How Prediction Engine Works
Machine learning models, expected value, and bankroll strategy — everything you need to make informed betting decisions.
What is Prediction Engine?
Prediction Engine is a quantitative sports analysis platform that generates daily NFL, MLB, NBA and NHL predictions using machine learning. Instead of relying on gut feelings, talking heads, or social media hype, every pick on the platform is produced by trained statistical models that analyze hundreds of features from historical data.
The platform covers team-level predictions (moneyline, spread/run line, totals), player props (hits, runs, RBIs, strikeouts, points, assists, rebounds, double-doubles, triple-doubles), and live in-game trackers that update predictions every 30-60 seconds as games unfold.
Every prediction includes the model's probability (with the market price next to it for reference where a book posts one), and every result is graded on the performance page — so you can see how much conviction the model has and how it has performed.
How Our Models Work
All models are built on XGBoost (gradient-boosted decision trees) and Random Forest classifiers/regressors — the same algorithms used across finance, healthcare, and competitive machine learning. We don't use deep learning or black boxes. Every model is interpretable and auditable.
MLB Models (3 Team + 6 Player Props + 6 Live)
Our MLB pipeline uses three separate team models — one each for moneyline (win/loss classification), run line (margin regression), and totals (over/under regression). Each model is trained on 42-44 engineered features including pitching matchups, bullpen usage, park factors, umpire tendencies, and rolling performance windows.
Player prop models cover five batter statistics (hits, runs, RBIs, home runs, combined HRR) and pitcher strikeouts. The pitcher K model gives the over/under probability for every starter's strikeout line, graded daily on the performance page.
The Live K Tracker runs four XGBoost checkpoint models (innings 2 through 5) that predict a pitcher's final strikeout total from in-game data — pace, pitch count, K rate, and game context. The Live Hits Tracker uses 59 features including park factors, platoon splits, umpire tendencies, and weather to predict remaining hits for batters mid-game.
NBA Models (5 Team + 5 Player Props + 9 Live)
From 2026-27 the NBA moneyline comes from Elo ratings built from results plus a preseason roster adjustment (a play when Elo gives its side 70% or more); the previous Random Forest moneyline went 156-41 (79.2%) from March 12 to June 13, 2026, playoffs included. Spread predictions use regression to estimate winning margin, and the totals model trains daily on rolling data from three seasons.
Player props cover points, assists, rebounds, double-doubles, and triple-doubles. The points and assists models use XGBoost regression combined with CDF (cumulative distribution function) probability to generate over/under confidence for each prop line.
The Live Player Stats Tracker runs nine XGBoost models — three stats (points, rebounds, assists) across three quarterly checkpoints (Q1, Q2, Q3). These models compare real-time projections against sportsbook prop lines to generate live over/under signals during games.
The Self-Training Loop
Models improve over time through a self-training feedback loop. Each day, the pipeline grades yesterday's predictions against actual results, collects new game data, and feeds it back into the training set. Monthly retrains incorporate the latest data to keep models calibrated against the current season. This isn't a static system — it learns and adapts.
What is +EV Betting?
Expected Value (+EV) is the mathematical foundation of profitable betting. A bet is +EV when its true probability of winning is higher than what the odds imply. Over a large enough sample, +EV bets make money regardless of any individual outcome.
How It Works
Every set of odds implies a probability. When a sportsbook offers +150 on a team, they're implying that team has about a 40% chance of winning. If your model calculates a 52% chance, that 12-point gap is your edge — and that bet is +EV.
Here's the math: if you bet $100 at +150 odds with a true 52% win probability, your expected value per bet is:
You won't win every bet. But over hundreds of bets, the math plays out. That is how any probability model should be judged: over hundreds of picks, not one night.
Edge vs. Confidence
Prediction Engine shows both confidence (model's probability a pick wins) and edge (the gap between model probability and Vegas implied probability). A pick can be high-confidence but low-edge if the sportsbook already agrees. The most valuable picks have both high confidence AND significant edge — meaning the model sees something the market hasn't priced in.
Why Most Bettors Lose
Sportsbooks build a margin (the “vig” or “juice”) into every line. Standard -110/-110 odds imply both sides have a 52.4% chance of winning — totaling 104.8%, not 100%. That extra 4.8% is the house edge. To beat the vig, you need to be right more than 52.4% of the time on standard bets. Casual bettors who pick based on fandom or narratives rarely clear that bar consistently.
Betting Strategies & Bankroll Management
Having an edge means nothing if you blow your bankroll on a bad run. The strategy you use to size your bets is just as important as the picks themselves. Here are the most common approaches, from simplest to most sophisticated.
Flat Unit Betting
The simplest and most popular approach. You bet the same fixed amount — one “unit” — on every play regardless of confidence. A unit is typically 1-3% of your total bankroll.
Example: $1,000 bankroll → 1 unit = $10-$30
Pros: Simple, limits variance, easy to track
Cons: Doesn't capitalize on high-confidence spots
Best for: Beginners, anyone who wants steady growth without overthinking bet sizing
Tiered Unit Betting
A step up from flat betting. You assign 1-3 units based on the model's confidence level. You can tier by the model's confidence — a high-confidence play gets 3 units, a standard play gets 1.
Example: 70%+ confidence = 3u, 60-69% = 2u, 55-59% = 1u
Pros: Allocates more capital to your best spots
Cons: Slightly more variance than flat
Best for: Intermediate bettors comfortable with the model's confidence tiers
Kelly Criterion
The Kelly Criterion is a mathematical formula that calculates the optimal bet size to maximize long-term bankroll growth. It accounts for both the probability of winning and the odds being offered.
where b = decimal odds − 1, p = win probability, q = 1 − p
Example: Model says 60% win probability at +120 odds (decimal 2.2). Kelly = (1.2 × 0.60 − 0.40) / 1.2 = 0.267, or 26.7% of bankroll. In practice, full Kelly is extremely aggressive.
Pros: Mathematically optimal for growth, proven in finance and gambling theory
Cons: Full Kelly is volatile — a bad streak can draw down 50%+ of bankroll
Best for: Advanced bettors who trust their model's probability estimates
Fractional Kelly (Recommended)
Most professionals use half Kelly or quarter Kelly — you calculate the full Kelly percentage then bet 25-50% of it. This dramatically reduces variance while capturing most of the long-term growth.
Example: Full Kelly says 26.7% → Half Kelly = 13.3% → Quarter Kelly = 6.7%
Pros: Best risk/reward balance, still mathematically grounded
Cons: Slower growth than full Kelly (but you stay in the game)
Best for: Anyone who wants optimal sizing without the rollercoaster
Key Bankroll Rules
1. Never bet more than 5% of your bankroll on a single play. Even the highest-confidence pick can lose. Preservation is the game.
2. Track everything. Use the My Dashboard page to log bets, see your win rate by market, and identify which play types work best for you.
3. The model is not a guarantee. Even a pick the model rates at 70% is expected to lose about 3 times in 10 — the graded hit rate for each market is on the Performance page. Variance is real — trust the process over a large sample, not any single day.
4. Set a stop-loss. If you lose 10% of your bankroll in a single day, stop. The model produces fresh picks every day — there's always tomorrow.
Pricing & Plans
The monthly and 3-month plans start with a 3-day free trial so you can see the models in action before committing; the yearly plan is paid when you subscribe.
Save 17% vs. monthly
- ✓ All daily predictions (NFL, MLB, NBA and NHL)
- ✓ Live in-game trackers
- ✓ Player props & team models
- ✓ Performance analytics
- ✓ Cancel anytime
- ✓ Everything, billed monthly
- ✓ Cancel anytime
Save 54% vs. monthly
- ✓ Everything, billed once a year
- ✓ Paid when you subscribe (no trial)
Frequently Asked Questions
How accurate are the predictions?
Accuracy varies by market, and every graded result is on the performance page. The previous NBA moneyline model went 156-41 (79.2%) at 60%+ confidence from March 12 to June 13, 2026, playoffs included; from 2026-27 the NBA moneyline uses Elo ratings, whose 70%+ picks won about 4 in 5 over the last three regular seasons, out of sample. The live trackers (K predictions, hits, player stats) are not graded yet, so we quote no accuracy figure for them. The NBA 156-41 is the graded live record and the Elo figure an out-of-sample walk-forward test — not cherry-picked.
What data do the models use?
Team models use game-level statistics from API-Sports (team records, odds, scores, head-to-head history) combined with engineered features like rolling averages, rest days, travel distance, and pitching matchups. Player prop models use individual box score data from the free MLB Stats API and nba_api. Live trackers use real-time play-by-play and box score feeds updated every 30-60 seconds.
How are “Top Picks” selected?
Top Picks are the highest-confidence predictions from each market that meet minimum thresholds. For example, the top 2 moneyline plays (MLB 60%+ confidence; NBA Elo side 70%+), the top NBA pick to win straight up at a projected margin of 5+ points (not against the spread), or the top player prop with 65%+ probability. These curated picks are what the Performance page tracks and grades daily.
Can I see the model's track record?
Yes. The Performance page shows live win/loss records for every market, updated daily. It displays both the backtest baseline (historical accuracy) and live season accuracy, so you can see exactly how the model is performing in real time. Every graded pick is tracked with full transparency.
What sports do you cover?
NFL, MLB, NBA and NHL: game lines, player props and season projections. Each sport gets its own set of dedicated models — we don't reuse a generic model across sports.
What are the live trackers?
Live trackers are in-game prediction models that update in real time. The MLB Live K Tracker predicts a pitcher's final strikeout total as the game progresses. The MLB Live Hits Tracker predicts remaining hits for batters. The NBA Live Player Stats Tracker predicts final points, rebounds, and assists totals from quarterly checkpoints. Each tracker compares its prediction to sportsbook lines and generates over/under signals.
Do I need a sportsbook account?
Prediction Engine provides analysis and predictions — it is not a sportsbook and does not process bets. You would use the insights on whatever legal sportsbook is available in your jurisdiction. Users must be 21+ and comply with local gambling laws.
How do I cancel?
You can cancel your subscription at any time from your account dashboard or by contacting us. There are no cancellation fees. Cancelling stops the next renewal and you keep access until the end of the period you paid for; periods already paid for, including 3-month or yearly plans, aren't refunded.
Is this guaranteed to make money?
No. No prediction model, no matter how sophisticated, can guarantee profit. Sports have inherent randomness, and even a pick rated at 70% is expected to lose about 3 times in 10. What Prediction Engine provides is a probability for every pick and a public, graded record of how often each kind of pick hits. Responsible bankroll management and realistic expectations are essential. Never bet money you can't afford to lose.
Deep Dive Guides
Want to go deeper on a specific topic? These guides break down the math, models, and market mechanics behind each sport and market type.
The week ahead, once a week
Last week's graded record — wins and losses — and the plays for the week ahead. One email, no daily noise, unsubscribe in a click.