Learn

Strategies hub

Playbooks for stress-testing ideas in the labs. These are educational frameworks — not tips to place real wagers.

Sports Casino Virtual Bankroll Fallacies

Sports

Odds threshold filter

Only bet when American odds clear a floor (for example +300+). Useful for studying longshot distributions.

Claim: Restricting to a price band changes both win rate and payout profile.

A +300 price implies a 25% break-even probability before edge.
Selecting rarer longshots lowers win rate while increasing payout per win.
Expected value still depends on true probability versus price.

Does not prove: That longshots are systematically underpriced in real markets. That a threshold alone creates positive EV.

Open in Strategy Lab →

Sports

Consistent price band

Always bet the same odds window (for example +200 to +299) to study a stable price profile.

Claim: Holding price constant clarifies how win rate and ROI interact inside one band.

Implied probability is nearly stable inside a narrow American odds band.
ROI differences then come mostly from true-probability assumptions and sample variance.

Does not prove: That a favorite band in simulation maps to a profitable live angle.

Open in Strategy Lab →

Sports

Favorite–longshot illustration

In multi-runner fields, compare chalk win rates to longshot priced bands and field overround.

Claim: Longshot prices can appear frequently on a scattered board while still winning rarely — price frequency ≠ value.

Field overround = sum of implied probabilities from priced odds.
Appearance rate of a +700–+999 band can be high even when that band’s win rate stays low.
Fair odds from simulated win% often differ from scattered board prices (CLV proxy).

Does not prove: That any particular longshot band is +EV in live racing markets.

Open Sports Lab field mode →

Casino

Martingale on even-money bets

Double after losses on near-even bets. Educational ruin demonstration against house edge.

Claim: Martingale does not remove house edge; it concentrates risk into rare catastrophic drawdowns.

Even-money roulette bets still face a house edge (European ~1.35%, American ~5.26%).
A long losing streak requires exponentially larger stakes.
Finite bankrolls and table limits cap the “recovery” path.

Does not prove: That Martingale can overcome casino edge in the long run.

Open in Casino Lab →

Virtual

Simulated edge filter

Bet only when the simulation’s true probability exceeds implied probability by a configured margin.

Claim: Filtering on known model edge improves expected ROI under the model’s assumptions.

Edge ≈ true probability − implied probability.
Positive model edge implies positive expected value at flat stake.
Observed ROI still wanders around expected ROI because of variance.

Does not prove: That you can observe true probability in live markets. That past simulated edges transfer to real sportsbooks.

Open in Strategy Lab →

Bankroll

Flat stake baseline

Bet the same unit size on every selected event to isolate price and probability effects from stake sizing.

Claim: Flat staking makes ROI comparable across odds buckets.

Profit on a win equals stake × (decimal odds − 1).
ROI = total profit ÷ total amount staked.
Win rate alone cannot rank strategies when prices differ.

Does not prove: That flat staking is optimal for growth. That any particular odds band has a real-world edge.

Open in Strategy Lab →

Bankroll

Percent-of-bankroll staking

Stake a fixed percentage of current bankroll to explore growth versus ruin risk.

Claim: Proportional staking couples bet size to equity path and can reduce absolute ruin speed versus aggressive fixed units.

Stake_t = bankroll_t × p.
Drawdowns shrink stakes; winning runs increase stakes.
Bankruptcy probability depends on edge, odds, and percentage chosen.

Does not prove: That any particular percentage is Kelly-optimal for real markets.

Open in Strategy Lab →

Fallacies

Hot streak chase

Increase or enable betting after a run of wins. Demonstrates how streak heuristics behave under independence.

Claim: Independent Bernoulli trials do not become more likely after recent wins.

If each event is independent, P(win | recent wins) = P(win).
Streak filters change when you bet, not the underlying probability.
Variance can still produce long winning and losing runs by chance.

Does not prove: That streak systems are profitable. That real sporting events are independent Bernoulli trials.

Open in Strategy Lab →

Fallacies

Random betting control

Bet a random subset of events as a control arm against selective strategies.

Claim: Without an information edge, random selection should track the market’s vig-adjusted expectation.

If every market embeds overround, random flat betting tends toward negative EV.
Comparing selective strategies to a random arm separates skillful filters from luck.

Does not prove: That beating a random arm in one simulation proves a transferable edge.

Open in Strategy Lab →