Probability Basics
Probability is the mathematical measure of how likely an event is to occur, expressed as a number between 0 (impossible) and 1 (certain), or as a percentage between 0% and 100%.
Basic formula: Probability = Favorable Outcomes / Total Possible Outcomes
Examples:
- Probability of rolling a 6 on a fair die: 1/6 = 16.67%
- Probability of drawing an Ace from a full deck: 4/52 = 7.69%
- Probability of getting heads on a fair coin: 1/2 = 50%
- Probability of being dealt a pocket pair in Texas Hold'em: 6/C(52,2) = 6/1326 = 5.88%
Independent vs Dependent Events
Independent events: The outcome of one event does not affect another. Each dice roll, each coin flip, each spin of a roulette wheel is independent. Previous results have no bearing on future outcomes.
Dependent events: The outcome changes based on previous events. Drawing cards from a deck without replacement is dependent — each card drawn changes the composition of the remaining deck. This is why card counting works in Blackjack.
Odds Formats
Odds express the likelihood of an event occurring and determine payouts. Three formats are used globally:
| Format | Example | How to Read | Implied Probability | Profit on $100 Bet | Common In |
|---|---|---|---|---|---|
| Decimal | 2.50 | Total return = stake x odds | 1/2.50 = 40% | $150 profit ($250 total return) | Europe, Australia, Asia |
| Fractional | 3/2 | Profit of 3 for every 2 staked | 2/(3+2) = 40% | $150 profit | UK, Ireland |
| American (Positive) | +150 | Profit on a $100 bet | 100/(150+100) = 40% | $150 profit | USA, Canada |
| American (Negative) | -200 | Stake needed to win $100 profit | 200/(200+100) = 66.7% | $50 profit | USA, Canada |
Converting Between Formats
- Decimal to Fractional: Decimal - 1 = Fractional (as a fraction). Example: 2.50 - 1 = 1.50 = 3/2.
- Decimal to American: If Decimal >= 2.00: American = (Decimal - 1) x 100. If Decimal < 2.00: American = -100 / (Decimal - 1).
- Decimal to Implied Probability: 1 / Decimal x 100%.
Expected Value (EV)
Expected value is the most important concept in mathematical gaming analysis. It calculates the average outcome of a decision if repeated many times.
Formula: EV = (Probability of Winning x Amount Won) - (Probability of Losing x Amount Lost)
Example: A coin flip game pays $2.10 for heads and you lose $1.00 for tails.
- EV = (0.50 x $2.10) - (0.50 x $1.00)
- EV = $1.05 - $0.50
- EV = +$0.55 per flip
A positive EV (+EV) means the decision is profitable long-term. A negative EV (-EV) means you'll lose money over time. Every decision in gaming can be evaluated through this lens.
House Edge by Game
| Game | House Edge | Player RTP | Skill Factor | Notes |
|---|---|---|---|---|
| Blackjack (basic strategy) | 0.5% | 99.5% | High | Lowest house edge with optimal play |
| Blackjack (average player) | 2.0% | 98.0% | High | Mistakes increase house edge significantly |
| Poker (rake) | 2.5-5%* | Varies | Very High | *Rake is taken from pot, not edge vs house; players compete against each other |
| Rummy (rake) | 5-15%* | Varies | Very High | *Platform rake per contest; skill determines player-vs-player outcomes |
| Baccarat (Banker) | 1.06% | 98.94% | None | Best non-skill bet available |
| Baccarat (Player) | 1.24% | 98.76% | None | Slightly worse than Banker bet |
| Craps (Pass Line) | 1.41% | 98.59% | None | One of the better bets in dice games |
| Roulette (European, single-zero) | 2.70% | 97.30% | None | Single zero is significantly better than double |
| Roulette (American, double-zero) | 5.26% | 94.74% | None | Avoid; the extra zero nearly doubles the edge |
| Slots (average) | 2-15% | 85-98% | None | Varies enormously by machine/game |
| Keno | 25-30% | 70-75% | None | One of the worst bets in any format |
RTP (Return to Player): The percentage of total wagered money that is returned to players over time. An RTP of 97% means that for every $100 wagered, $97 is returned and $3 is the house's profit on average.
Variance & Standard Deviation
Variance measures how spread out results are from the expected value. High variance means wild swings; low variance means more predictable results.
- Low variance games: Blackjack, Baccarat. Results stay close to the expected value. Fewer big wins or big losses per session.
- Medium variance: Poker, Rummy. Results depend heavily on skill and hand distribution within a session.
- High variance: Slots, tournament poker, lottery. Individual sessions are unpredictable. Need many sessions for results to converge toward EV.
The Law of Large Numbers
As the number of trials increases, actual results converge toward the mathematical expectation. This means:
- Short term (10 hands): Anything can happen. Results are essentially random.
- Medium term (1,000 hands): Patterns begin to emerge but significant deviation is still normal.
- Long term (100,000+ hands): Results closely approximate the expected value.
Sample Size Matters
Judging your skill or a strategy based on small samples is one of the most common analytical errors:
| Sample Size | Reliability | What You Can Conclude |
|---|---|---|
| 10 sessions | Very Low | Almost nothing; too small to distinguish skill from luck |
| 50 sessions | Low | Broad trends visible but still high variance |
| 200 sessions | Moderate | Reasonable confidence in general win/loss rate |
| 500 sessions | Good | Strong statistical basis for evaluating performance |
| 1,000+ sessions | High | Results reliably reflect your true skill level |
Common Probability Fallacies
| Fallacy | Description | Why It's Wrong | Reality |
|---|---|---|---|
| Gambler's Fallacy | "Red has come up 5 times, black is due" | Each spin is independent; the wheel has no memory | Probability of red/black remains the same every spin |
| Hot Hand Fallacy | "I'm on a winning streak, I can't lose" | Past wins don't change future probabilities in independent games | Winning streaks are statistically expected and inevitable; they don't predict future success |
| Availability Bias | "People win jackpots all the time" (from seeing ads/news) | Winners are publicized; the millions of losers are not | The base rate of winning is extremely low regardless of what media shows |
| Sunk Cost Fallacy | "I've lost $500 so I can't quit now" | Money already lost is irrelevant to future decisions | Each decision should be evaluated independently on its own merits |
| Near-Miss Fallacy | "I almost won! I'm getting closer!" | A near-miss is the same as any other loss mathematically | Slot machines are designed to show near-misses to encourage continued play |
| Illusion of Control | "I have a system that beats the odds" | No betting system can overcome a negative expected value | Martingale, Fibonacci, and other systems all fail mathematically long-term |
Using Math to Make Better Decisions
- Always check the house edge: Before playing any game, know the mathematical cost. Choose games with the lowest house edge.
- Calculate EV before big decisions: In poker or rummy, estimate your pot odds and compare them to your winning probability.
- Focus on skill games: In poker and rummy, the house takes a rake, not an edge. Your profit comes from being better than opponents, which is achievable through study.
- Respect variance: Don't confuse a losing streak with lack of skill or a winning streak with guaranteed ability. Look at long-term data.
- Set rational limits: Use the expected cost of a gaming session (house edge x total wagered) as your budget, not an arbitrary number.
- Avoid negative EV side bets: Insurance in Blackjack, tie bet in Baccarat, and bonus rounds with high house edges are all mathematically poor decisions.
Sources & References
Epstein, Richard A. (2012), "The Theory of Gambling and Statistical Logic." Haigh, John (2003), "Taking Chances: Winning with Probability." House edge data compiled from Wizard of Odds (wizardofodds.com), a peer-reviewed gambling mathematics resource. Law of Large Numbers formally proven by Jacob Bernoulli (1713), "Ars Conjectandi." Expected value theory from Pascal and Fermat correspondence (1654).