N NYCE

Live Dashboard

Model Active

⚾

NYCE MLB Mode Activated 🔥

Today | Vegas Feed Live

Ensemble Overview & Daily Execution

Full ensemble active: Bayesian NB Monte Carlo (10k) + XGBoost Scikit-Learn Hybrid (L20) + Mack wRC+ + Strikeout Regression + Yang-Swartz Two-Stage MCMC Weights + Victory Bombs Statcast Framework.

⚡ Value Discovery Terminal

Auto-syncing w/ Backend

Running Poisson Math & Line Shopping...

Live Market Grid

The Odds API Integrated

Backend Protected Feed

Odds API data now routes through FastAPI

Provider keys stay server-side. The Live Slate and Prop Engine read the normalized backend payload instead of exposing browser secrets.

Local R Sync

🖥️ RStudio Bayesian Output

Pick a live slate matchup to write R inputs, then sync the exported model output.

R Model Matchup

Waiting for R export

Pitchers pending

Projected Winner

Away Win Prob

--%

Home Win Prob

--%

Away Total

Home Total

Waiting for RStudio graph export

#1 POTD

#1 Algorithmic Edge

Awaiting Sim... Pending

--%

MC Prob

Odds

Score --

+EV Edge

--%

Prop Read

Market N/A | Model -- | Fair --

Confidence

True Math Analysis: Awaiting execution of backend logic parameters... Expert Analysis Thoughts: Awaiting NYCE AI brain context...

Top 3 ML Parlay

🔥 Top 3 ML Edges

-- Pending

NYCE Parlay of the Day

Model-Driven +EV Slip

Pregame odds only

Building a model-driven Parlay of the Day from current priced NYCE edges...

📊 Live MLB Slate

Fetching Live Vegas Intel...

Matchup & Form	Arena / Time	Market vs True Odds	Total (O/U)	AI Analyst

📰 LIVE MLB NEWS & SOCIAL

DUAL-SYNC

Loading Official MLB News...

MLB Prop Engine

Player Props Command Board

Backend-fed prop picks, model grading, and saved parlay tracking. No browser-side sportsbook keys.

Loading backend Prop Engine...

#1 MLB Prop

Awaiting data

-- --

Parlay of the Day

Top Moneyline

Awaiting slate

Tracker

Open

Cash

Saved

🆕 New Today 0 props Results idle

		LU #	Batter Stats	VS Pitcher	Platoon	BvP / Bullpen		Form L5/L10/L20L5L10L20				Rating		Consensus Diff	EV %	Model Diff		Park & WX	Opp vs Stat	Add

Prop Card Glossary

Reading the card fast

Gray = Waiting Yellow = Live Green = Cashed Red = Miss

Good Prop

Look for green Power or Contact, ML Model above the break-even zone, positive Real Edge, and a Bounce-Back Regression or Hot Form Confirmed signal.

Elite Prop

Best cards have model agreement, API Odds or Statistical support, strong L5/L10/L20 context, and no major risk badge fighting the pick.

Trap / Risk

Be careful when Power or Contact is red, the card shows Cool-Down Risk or Risky Regression, Real Edge is weak, or the projection leans on model-only data with no live book confirmation.

Power 70%+ Contact 65%+ Statcast Bat Track ML Model Statistical API Odds

🔍 Live Player Props Search Engine

Player Name

🎯 Pitcher Strikeout (E[K]) Engine

Calculates true bat-missing talent by regressing Swinging Strike% and Foul Strike%, adjusting for opponent K-rate and projected IP.

SwStr%

16.4%

Opp K Adj

+1.12

True Expected K's (E[K])

7.42

📈 wRC+ Lineup Strength Module

Isolates handedness splits to calculate the exact Adjusted Batter Runs (AdjBat) for today's specific 9 starting hitters.

Opp SP Arm

LHP

Lineup wRC+ v L

114

Dynamic Run Modifier

+0.68 Runs

⛽ Odds API Usage

Loading usage…

—

est. credits used today

NYCE Analyzer

Analyze & Optimize any pick, list, or ticket

Type a player or a list, upload a player photo, or drop a FanDuel / DraftKings / PrizePicks slip. NYCE grades every pick against its own models — model probability, edge, EV, model points, confidence — then builds the best parlay with Kelly stake and a Monte-Carlo win%. Save to log picks to the Accuracy tracker.

Player name or list (one per line — optional "Name - market")

Upload image or paste / drop a screenshot (Ctrl+V)

Recent analyses

Loading…

Nyce Recom Bot

Model Recommendation Feed

Best current Prop Engine ideas ranked by model probability, real edge, and analyst explanation coverage.

Loading Nyce Recom Bot...

Safe Workbook Bridge

Master Model Sync

SQLite is the shared brain. The master workbook imports model weights and seed tables into the backend, and exports create a new live copy with current props, parlays, and tracker data.

Status

Checking...

Source

Imported Rows

MLB Workbook Signal

Public team-power ingest for moneyline and NYCE Recom context.

Status

Checking...

Last Refresh

Assets

Rows Imported

Pitcher Prop Workbook Bridge

Super Model v5 Sync

NYCE_PRO_MLB_Super_Model_v5.0 imports pitcher prop hub rows, AI agent rules, data-quality checks, attack/defense ratings, and final pitcher ERA into SQLite. The app reads it only; your upgraded XLSX stays untouched.

Status

Checking...

Source

Imported Rows

🤖 XGBoost & Scikit-Learn Hybrid Engine

Python Integration Active

The NYCE Ensemble now merges Bayesian priors with a non-linear Extreme Gradient Boosting (XGBoost) model. Trained via Python/Retrosheet, this module analyzes L20 (Last 20 Games) rolling stats to capture current momentum, bullpen fatigue, and hidden conditional variables.

Holdout AUC

Brier Score

Log Loss

Top XGBoost Features

Loading model metrics...

Relational MPNN Challenger

Pitch Decision Graph Model

Loading shadow model status...

Shadow

Live Weight

Top-1

Brier

Gate

Research

> import pandas as pd > import xgboost as xgb > from sklearn.metrics import mean_squared_error > > # Loading L20 Rolling Retrosheet Data > data = load_mlb_rolling_stats(window=20) > model = xgb.XGBRegressor(objective='reg:squarederror', n_estimators=1000) > model.fit(X_train, y_train) > > # Ensemble execution bridging JS Frontend > print(f"Model RMSE: {mean_squared_error(y_test, preds, squared=False):.2f}") [SYSTEM] XGBoost L20 Projections successfully synced to NYCE Engine.

The 5-Core Ensemble Architecture

Master LogReg Engine

📐 1. Pythagorean GSSD

Uses current and historical run differentials to predict expected win rates based on Bill James' Pythagorean theorem of baseball, adjusted for Game Score Standard Deviation.

⚾ 2. Starting Pitcher BTM

Batters Faced to Missing. Evaluates true starting pitcher talent against the expected lineup, prioritizing swing-and-miss stuff (SwStr%) over traditional variance-heavy ERA.

🏃 3. BaseRuns (TOOR)

Estimates how many runs a team *should* have scored/allowed based on underlying offensive events (singles, walks, HRs), eliminating sequence luck and clustering noise.

📊 4. Bayesian NB Clusters

Predicts the run environment applying a Negative Binomial distribution. Integrates the Yang-Swartz Two-Stage MCMC to dynamically draw parameter uncertainty weights.

⚔️ 5. wRC+/FIP Module

A matchup-specific adjustment comparing the 9-man offensive lineup's weighted Runs Created (wRC+) against that specific pitcher's FIP and handedness splits.

🧠 Master Ensemble

A Logistic Regression (LogReg) aggregator that ingests the 5 independent component outputs, weighting them dynamically to output the final Vegas-beating Win Probability.

⚡ Negative Binomial Deep-Sim Sandbox

Yang-Swartz MCMC Two-Stage

Select parameters below to execute a live, 10,000-iteration Monte Carlo simulation based natively on the Bayesian Log-Linear Run Cluster architecture.

Away Team

Offense (wRC+ Base)

Starting Pitcher (Run Sup)

Home Team

Offense (wRC+ Base)

Starting Pitcher (Run Sup)

🎯 OPal Prop Monte Carlo Simulator

Stuff+ · xCSW · Exit Velo

Pick a player prop, set the line and price, then dial the three drivers that move the model. Each run executes a real per-at-bat Monte Carlo (10,000 games) and prices the result against the book to surface the OPal Win Probability and Edge EV%. Plays clearing +5% EV are flagged.

Prop Market

Line

Odds & Side

Pitcher Stuff+ 100

100 = league average · higher suppresses batters

Pitcher xCSW% 28

Called-strike + whiff · the strikeout driver

Batter Exit Velo 89

Avg mph · drives power (HR ≫ TB > hits)

📏 Prove-Edge Report — Model vs Market

loading…

Every archived pick is graded against real outcomes into a permanent ledger, then scored the way forecasters are judged: calibration first, then hit-rate and ROI. Buckets under n=20 are flagged and never drive verdicts. Model-only picks carry fair odds, so their ROI is definitional — read their calibration.

Calibration — predicted vs realized (props)

Band	n	Predicted	Realized	Gap

By market — record, calibration, ROI at archived odds

Market	W-L	Hit%	Pred%	ROI

2026 Platoon Split Model

Stevenson's update proves lateral pitch movement generates severe platoon splits.

The 5-Model Ensemble "Reality Check"

Comparing Out-Sample individual model returns vs. the Master LogReg Aggregator.

3-Year Sample

Model Component	Brier Score	Log Loss	Target EV+ Cutoff	Out-Sample ROI
🏛️ Vegas Closing Line	0.245	0.691	N/A	-4.54% (Vig)
1. Pythagorean GSSD	0.241	0.685	> 2.0%	+2.12%
2. Starting Pitcher BTM	0.240	0.680	> 2.5%	+1.85%
3. BaseRuns (TOOR)	0.238	0.672	> 2.5%	+3.40%
4. Bayesian NB Run Clusters	0.221	0.641	> 3.0%	+4.51%
5. wRC+/FIP Component	0.218	0.628	> 3.5%	+5.12%
🌟 Master Ensemble (LogReg)	0.187	0.593	> 4.5%	+8.24%

Log Loss Progression (Lower is Better)

NYCE Model

🎯 Beat the Streak

Tiers: ⚡ Elite 72%+ 💎 Strong 65–72% 🔥 Good 58–65% 📊 Fair 50–58% ⬜ Weak <50%

🔮 This is a Prediction Board

The percentages show how likely each player is to get at least 1 hit today — based on Statcast data, park factors, and pitcher matchups. Games haven't been played yet. Results appear after each game finishes.

🃏 Reading a Player Card

75.4% — Model's hit probability for today
72–79% — Confidence range
Confirmed #1 — Batting order spot (in lineup, not a hit)
xBA — Expected batting average from Statcast
Pit× — Pitcher matchup multiplier
Park — Ballpark environment multiplier
E[AB] — Expected at-bats today

🎯 How to Play Beat the Streak

Click any player card to select them as Pick 1 or Pick 2
Click 🔒 Lock In to confirm your picks for today
After games finish, click ↺ Check Results
Both players must get a hit to extend your streak 🔥

✅ Hit confirmed ❌ No hit DNP Did not play ⏳ Game pending

📘 Stats Glossary

xBA: expected batting average, the hitter's baseline chance of batted balls becoming hits.

Pit×: pitcher multiplier. Above 1.00 helps the hitter; below 1.00 means the pitcher suppresses hits.

Park: venue multiplier for how the stadium plays for hits today.

E[AB]: expected at-bats, usually around 3.9 for a projected starter.

Hit%: baseline hit chance using xBA, pitcher, park, and expected at-bats.

Top % / Range: final model probability plus confidence range after all adjustments.

L5 / L20: recent hit rates over the last 5 or 20 logged games when available.

Platoon / Wind: matchup and weather edges when the backend supplies them.

🎯

My BTS Picks — Today

Click any player card below to add to your picks (max 2)

Day Streak 🔥

Pick 1 — click a card below

Pick 2 — click a card below

Scoring today's lineup…

✨ NYCE AI: Matchup Deep Dive

🔴 Game Center

Matchup Overview

Probable Pitchers

Game Details

Matchup Stats

Fun Facts

✨ NYCE AI & Monte Carlo Deep Dive

💬 Ask NYCE AI Expert

📰 Post Details

NYCE MLB Mode Activated 🔥

Ensemble Overview & Daily Execution

⚡ Value Discovery Terminal

Live Market Grid

Odds API data now routes through FastAPI

🖥️ RStudio Bayesian Output

Awaiting Sim... Pending +

🔥 Top 3 ML Edges

Model-Driven +EV Slip

📊 Live MLB Slate

📰 LIVE MLB NEWS & SOCIAL

Player Props Command Board

Awaiting data

Awaiting slate

Reading the card fast

🔍 Live Player Props Search Engine

Name

🎯 Pitcher Strikeout (E[K]) Engine

📈 wRC+ Lineup Strength Module

Analyze & Optimize any pick, list, or ticket

Model Recommendation Feed

Master Model Sync

Super Model v5 Sync

🤖 XGBoost & Scikit-Learn Hybrid Engine

The 5-Core Ensemble Architecture

📐 1. Pythagorean GSSD

⚾ 2. Starting Pitcher BTM

🏃 3. BaseRuns (TOOR)

📊 4. Bayesian NB Clusters

⚔️ 5. wRC+/FIP Module

🧠 Master Ensemble

⚡ Negative Binomial Deep-Sim Sandbox

Away Team

Home Team

🎯 OPal Prop Monte Carlo Simulator

📏 Prove-Edge Report — Model vs Market

2026 Platoon Split Model

The 5-Model Ensemble "Reality Check"

Log Loss Progression (Lower is Better)

🎯 Beat the Streak

Awaiting Sim... Pending