Read the numbers properly.

A counting stat rewards opportunity. A rate stat measures quality. Confusing the two is the most common mistake in sports analysis, and it is why the league leader in a category is often not the best player in it.

A .300 average over 40 at-bats is consistent with a true talent anywhere from .158 to .442. Small samples do not produce small errors — they produce wide ranges that get reported as single numbers.

An average collapses a distribution into one number and throws away the shape. Two players averaging 18 points can have completely different floors, ceilings, and consistency — and the average cannot tell them apart.

Extreme performances tend to be followed by less extreme ones — not because anything changed, but because extremes are partly luck, and luck does not repeat. This is the most useful idea in sports analysis and the most consistently misread.

Before asking whether something is impressive, ask how often it happens anyway. Most sports claims that sound remarkable are describing an ordinary base rate in dramatic language.

Teams that run the ball a lot tend to win. That does not mean running the ball causes winning — teams run because they are already ahead. Most sports narratives are causation claims resting on correlation evidence.

If you search 500 players for a ten-game streak, you will find several — even if every player is a coin flip. The streak is real; the search that found it is what makes it meaningless.

A projection is a distribution reported as a number. The single figure is the middle of a range, and for most players that range is wide enough that the point estimate is the least interesting part of it.

Not all statistics become trustworthy at the same rate. Some are meaningful within weeks; others need more than a season. Knowing which is which tells you what you are allowed to conclude in May.

Expected statistics estimate what should have resulted from the quality of what a player did, ignoring where the ball happened to land. The gap between expected and actual is usually luck, and usually closes.

Raw statistics are not comparable across decades, because the conditions producing them changed. Era adjustment compares a player to his own league instead of to a fixed number.

Every ranking is a set of weighting choices wearing the appearance of an objective list. Two credible systems can disagree sharply and both be defensible — the disagreement is usually about what to value, not about the facts.

The same swing is a home run in one stadium and a fly out in another. Park factors adjust for that, and without them you are comparing players who played different games.

Points per game mixes three separate things: how fast a team plays, how much of the offense runs through a player, and how efficient he is with it. Pace and usage let you pull them apart.

A 7-3 record against the league's easiest schedule and a 6-4 against its hardest are not the same achievement. Strength of schedule is the adjustment, and it matters most early in a season.

Home advantage is real, smaller than most people assume, and varies a lot by sport. It is also not one effect — it is several, and they do not all move together.

Fatigue is one of the few context effects that is both well documented and visible in advance, because the schedule is published months ahead.

Splits divide a season into smaller pieces, which means every split is a small sample. Some splits reflect genuine, stable skill differences; most are the sample size problem wearing a costume.

The same player is worth different amounts in PPR and standard scoring. Format is not a detail you adjust for at the margin — it reorders the position.

Fantasy points are volatile week to week; role is not. Snap share and target share tell you what a player is being asked to do, which predicts future production far better than past production does.

Fantasy seasons are decided in three specific weeks. Which opponents your players face in those weeks is knowable in August and is routinely ignored until December.

Average Draft Position measures where drafters are taking a player. That is a fact about drafters, not about the player — and the distinction is the entire point of using it.

When a starter goes down, the replacement rarely inherits the whole role. Work out which parts of the job transfer before assuming the backup inherits the production.

The waiver wire rewards acting on role changes before they turn into points. That means reading usage trends, not last week's box score, because everyone can see the box score.

American odds are just probabilities wearing a costume. -150 means the book thinks it hits about 60% of the time. Learn the conversion and every line on every board becomes readable.

Expected value is the long-run average of a bet — what you'd earn per bet if you could make it a thousand times. A +EV bet can lose tonight and still be the right bet. That distinction is what separates process from luck.

Both sides of a -110/-110 line imply 52.4% each — 104.8% total. That extra 4.8% is the book's fee. You must be better than the vig, not just better than a coin flip, to profit.

The same bet is priced differently at different books. Always taking the best available number is the closest thing to free money in this hobby — worth several percent a year with zero prediction skill required.

Tickets% counts bets; money% counts dollars. When 70% of tickets are on the Over but only 45% of the money, a few big, informed bets are quietly on the Under. That gap is one of the market's most honest signals.

Lines move when money or news arrives. A price shortening fast (steam) means informed money is piling on; a slow drift usually means book rebalancing. Beating the closing line is the classic proof you're betting well.

Decide what your bankroll is, bet 1-2% of it per play, and never change stake size because of how yesterday went. Sizing discipline is the difference between a rough week and a wiped-out account.

A 90% L10 hit rate is 9 games of evidence — enough to notice, not enough to trust blindly. Combine short-term form with larger windows, matchup context, and price before believing a streak.

Across decades of markets, longshots are systematically overpriced and heavy favorites slightly underpriced — because big payouts are fun to bet. Our own research found the same pattern in props. Respect it.

Parlays multiply each leg's vig together, so the house edge grows with every leg you add. They're a legitimate entertainment product — just know the tax rate rises with the payout.

Betting should cost what a hobby costs and feel like a hobby feels. Set limits before you start, never bet money that has a job, and know the warning signs and the help line: 1-800-GAMBLER.