Sports Betting Glossary

A pick is a call Agatha stands behind on the record. It starts as a card on the board, freezes 15 minutes before its start, takes a ticket number and is sealed. From then on its side, line and price never change, and it is graded win, loss, push or void at one unit. Only picks count on the track record.

A prediction is Agatha's call on a game shown on a schedule or game page when that game is not on the board. It says which side she favors, but it is never listed as a pick: it does not freeze, takes no ticket number, is not sealed and is never graded on the track record. Treat it as her read, not as a bet she stands behind.

A lean is a pick tagged LEAN: Agatha's call on a game no validated model covers yet. It is listed, frozen, numbered, sealed and graded exactly like an edge, and it carries the same one unit. The tag tells you where the call came from, not how much it is worth or how sure she is; her confidence number says that.

An edge is a pick tagged EDGE: a pick from a model validated against its own graded history. Before it could publish, that model had to hold up on seasons it never saw while it was being built. An edge is staked, frozen, sealed and graded exactly like a lean, at one unit. The tag describes where the call came from; it is not a promise about the result.

DOG marks a moneyline pick at plus odds, 2.00 or longer in decimal, which is +100 or longer in American odds. It shows on edges and leans alike. Props, spreads and totals never carry it, whatever their price, and neither does a race or tournament winner. The tag describes the price only; the stake stays one unit, like every other pick.

Confidence is Agatha's calibrated probability that a pick wins, and nothing else. When she says 65, she expects to lose 35 of every 100 such picks. It does not set the stake, which is one unit on every pick. Read it next to the implied probability of the price: a call is only worth a bet when her number is higher than the price's.

A unit is the stake size the record is kept in. Every pick is one unit, edge and lean alike, so the hit rate and the units line tell the same story and no stretch can be dressed up by pressing harder on some picks. Picks listed before this rule keep the size they were listed at, because a listed pick is never edited. A unit is worth whatever you decide.

Every pick takes a ticket number at its freeze, 15 minutes before its start, from one counter shared by every sport and both tags. Numbers run in order from #1 and are never reused. Enter a number in the ticket lookup on the track record to open that pick's sealed receipt, or open its verify page directly.

On the record Tickets so far run from #1 to #1529.

Sealing publishes a pick before its game can decide it. At the freeze, 15 minutes before the start, the pick is encrypted and anchored on Stellar with a fingerprint of its exact text, so nobody, Agatha included, can change it afterwards. Once that day's games are graded its key is released, and anyone can decrypt the pick and check it against the fingerprint.

On the record Right now every pick is sealed on the Stellar test network.

A receipt is the public proof page for one sealed pick, at /verify followed by its ticket number. It shows the fingerprint, the Stellar transaction that carries the sealed bytes, when the seal landed against the game's start, and the day key once it is released. With those, you or an AI assistant can decrypt the pick and check it was not changed after the fact.

A dynamic card is a card on the board that has not frozen yet. Its best price and book can still move with the market, and it can still be delisted if it drops below the board's bar. It becomes a pick 15 minutes before its start, when it freezes, takes its number and is sealed. Nothing about a dynamic card is graded.

A delisted card was pulled from the board before its freeze, for example because its price drifted below the board's bar. It was never a pick: it took no ticket number, was never sealed and is never graded or counted anywhere. The board's timeline keeps a line saying it was removed and why, so the pull itself stays on view.

Withdrawn is the label on older record rows that were taken back after they were listed, under a rule the site no longer uses. A withdrawn row stays visible with its tag, but it is never graded and counts nowhere: not in the win and loss record, the units or the ROI. A card pulled before its freeze today is delisted instead, because it was never a pick.

Agatha's board day runs from 3:00 AM to 3:00 AM Eastern, not midnight to midnight. A game that starts before 3:30 AM Eastern belongs to the previous board day, so a late fight card or a West Coast game after midnight stays with the slate it was part of. The board day decides where a pick is filed and whether a moved game was played on its day.

The track record is the public ledger of every graded pick since the record start: wins, losses and pushes at the frozen price, one unit each, with net units and ROI by sport and bet type. Nothing on it is edited or deleted, losses included, and any numbered pick opens to its sealed receipt, so every result can be checked.

On the record 2,630 graded picks on it so far, 1,531-1,095-4, -58.96u at one unit each.

The record start is the first board day the public track record counts. Picks from before it are not part of the record and are never shown as record. A league whose lane was rebuilt can carry a later start of its own, and its picks count from that later date. No model publishes until it passes validation for its sport.

On the record Agatha's record counts from August 24, 2026.

A pick wins when its side comes in at the frozen line and loses when it does not. A win pays the frozen price on one unit, so a win at -150 nets 0.67 units and a win at +150 nets 1.5; a loss costs the one unit staked. Both count in the win and loss record, the hit rate, the units and the ROI.

A push is a tie with the line: the result lands exactly on the number, like a spread decided by exactly the points given. The stake comes back, so a push nets zero, but it still counts as one unit staked because the bet was made. Pushes are left out of the hit rate. A two-way moneyline that ends level also pushes.

On the record On Agatha's record, 4 of 2,630 graded picks pushed.

A void means no action: the pick settles as if it was never made. Agatha voids a pick when its game is postponed, canceled or moved off its day, when a named player or pitcher does not take part, or when a sport's rule says the market cannot settle. A void counts nowhere: not in the win and loss record, the units, the ROI or the hit rate.

On the record On Agatha's record, 78 of 2,708 settled picks were void (2.9%).

Under Agatha's day rule, a game not played on its board day is void. A postponed or canceled game voids at once, and so does a game whose start moves to another board day, however small the move. A game that starts, is suspended and later resumes keeps its pick until it finishes. After a postponement the game can be listed again on the day it is played.

A shortened game stops before its full length, most often a baseball game called for rain. An official shortened MLB game still settles Agatha's moneyline picks. A strikeout prop needs a full game, nine innings or eight and a half with the home side ahead; short of that it is void, unless the Over had already cleared its line when play stopped.

In tennis a retirement is a player quitting a match already under way, usually through injury, and a default is a player being disqualified. Agatha voids a match pick when the match ends either way, at any stage, one game in or two sets up, even though the scoreboard names a winner. A walkover, where the match is never played at all, is void too.

A player prop needs its player to take part. When the official box score or game roster says the player did not play, the pick is void. Taking part and recording nothing is different: a receiver who plays and catches no passes is graded at zero, so the Under wins. When the evidence is unclear, the pick waits for clear evidence instead of being guessed.

American odds show a price around a base of 100. A minus number is what you stake to win 100, so -150 means staking 150 to win 100. A plus number is what a stake of 100 wins, so +150 wins 150. Both -100 and +100 are even money, and nothing strictly between them is a valid American price.

Decimal odds show the whole return on a stake of 1, stake included. A price of 2.50 returns 2.50 for every 1 staked: 1.50 of winnings plus the stake back. Even money is 2.00. The implied probability is 1 divided by the price, so 2.50 implies 40 percent. Decimal odds make parlays simple: multiply the legs.

Fractional odds, common in Britain and Ireland, show winnings over stake. 5/2 means 5 won for every 2 staked, the same price as +250 or 3.50 decimal. Evens is 1/1. A fraction below one, like 1/2, is a favorite: stake 2 to win 1. Add 1 to the fraction to get the decimal price.

Implied probability turns a price into the win chance it charges for. For decimal odds it is 1 divided by the price. For American odds, -150 implies 150 out of 250, or 60 percent, and +150 implies 100 out of 250, or 40 percent. Because books add a margin, the sides of a market imply more than 100 percent together.

The break-even win rate is how often a bet at a given price has to win just to come out even over time. It equals the implied probability: 52.38 percent at -110, 50 percent at +100, 40 percent at +150. At a single price, a record is ahead only when it beats that rate; across mixed prices, which picks won matters too, so the return is the final check.

On the record The prices on Agatha's graded picks needed 59.4% on average; she has won 58.3%.

The vig, or juice, is the book's cut built into its prices. On a market priced -110 on both sides, each side implies 52.38 percent, 104.76 percent together, and the extra 4.76 percent is the vig, stated as the overround. You pay it whether you win or lose, which is why the same bet at a better price is worth more.

The overround is how far a market's implied probabilities add up past 100 percent: the sum of 1 divided by each decimal price, minus 1. Two sides at 1.909 each give 0.5238 twice, 1.0476 in all, an overround of 4.76 percent. It is the book's margin measured on the prices; hold measures the same margin on the money.

Hold is the share of the money bet on a market that the book expects to keep when the action matches its prices: 1 minus 1 divided by the sum of the implied probabilities. A -110 and -110 market sums to 1.0476, so its hold is 1 - 1/1.0476 = 4.55 percent, a little under its 4.76 percent overround.

No-vig odds, also called fair odds, are a market's prices with the book's margin taken out, so the implied probabilities add up to exactly 100 percent. The simplest method scales each side down by the overround: a -110 and -110 market becomes 50 percent a side, fair odds of +100. Comparing a price with fair odds shows what the margin costs you.

Line shopping means checking several books and taking the best price on offer for the side you want. Books often differ by a few cents on the same game, and over hundreds of bets that gap adds up. Each Agatha card shows the best price among the books it tracks, and a pick is graded at the price it froze with.

The closing line is a market's last price before the game starts. By then the most information and the most money have gone into it, so it is the best single estimate the market offers of each side's chances. For each pick, Agatha stores the last pregame price at the same book, where that book quoted one, and compares it with the frozen price.

Closing line value, or CLV, measures whether you got a better price than the close. On Agatha's record it is the frozen decimal price divided by the same book's decimal closing price, minus 1: taking 2.10 when the close is 2.00 is +5 percent. Prices settle faster than results, so CLV says more about a small sample than a win and loss record does.

On the record Agatha's picks with a stored close average +0.88% CLV, over 1,839 picks.

Expected value, or EV, is what a bet returns on average per unit staked if it could be repeated forever: the win probability times the net win, minus the loss probability times the stake. At +110 with a true 50 percent chance, EV is 0.5 x 1.1 - 0.5 x 1 = +0.05, or +5 percent. The hard part is knowing the true probability.

ROI, return on investment, is net units won divided by units staked. On a record where every pick is one unit it reads as the average result per pick: -2 percent means losing two units for every 100 staked. Pushes add to the units staked and nothing to the net; voids add to neither. Over a few hundred picks ROI moves a lot by chance.

On the record Agatha's ROI is -2.26% on 2,607 units staked.

Hit rate is the share of decided picks that won: wins divided by wins plus losses, with pushes and voids left out. On its own it says little, because 60 percent on heavy favorites can lose money while 45 percent on long prices can make it. Always read a hit rate next to the break-even win rate of the prices that were taken.

On the record Agatha's hit rate is 58.3% over 2,626 decided picks.

A moneyline is a bet on who wins the game outright, with no points given or taken. Favorites carry minus prices and underdogs plus prices. On Agatha's record a two-way moneyline that ends level is a push, a pitching change never voids one, and an official shortened baseball game still settles it on the final score.

A three-way moneyline, the standard soccer market, prices three results: home win, draw and away win. A pick on a team loses if the match is drawn. A knockout match settled in extra time or on penalties counts as a draw at 90 minutes, so a team pick loses there too, even when its team goes through. The no-vig calculator takes all three prices.

A spread bet gives the underdog points, or takes them from the favorite, to even the game out. A spread pick wins when its side's final margin plus the line is above zero, loses below zero and pushes at exactly zero. At -7.5 a team must win by 8 or more; at +3 it covers by losing by 2 or less and pushes on a 3 point loss.

A total, or over/under, is a bet on the combined score of both sides against a posted number. The Over wins when the final combined score goes above the number and the Under when it stays below it. Landing exactly on a whole-number total is a push, while a half-point total like 8.5 can never push. Both sides are usually priced near -110.

A player prop is a bet on one player's own number, such as a pitcher's strikeouts, assists or made threes, receptions or receiving yards, set against a line with an Over and an Under. It needs the player to take part: if the official record says the player did not play, the prop is void. A result landing exactly on the line pushes.

A strikeout prop is a bet on how many batters a starting pitcher strikes out, Over or Under a line. Agatha grades it on the pitcher's official strikeouts once the game is final. If the named pitcher does not start the game, the pick is void, and a game called before it is a full game voids it unless the Over had already cleared.

An outright is a bet on the winner of a whole event with a field of entrants, such as a race or a golf tournament, instead of a single game. Prices run long because most of the field loses: +1000 implies about 9 percent. On Agatha's race and golf pages her call on the winner is an outright call, and it never carries a DOG tag.

A parlay joins two or more picks into one bet that wins only if every leg wins; the payout is the legs' decimal prices multiplied together. A leg that pushes or is void drops out and the payout is worked out again from the legs left. Each added leg also multiplies the book's margin, which is why parlays cost more than single bets over time.

A hedge is a second bet on the other side of an open bet, placed to narrow the gap between outcomes, for example backing the opponent in the last leg of a parlay. It trims the downside at a price: the hedge pays the book's margin again, so on average it lowers what the position is worth. It can reduce risk; it never removes the cost.

A bankroll is money set aside for betting and nothing else, money you can afford to lose in full. Stakes are sized as a share of it, which is what units express; a common convention is one unit at 1 to 2 percent of the bankroll. Setting it before you bet, and not topping it up after losses, is the simplest limit there is.

The Kelly criterion sizes a bet as a share of the bankroll from its expected value and its price: the EV per unit divided by the decimal odds minus 1. It grows a bankroll fastest only when the win probability is exactly right and swings hard when it is not, so most people who use it bet a fraction, often a half or a quarter.

A drawdown is how far a bankroll or a record has fallen from its highest point, measured in units from that peak to the lowest point after it. Every betting record has them, winning ones included, and their depth depends on the prices and the number of bets as much as on skill. The deepest one so far is the maximum drawdown.

On the record Agatha's deepest drawdown so far is 84.38 units below her running peak, counted pick by pick, reached on September 27, 2026.

Variance is how widely results scatter around what a bettor should expect. Over a small number of bets it swamps skill: at even money, a bettor who truly wins 52 percent of the time still finishes 100 bets below 50 percent about three times in ten. It shrinks slowly, with the square root of the number of bets, which is why a record needs hundreds of picks.

A losing streak is a run of losses in a row, and they run longer than intuition expects. At a 55 percent win rate the longest losing run expected over 1,000 bets is about eight, and a run of six or more is close to certain. A streak on its own says little about skill; the prices paid and the closing line value say more.

On the record Agatha's longest losing run is 7 picks; over 2,626 decided picks at her win rate, the longest run expected by chance is about 8.5. Every pick in that run started at the same time.

Calibration is whether stated probabilities come true as often as they say. A calibrated 60 wins about 60 of every 100 picks; if those picks win 50, the number is overconfident. Checking it takes many picks in each band. Agatha publishes her confidence on every pick so her calibration can be checked against the graded record instead of taken on trust.