Breakdowns guide for betting mathematics

Educational disclaimer

Novus Odds is a simulation laboratory. Nothing here is a tip to wager.

Every market, team, player, runner, and casino round in Novus Odds is synthetic. There are no live sportsbook feeds, no real casino tables, and no wagering.

Positive expected value in a lab means the model’s assumed true probabilities beat the model’s quoted prices under those assumptions. It is not proof you will win in a real book or casino. Sportsbooks and casinos retain vig or house edge. Variance means you will not always win even when a model edge exists.

Use this guide to learn how to interpret experiment output — not to place bets.

• Synthetic data only — no live odds.
• Not a sportsbook, casino, or tip service.
• House edge and vig persist in real products.
• Short samples can look great or terrible regardless of long-run EV.

Core metrics: edge, EV, vig, ROI, drawdown

Shared vocabulary used across Odds, Strategy, Sports, Props, and Casino labs.

• Implied probability — break-even rate implied by the American price alone (before any model of “true” chance).
• Simulated true probability — the lab’s underlying chance used to settle random outcomes. Only known inside the simulation.
• Edge (true − implied) — how much the model’s true chance exceeds the price’s break-even chance. Positive edge is a model advantage, not a live-market guarantee.
• Expected value (EV) / expected ROI — average profit per unit stake if the true probability is correct, before random outcomes are drawn.
• Observed ROI — profit or loss from the actual simulated wins and losses. Differs from EV because of variance.
• Vig / overround — extra margin baked into prices so combined implied probabilities exceed 100%. Raises the bar for a profitable edge.
• Drawdown — peak-to-trough decline in bankroll or cumulative profit. Illustrates path risk even when EV is positive.
• Win rate — fraction of selected events that won. Insufficient alone: a low win rate at long prices can beat a high win rate at short prices.

Bankroll paths and drawdown

Equity and bankroll charts show path risk, not just end profit.

Max drawdown answers: “How deep did the trough get relative to a prior peak?” Strategies with similar end ROI can have very different drawdown profiles — especially progressive stakes like Martingale.

Use drawdown together with longest losing streak, bankruptcies (Casino Lab), and confidence intervals on win rate when available.

Odds Lab overview

Synthetic two-outcome markets with filters, buckets, A/B compare, and full breakdown suite.

Configure event count, odds range, market overround, pricing uncertainty, stake, bankroll, and selection mode (all, favorites, underdogs, edge filter, or odds threshold). Optional A/B compare runs a second selection rule on the same tape.

After a run you get summary cards, equity/drawdown/histogram charts, example tickets, and the breakdown panels described below.

Mathematical advantages panel

Average edge, expected ROI, vig setting, CLV proxy, and Kelly-lite hints from the simulated tape.

• Avg edge (true − implied) — mean model edge across selected (or tape) events.
• Expected ROI / unit — model EV per unit stake.
• Break-even / avg implied — average price-implied probability.
• Avg simulated true prob — average of the lab’s true probabilities.
• Vig / overround setting — the margin you configured; it is a cost you chose to study, not a live book’s exact hold.
• CLV proxy — comparison of fair vs priced implied probabilities when fair odds exist; otherwise falls back to edge.
• Kelly-lite — capped fractional Kelly using assumed true probabilities. Educational stake hint only; not advice to size real bets.

Odds bucket / example bet frequencies

How often each American-odds band appears on the tape, with win rates and which strategies would select them.

Each row is a price band (for example +300 to +399) with share of tape, representative odds, average win rate, implied vs true probabilities, and strategies that would pick that band.

This is a frequency map of the synthetic market — useful for understanding how rare long prices are under your generator settings.

Strategy overlay breakdown

Same synthetic events, different selection rules — selection share, win rate, EV ROI, observed ROI.

Overlays answer counterfactuals: if every rule had seen this tape, which would have selected how much, and with what model EV vs realized ROI?

Selection share matters. A rule that picks 2% of events can show noisy observed ROI even when EV looks attractive.

Hot streak lens

Stress-test the gambler’s fallacy / hot-hand heuristic: win rate after recent wins vs baseline.

Windows count how often a streak threshold was hit and what the next-event win rate was versus the overall baseline.

In a fair independent model, next-event rates should hover near baseline. Large gaps in small samples are variance, not proof of momentum.

Consistent price-band selection

Always selecting the same odds band — win rate and ROI when price is held roughly constant.

Consistent bands isolate price. Instead of chasing edges or streaks, you always take events inside a fixed American-odds window.

Win rate should roughly track the band’s typical break-even probability after vig; ROI shows whether that was enough under the simulated true probabilities.

Strategy Lab

Stake rules and selection heuristics on synthetic market tapes, with overlays, bands, and streak lenses.

Families include flat, threshold, edge filters, hot-streak chase, bankroll-percent, Martingale, random, and more. Optional compare family runs a second rule.

Read the strategy matrix and bankroll path alongside overlays, consistent bands, and hot-streak panels — selection quality and stake sizing are different questions.

Sports Lab

Fictional matchups, soccer 1X2, and multi-runner fields with fair vs priced markets.

Matchup mode rates Team A vs Team B with spreads and totals. Soccer mode produces 1X2 prices. Field mode builds a multi-runner board with odds scattered across longshot bands.

Math advantage, hot streaks, and consistent bands still apply. Field mode adds horse-style frequency tables.

Horse-field / multi-runner frequencies

How often each longshot odds band appears in an N-runner synthetic field, and win rates when present.

Appearance rate is not the same as win rate. A +700-style price may appear often in an 8-runner field yet win rarely when it does — that is the longshot structure you came to inspect.

Example odds and average odds help you picture what “+700-style” meant in that run.

Props Lab

Fictional player distributions, line sweeps, and over/under EV curves.

You set a player’s scoring distribution (mean, spread, minutes, usage, opponent modifier) and a prop line with juice. The lab sweeps nearby lines to show how probability and EV change.

There is no live player data. Treat names and stats as placeholders for distribution mathematics.

Casino Lab

House-edge games with transparent bet math, bankroll paths, and strategy comparisons.

Games include roulette (European/American), blackjack, baccarat, craps, and slots with configurable RTP/volatility where relevant.

Unlike sports labs that can inject model edge via mispricing, casino bets are designed so the house edge is known and usually negative for the player.

Casino bet tables and house edge

Fair probability, payout, house edge, and player EV for each listed bet type.

• Fair probability — chance of winning under the rules model.
• Payout — stated odds paid on a win (e.g. 1:1 even money).
• House edge — long-run expected fraction of each stake kept by the house.
• Player EV — typically the negative of house edge for these tables.
• Strategy breakdown — flat vs configured stake path on the same round sequence.

Datasets lab

Exportable fictional seasons — teams, games, and sample odds for offline study.

Choose sport, team counts, seasons, games per season, universes, and seed. Summaries and sample rows are what you keep; full Monte Carlo ticks stay in memory.

Use datasets when you want a portable synthetic league history rather than a single interactive experiment chart.

Practical educational use cases

Suggested lab workflows for learning — still not wagering tips.

When something looks too good, check sample size, selection share, vig, and whether the lab injected a known model edge. When something looks too bad, check drawdown and variance notes before blaming the math.

• Sports Lab field mode — measure how often +700-style prices appear in an 8-runner synthetic field and what their win rate is when present.
• Odds Lab edge filter — quantify how much selection share you give up to raise average model edge.
• Strategy Lab hot-streak — show that post-streak win rates need not beat baseline under independence.
• Consistent bands — hold price fixed and study ROI as a function of band alone.
• Casino Lab Martingale — demonstrate ruin risk despite short winning streaks when house edge is negative.
• Props Lab line sweep — find the line where EV flips sign for a known distribution and juice.
• A/B compare — teach “same tape, different rules” before debating which heuristic “works.”
• Seed control — reproduce a surprising session for discussion, then change seed to show it was one history among many.