competition.

Round ... Analytics Active

ML Predictions

Ensemble algorithms projecting margins, scores, and win rates.

ELO Rankings

Real-time team strength indices and multi-season rating trajectories.

News Sentiment

Gemini AI parsing articles, injury wards, and late-mail intelligence.

Player Stats

Search and filter 30+ individual player metrics across 10 seasons.

Round Schedule

Home Team Away Team

Power Rankings

The Edge Pick

AI High Confidence

System Matrix Sync Status

NRL Draw (Fixtures) ...

Market Odds ...

News Sentiment ...

Team Lineups (Squads) ...

Player Stats (Hist.) ...

Last Overall Sync: ...

What is the Player Leaderboard?

The Player Leaderboard ranks every NRL player across any statistical category — from run metres and tries to missed tackles and penalties. Filter by season, team, position and stat to uncover who's leading the competition in the areas that matter most to you.

Choose Your Stat

Pick from 30+ stats across attack, defence, kicking, and discipline — from run metres to 40/20s.

Best Game vs Season

Switch between a player's best single performance, their season total, or their per-game average.

Multi-Season

Hold Ctrl and click to select multiple seasons at once — compare across years or see all-time leaders.

Position Filter

Narrow to a specific position like Halfback or Prop to see who leads their position group.

Season(s)

Ctrl+click to multi-select.

Team (optional)

Stat

Mode

Top N

Position

What are Season Trends?

Season Trends plots any statistic round-by-round across the entire competition, letting you visually track how teams evolve through a season. Spot momentum shifts, identify consistent performers, and see which teams peak at the right time.

Round by Round

Each point on the chart represents a round's performance — see exactly when a team's form rises or falls.

Toggle Teams

Use the team chips below the controls to show or hide individual teams for a cleaner comparison.

Avg vs Total

Switch between the team's total output per round or per-player average to normalise for bench usage.

Any Stat

Chart everything from run metres and tries, to errors and missed tackles to spot defensive trends.

Season

Stat

Aggregation

Select Teams to Display

Load a season first…

What is a Player Profile?

The Player Profile gives you a complete statistical breakdown of any NRL player — career history, season-by-season averages, recent form, and how they compare to the league average at their position. Start typing any player's name to pull up their full dossier.

Career Stats

Full career stats from 2015 onwards, broken down by season with totals and per-game averages.

Recent Form

Game-by-game performance for recent matches — see whether a player is trending up or down heading into the weekend.

Search Any Player

Just start typing — the search auto-suggests matching players from the full NRL database.

Type a player name above to load their profile.

What is Round Results?

Round Results shows every completed NRL match with scores, key stats and team details — from the current round all the way back to 2015. Use it to review how any round played out and compare actual results against model predictions.

Full History

Browse results from any season and round going back to 2015 — all stored in our local database.

Scores & Margins

Each match shows the final score, winning margin, and which team was home vs away.

Cross-Reference

Use alongside the Predictor tab to see how the model's predictions compared to actual outcomes.

Season

Round

What is an ELO Rating?

ELO is a self-correcting rating system originally developed for chess and now widely used in sport analytics. After every match, points are transferred from the losing team to the winner — with the size of the transfer depending on how surprising the result was. Beating a strong opponent earns more points than beating a weak one.

Starting at 1000

Every team begins at 1000. Ratings above 1000 indicate above-average strength; below 1000 is below average.

Margin Weighted

Winning by 30 moves the rating more than winning by 2. Dominant teams climb faster.

Historical Chart

The line chart tracks every team's ELO progression from 2015 to today — revealing dynasties and declines.

Feeds Predictions

ELO ratings are one of the key inputs into our ML prediction models, capturing current team strength.

Current Rankings

Historical Performance

Season Performance (Round by Round)

Select Season:

* ELO ratings are calculated based on match results, opponent strength, and point margins. The history chart shows the progression of each team's rating over the seasons.

How does the Match Predictor work?

The Match Predictor runs every upcoming fixture through our ensemble of 7 machine learning models — XGBoost, LightGBM, CatBoost, Logistic Regression, Random Forest, and two blended composites. Each model independently analyses rolling form, ELO strength, scoring patterns, and rest days to produce a win probability and predicted score margin.

Win Probability

Each model outputs a probability (0–100%) that the home team wins, based purely on statistical inputs — no opinion.

Predicted Margin

A dedicated margin regression model predicts the point difference — positive means home team wins by that many.

Blended Model

The Blended result combines XGBoost and Logistic Regression outputs (60/40 weight) for the most stable prediction.

Contextual Factors

Each prediction card also shows supporting context: recent form, H2H record, venue stats, and squad tiers.

Predictor Component Breakdown

Game Summary Card

Shows match metadata (venue, local kickoff, round) alongside the models' consensus margin, home/away win rates, and a custom ELO-based handicap line.

Simulated Markets & Live Feed

Simulates market odds derived from model win probabilities, compares them against real-time bookmaker lines to flag value discrepancies, and overlays the live news sentiment stream for the two clubs.

Team List Card

Displays official team lineups side-by-side with individual player rankings, tactical positions, and color-coded ELO impact ratings to contrast depth and talent.

Performance Rank (e.g. A+, B): Raw statistics-based grade (PIR) compared to other starters in their positional group.

ELO Impact Tier (e.g. Elite, Solid): Roster tier reflecting their strategic impact on predictions (Gold = Elite, Green = High, Blue = Solid, Gray = Depth).

Recent Form Analysis

Tracks both teams' actual scores, margins, and ELO rating shifts over the last 5 rounds. Explores scoring averages, points conceded, and opponent difficulty levels to measure momentum.

Last 10 Head-to-Head

Details the win-loss split, average points, and match-by-match historical scorecards for the last 10 encounters between the two clubs.

Model Decision Drivers

Exposes the SHAP-based feature importance weights showing which top 5 factors (e.g. defense ELO, travel distance, rest days, line breaks) most heavily influenced the AI's final match prediction.

Head-to-Head Rolling Averages (Last 5 Games)

Plots a visual historical trend of statistical averages (runs, tackles, line breaks, completion rates) during matches played between these two clubs, illustrating historical playstyle matches and tactical superiorities.

Season

Round

Model Weight Distribution

XGBoost 70%

CatBoost 0%

LightGBM 5%

Random Forest 20%

Current Mix: 70% XGBoost / 5% Logistic Regression / 0% CatBoost / 5% LightGBM / 20% Random Forest

What is Model Analytics?

Model Analytics evaluates how each of our ML models would have performed as a betting system — backtested against real bookmaker closing odds from 2015 to today. It answers the critical question: "If I had followed this model's tips, what would my return on investment be?"

Four Markets

Performance is tracked across Head-to-Head, Line (handicap), Totals (over/under), and Margin bucket bets.

Real Odds

Profits are calculated using Pinnacle, bet365, and BlueBet closing prices — not theoretical fair odds.

W/L & Profit

Each cell shows wins–losses, win percentage, and total profit in units (1 unit = 1 bet of equal stake).

Season Trend

Scroll down to the Season Breakdown table to see year-by-year performance for any model.

Model Analytics

Evaluate historical model performance across different timeframes.

OPEN SOURCE ACKNOWLEDGEMENTS

Ryno's Edge relies on several incredible open-source libraries and mathematical models to power our predictive analytics. We gratefully acknowledge the developers and contributors of the following software:

XGBoost

Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with the License. You may obtain a copy of the License at: http://www.apache.org/licenses/LICENSE-2.0

Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.

CatBoost

LightGBM

Licensed under the MIT License.

Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.

scikit-learn (Random Forest & Logistic Regression)

Licensed under the 3-Clause BSD License.

Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met:

Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer.

Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution.

Neither the name of the copyright holder nor the names of its contributors may be used to endorse or promote products derived from this software without specific prior written permission.

What is the Knowledge Hub?

The Knowledge Hub serves as a central library explaining the models, data sources, and methodologies used throughout Ryno's Edge. Select any topic below to learn how each algorithm operates and how to read key sports betting concepts.

1. Betting Guides

Understand the structure of standard sports betting markets, including Head-to-Head, Line, and Totals.

2. Model Details

Dive into our individual machine learning models: XGBoost, LightGBM, CatBoost, and others.

3. Technical Concepts

Learn about ELO ratings, feature engineering, cross-validation, and target metric selections.

4. Interactive Reading

Use the buttons below to open article overlays with in-depth descriptions, configuration stats, and guides.

KNOWLEDGE HUB

Betting Guide

Master the basics of H2H, Line, and Over/Under markets in the NRL.

Model Insights

Understand the math behind the Blended Model and ELO ratings.

NRL Rules

Essential rules, interchange laws, and scoring mechanics explained.

Edge Index Model

Statistical Normalization & Probability

1. The Concept of Normalization

The Edge Index model doesn't just look at raw numbers; it looks at Relative Performance. We compare every team's stats against the rolling League Average to account for seasonal trends (e.g., higher scoring in dry tracks).

Normalized Score

Team Stat / League Average

Recent Form Window

Rolling average of the last $N$ games

2. Pythagorean Expectation (Exponent 8)

To convert normalized scores into a win probability, we use a modified Pythagorean Expectation formula. While MLB uses an exponent of 1.83, the high scoring and variability of the NRL require an exponent of 8 to achieve maximum predictive accuracy.

Win % = Home⁸ / (Home⁸ + Away⁸)

Where "Home" and "Away" are normalized power scores

3. The 4-Factor Advanced Formula

To reach 62% predictive accuracy, the model balances four distinct aspects of the game:

Core Performance (Slider Weight) Custom

Territorial Pressure (FDO, Kick Metres) 15%

Discipline & Efficiency (Errors, Penalties) 15%

Positional Influence (Spine, Engine Room) 10%

Ensemble Integration

The Edge Index provides the "Base Logic" for the Ryno's Edge system. It is blended with XGBoost and LightGBM outputs to form the final prediction you see on the dashboard.

Global Ensemble

Blending Independent Models into a Single Prediction

1. What is an Ensemble?

No single machine learning model is perfect. Each algorithm sees the data through a different lens, finding patterns the others might miss. An Ensemble combines the outputs of multiple independent models into a final, shared prediction — similar to the concept of a "wisdom of the crowd."

Reduces Variance

When models disagree, the average is more stable than any single model's guess.

Reduces Bias

Different algorithms capture different patterns, limiting the collective blind spots.

2. The Model Stack

The Ryno's Edge Blended Model uses five independent classifiers, each trained on historical match data and player performance stats:

XGBoost Gradient boosted trees. The workhorse — 70% Blended Weight due to its superior 91.78% training accuracy.

Rand Forest Aggregates many uncorrelated decision trees. Great at detecting non-linear feature interactions. 20% Blended Weight.

LGBM Leaf-wise gradient boosting. Fast and efficient on high-dimensional data. 5% Blended Weight.

CatBoost Symmetric trees with ordered boosting. Robust to categorical features and noise. 0% Blended Weight (used for Margin/Total only).

Log. Reg. Logistic Regression. A simple linear baseline that helps prevent overfitting. 5% Blended Weight.

3. Weighted Average Blending

Each model outputs a win probability between 0–100%. These are combined using a weighted average. Weights were determined via a grid search to maximize back-tested historical accuracy:

Final % = (XGB × 0.70) + (RF × 0.20) + (LR × 0.05) + (LGBM × 0.05)

Grid-search optimized weights — back-tested to 91.78% accuracy

4. Margin & Total Predictions

Beyond win probability, the ensemble also provides separate regression models that predict the winning margin in points and the total match score. These use the same features but are trained as regression tasks (not classification), with their own ensemble of models including CatBoost and LightGBM regressors.

Pro Tip

You can adjust the weight of each model using the sliders in the Match Predictor tab. If you believe XGBoost is overfit for the current NRL season, reduce its weight and redistribute to Random Forest or LGBM.

Betting Guide

Mastering the NRL Markets

1. Understanding the 'Edge'

"The Edge" is the core metric of this system. It represents the difference between the Model's Estimated Probability and the Market's Implied Probability.

Implied Probability = 1 / Decimal Odds

If the market offers odds of $2.00 (50% implied), but our model predicts a 65% chance of winning, you have a 15% Edge. This is where value is found.

2. Market Types Explained

Head-to-Head (H2H)

Simply picking the winner of the match. Our primary focus for probability analysis.

Line / Handicap

Betting on a team with a points head-start. E.g., Panthers -6.5 means they must win by 7+.

Over/Under

Predicting whether the total points scored by both teams will be above or below a set number.

3. Strategy & Bankroll Management

Never bet more than you can afford to lose. We recommend using a Unit System where 1 unit equals 1% of your total bankroll.

The Banker (85%+)

4-5 Units

Highest probability logic. Peak confidence. These represent the strongest value propositions found by the blended models.

High Conviction (75-84%)

2-3 Units

Strong statistical evidence. High historical win rate. Reliable core plays for a long-term strategy.

Strong Play (65-74%)

1-2 Units

Reliable edge with consistent value. These plays form the foundation of a disciplined betting approach.

Solid Lean (55-64%)

0.5-1 Unit

Favorable matchup with higher expected variance. Good for filling out a portfolio with smaller stakes.

Coin Flip (<55%)

Caution / Pass

Speculative or high-risk scenarios. Low statistical edge. Best used for small speculative plays or total avoids.

Pro Tip: The Kelly Criterion Explained

The **Kelly Criterion** is a mathematical formula used by professional bettors to determine the optimal size of a bet. Its goal is simple: maximize the long-term growth of your bankroll while minimizing the risk of losing everything.

The "Edge" Concept

To use Kelly, you first need to find your **Edge**. This is the difference between what you think the probability is vs. what the bookmaker thinks.

Example: If bookie odds are $2.00 (implied 50% chance), but our model predicts a **55%** chance, you have a **5% Edge**.

Real-World Scenario ($1,000 Bankroll)

Using our **5% Edge** example from above, here is how a professional would determine their bet size:

$50.00

Full Kelly (5%)

Mathematically optimal but very aggressive. High risk of large swings.

$25.00

Half Kelly (2.5%)

The "Aggressive Standard". Balances growth and safety.

SAFE

$12.50

Quarter Kelly (1.25%)

The "Professional Standard". Highly safe for long-term survival.

The ratings in this dashboard (Banker, High Conviction, etc.) use a simplified version of this logic to suggest unit sizes that keep your bankroll safe while still capitalizing on high-edge opportunities.

XGBoost Model Deep Dive

Technical Documentation & Insights

How XGBoost Works

XGBoost (Extreme Gradient Boosting) is an ensemble learning method that combines the predictions of multiple "weak" decision trees to form a single "strong" predictor. It works through a process called Boosting, where each new tree is specifically designed to correct the prediction errors of all previous trees.

The Process

1. Starting with a base prediction (usually 0.5 win probability).
2. Building a decision tree to predict the "Residuals" (errors).
3. Adding the tree's output to the tally with a "Learning Rate" penalty.
4. Repeating this 300+ times to refine the final probability.

Statistical Inputs (The "Features")

The model analyzes Rolling 5-Game Averages for both teams. This ensures it focuses on current form and momentum rather than historical reputation.

Attack Stats

Tries, Linebreaks, Tackle Busts, Offloads, Run Metres, Post-Contact Metres, Kick Return Metres.

Defence & Discipline

Tackles Made, Missed Tackles, Errors, Penalties Conceded, Tries Conceded.

Model Config

Total Trees 300

Max Depth 5

Learning Rate 0.05

Subsample 80%

Did You Know?

XGBoost handles "missing data" automatically. If a team's missed tackles aren't recorded, it uses its learned historical patterns to estimate the impact.

Random Forest Analysis

Ensemble Learning & Bagging Logic

How Random Forest Predicts

Random Forest is an ensemble learning method that serves as a "Wisdom of the Crowd." Unlike a single decision tree which can be biased, Random Forest grows hundreds of different trees and lets them "vote" on the outcome. This process is called Bootstrap Aggregating (Bagging).

The Logic

1. Sampling: Creating random subsets of the match data.
2. Feature Shuffling: Selecting random stats for each tree to consider.
3. Voting: Each of the 300 trees makes a win/loss prediction.
4. Consensus: The final percentage is the ratio of trees that picked the winner.

Classification vs Regression

While Random Forest is elite at predicting Win Probabilities (Classification), it can sometimes produce inconsistent results for continuous variables like scores (Regression).

Optimization Note

To ensure the highest accuracy in our Blended Model, we have suppressed the Margin and Game Total outputs from the Random Forest model. This forces the consensus to lean on more robust regressors like XGBoost and LightGBM for score predictions, while still utilizing RF's strength in picking winners.

Model Config

Estimators 300

Max Depth 10

Random State 42

Min Samples 2

Key Features

The model heavily weights ELO Difference and Post-Contact Metres. It excels at identifying when a dominant forward pack will eventually break down the opposition.

LightGBM Deep Dive

Gradient Boosting & Leaf-wise Efficiency

How LightGBM Works

LightGBM is a high-performance gradient boosting framework developed by Microsoft. It is designed to be fast, distributed, and efficient. While it shares the "Boosting" DNA with XGBoost, it introduces several innovations that make it ideal for the high-dimensional data used in NRL prediction.

Technical Innovations

1. Leaf-wise Growth: Growing trees by splitting the leaf that reduces loss most, rather than level-by-level.
2. GOSS: Retaining training instances with large gradients while randomly sampling those with small gradients.
3. EFB: Bundling mutually exclusive features (like team IDs) to reduce the number of features.
4. Histogram Learning: Bucketing continuous stats into discrete bins for near-instant split finding.

LGBM in the Blended Model

In our NRL predictor, LightGBM acts as the "Speed Demon." It processes the thousands of data points from our Rolling 5-Game Averages faster than any other model, making it the primary model we use for high-frequency updates or real-time odds simulations.

Handling Complexity

LightGBM excels at finding subtle interactions between stats (e.g., how "Post-Contact Metres" impact "Missing Tackles" later in the game).

Memory Efficiency

Its histogram-based approach uses less memory, allowing us to train deeper models on more seasons of historical data.

Model Config

Total Trees 500

Num Leaves 31

Learning Rate 0.05

Subsample 80%

Speed Advantage

While XGBoost takes roughly 15-20 seconds to train on our full dataset, LightGBM often completes the same task in under 5 seconds with comparable accuracy.

CatBoost Model Deep Dive

Categorical Boosting & Symmetric Trees

The CatBoost Edge

Developed by Yandex, CatBoost (Categorical Boosting) is a powerhouse specifically optimized for datasets with many categories—like NRL team names, match venues, and weather conditions. It is the only model in our ensemble that handles these labels natively without losing information.

Core Principles

1. Symmetric Trees: Building balanced decision trees where every node at a level uses the same split condition.
2. Ordered Boosting: A unique training method that prevents "target leakage" by strictly respecting the data sequence.
3. Native Categorical Support: Converting team and venue names into high-precision vectors using advanced Target Statistics.

Why We Use MAE for CatBoost

For our score predictions, we configured CatBoost to optimize for Mean Absolute Error (MAE). Unlike standard models that try to minimize distance to "perfect" scores, MAE models prioritize being "roughly right" most of the time.

Performance Note

By using MAE, CatBoost doesn't get "overwhelmed" by massive 50-point scorelines that occur once or twice a season. Instead, it maintains its focus on predicting the competitive 1-12 point margins common in top-tier NRL matches.

Model Config

Iterations 500

Tree Depth 6

Loss Func MAE

Random Seed 42

Inference Speed

Due to its symmetric tree structure, CatBoost can generate predictions for a full NRL round in under 10 milliseconds, making it the fastest live-inferencing model in the group.

What is an Inside Edge?

An "Inside Edge" represents a mathematical advantage over the sports betting market. It is identified when our ensemble of machine learning models projects a higher likelihood of an outcome occurring than the odds offered by bookmakers imply. By strictly targeting these mispricings, you shift the mathematical advantage back in your favor, mimicking the strategies used by quantitative trading firms.

1. Implied Probability

Every bookmaker price translates to a percentage chance. For instance, odds of $2.00 imply a 50% chance. We convert bookmaker odds back into implied probability to establish the baseline.

2. Model Probability

Our advanced machine learning models (XGBoost, LightGBM, CatBoost, etc.) analyze hundreds of features—such as team form, weather, rested days, and player tiers—to compute a purely statistical, unbiased probability.

3. The Edge %

The subtraction of Implied Probability from Model Probability. If our model projects a team has a 60% chance to win, but the market odds imply a 50% chance, we have identified a +10% Edge.

4. Expected Value (EV)

The expected percentage return on investment. Calculated as: (Model Prob * Market Price) - 1 A positive EV (e.g. +7.5%) indicates that if you made this bet 1,000 times, you would average a 7.5% return.

Statistical Safety Guard: To reduce variance, the dashboard only shows high-confidence edges with a Model Probability ≥ 40% and Expected Value ≥ 7.5%.

Optimized Top 10 Active Edges

Inside Edge

Top 10 Positive Betting Margins vs Market Price

Target Model

Target Round

—

Bets Found

Max Edge

—

Rank	Match Context	Market	Selection	Market Price	Model Odds	Model Prob	Edge %	EV %

Model Prob

The winning probability assigned by the selected AI model. This represents the consensus or specific engine's view of the outcome's likelihood based on historical data.

Edge %

The mathematical gap between our Model Probability and the Market's implied probability. A higher positive edge indicates a potential market mispricing.

EV %

Expected Value (EV) measures the projected ROI over time. Calculated as (Model Prob * Market Price) - 1. Positive EV indicates a long-term "value" bet.

What are Squad Intelligence Tiers?

Squad Intelligence Tiers rank NRL teams by the collective quality and depth of their roster, breaking down squad values into distinct tiers. Rather than looking only at ladder position, this system uses player-level performance models to evaluate the true potential, depth, and vulnerability of each team's current line-up.

1. Tier Distribution

Players are classified from Elite (Tier 1) to backup/fringe (Tier 4), mapping out how top-heavy or deep each squad is.

2. Roster Value

An aggregate numeric rating based on active player tiers, allowing quick visual comparison of overall team strengths.

3. Injury Impact

Instantly highlights when a team's key contributors (Tier 1 & 2 players) are sidelined, affecting match predictor weights.

4. Predictive Utility

Roster value differences directly adjust team ratings in our ML models, serving as a primary driver of score projections.

Reading Player Cards

Performance Rank (A+, A, B, C, D)

A statistics-based grade derived from our Player Impact Rating (PIR) formula. PIR evaluates tries, try assists, line breaks, tackle busts, run metres, tackles made, missed tackles, and errors. Players are ranked against others in the same positional group (e.g. all Fullbacks, all Props) and graded on a percentile curve.

ELO Impact Tier (Elite, High, Solid, Depth)

The strategic tier reflecting how much a player's presence or absence swings our prediction models. Gold = Elite (Tier 1), Green = High (Tier 2), Blue = Solid (Tier 3), Gray = Depth (Tier 4).

Form Trend Arrows (▲ ▼)

Compares a player's PIR across their last 3 games against their season-wide average. ▲ Green = recent form is 15%+ above their average. ▼ Red = recent form is 15%+ below. No arrow means they are performing at their expected level.

Squad Intelligence Tiers

Live Roster Power Rankings & Player Value Analysis

Initializing Squad Database...

What is the Tactical News Hub?

The Tactical News Hub automatically aggregates and monitors news reports, press conferences, and injury lists. By applying advanced sentiment analysis, it evaluates team readiness and translates qualitative information into statistical factors for the predictor model.

1. Injury Intel

Tracks and classifies player injury statuses, recovery timelines, and late fitness tests to determine roster viability.

2. Sentiment Score

Computes a sentiment ratio based on media coverage, coach statements, and camp reports to gauge team morale.

3. Betting Impact

Translates tactical news directly into betting impacts — indicating if an event changes the line, totals, or H2H value.

4. Injury & Tactical Alerts

Highlights breaking, high-impact alerts that indicate major player lineup changes, injuries, suspensions, or late-mail shifts.

Tactical News Hub

AI-Powered Injury Analysis & Betting Impact Reports

Latest Tactical News Feed

Scanning Global News Sources...

Casualty Ward by Club

Loading Casualty Ward...

ADMIN PORTAL

User Management & System Access

Require Auth

User Administration

Auto Refinement

Idle

The system automatically syncs new round results and retrains the ML models on a 12-hour background loop.

Trained: ...

Scraped: ...

Refinement Console Log

// Waiting for refinement logs...

Open Source Acknowledgements

Comments, Suggestions and Feedback? Please email info@hamblion.com

Authentication

RYNO'S EDGE NRL EDITION

Welcome to Ryno's Edge

ML Predictions

ELO Rankings

News Sentiment

Player Stats

Recent Results (Round ...) WRAPPED UP

Round Schedule VIEW ALL

Power Rankings ELO DATA

The Edge Pick

System Matrix Sync Status

What is the Player Leaderboard?

Choose Your Stat

Best Game vs Season

Multi-Season

Position Filter

What are Season Trends?

Round by Round

Toggle Teams

Avg vs Total

Any Stat

What is a Player Profile?

Career Stats

Recent Form

Search Any Player

What is Round Results?

Full History

Scores & Margins

Cross-Reference

What is an ELO Rating?

Starting at 1000

Margin Weighted

Historical Chart

Feeds Predictions

Current Rankings

Historical Performance

Season Performance (Round by Round)

How does the Match Predictor work?

Win Probability

Predicted Margin

Blended Model

Contextual Factors

Predictor Component Breakdown

Game Summary Card

Simulated Markets & Live Feed

Team List Card

Recent Form Analysis

Last 10 Head-to-Head

Model Decision Drivers

Head-to-Head Rolling Averages (Last 5 Games)

Stat Weights Remaining: 100%

What is Model Analytics?

Four Markets

Real Odds

W/L & Profit

Season Trend

Model Analytics

OPEN SOURCE ACKNOWLEDGEMENTS

XGBoost

CatBoost

LightGBM

scikit-learn (Random Forest & Logistic Regression)

What is the Knowledge Hub?

1. Betting Guides

2. Model Details

3. Technical Concepts

4. Interactive Reading

KNOWLEDGE HUB

Betting Guide

Model Insights

NRL Rules

Edge Index Model

1. The Concept of Normalization

2. Pythagorean Expectation (Exponent 8)

3. The 4-Factor Advanced Formula

Global Ensemble

1. What is an Ensemble?

2. The Model Stack

3. Weighted Average Blending

Recent Results (Round ...)

Round Schedule

Power Rankings