Glossary

The bet itself

Moneyline: A bet on who wins, at a price. No margin, no line to cover.
Spread: A bet on the margin. The favourite gives points and has to win by more than the number; the underdog takes them and can lose by less.
Over / under (the total): A bet on the two sides' combined score against a number, also called the total. Over wins above it, under below it, and landing exactly on it is a push.
To win: Profit if the bet wins, on top of your stake. The total the book hands back is this plus the stake, so a $100 bet at +150 wins $150 and returns $250.
Odds format: Prices here are American throughout: a minus is the stake needed to win 100, a plus is the profit on a stake of 100. There is no decimal or fractional view.
Implied probability: The chance a price implies, before the book's margin is taken out. −150 implies 60 percent, +150 implies 40; the two sides of a real market add to more than 100, and the excess is the vig.
Free bet: Credit the book handed over that can be staked but not withdrawn. Winnings come back as cash; the stake does not. In your record a lost free bet still costs its face value, because a won one is counted in full; counting the win and not the loss would flatter every total.
Live bet: A ticket placed after its game had started; its row carries the live badge.

How a bet ends

Graded: A bet the book has settled, so it has a result and its money has moved. Everything else is open.
Push: The result landed exactly on the line, so the bet is a tie. Your stake comes back in full and nothing is won or lost. It counts in neither side of the win rate.
Void: The bet was cancelled: a game postponed, a player scratched, a line pulled. Your stake comes back in full, as with a push, and it counts in nothing.
Cashed out: You took the book's offer to settle early. What comes back is the cash-out value, not the full return: the book keeps a margin for taking the risk off you, so it is always less than the bet would pay if it went on to win.

Your record

Record: Wins, losses and pushes on graded bets. A push is a tie with the line and counts as neither.
Win rate: Wins divided by wins plus losses. Pushes and voids are left out of both sides.
Profit: Profit or loss on every graded bet, after stakes. A lost free bet counts as its stake, the same as cash.
ROI: Profit divided by the money staked on graded bets. Plus 5 percent means five dollars back for every hundred risked.
Graded stake: The stake on graded bets, the figure ROI is taken on.
At risk: Stake on open bets, and what they would win on top of that stake if every one of them came in. It is the profit, not the total returned.
Units: Every figure divided by one unit size, set in Settings. One size is applied to the whole history, so a unit figure is what a result would have been at today's stake.
Profit over time: Running profit, one point per game day, over the span the chart names: the stretch itself for a week or all time, the trailing month for a single day. Flat stretches are days with nothing graded.
Drawdown: How far the running total fell from its highest point so far. A normal losing week shows here as a dip that recovers; a broken model does not recover.
Track: Infra tickets are plumbing tests and are left out of every total. Any other track counts normally.

Games in play

Pace: Where a live ticket is heading against the line you took, on a half-circle gauge: red for heading to lose, amber for too close to call, green for heading to win. A total or a player's number at its rate so far (a pitcher's over a six-inning start), a spread with the rest of the game going as the line expects, a moneyline by its lead against the time left.
In play split: The bets on games under way, by where each is heading on its gauge: green on track, amber too close to call, red behind, grey nothing to read.
Projected record: The record if every bet in play ends the way it is heading: its pace gauge's reading where it has one, else how it stands. A bet heading for a win counts as a win, one heading for a loss as a loss, one too close to call or with nothing to read as neither. It moves with the scores.
Projected profit: The stretch's profit if every bet in play ends the way it is heading: the graded profit, plus what each bet heading for a win would win, less the stake of each heading for a loss; a bet too close to call, or with nothing to read, adds nothing. It moves with the scores. In the ROI mode it is that profit over the stake it is taken on.

Closing line value

Closing line value (CLV): The price you took against the price the market closed at. Beating the close over many bets is the clearest sign of an edge, because it does not depend on luck.
Closing line: The last price captured at the bet's own book before kickoff. Sides and totals are captured on a schedule, usually within 20 minutes of the start; a prop only when the desk or the bot priced it, so its close can be older.
Beat the close: The share of bets whose price was better than the closing price. The market itself sits near 50 percent; a lasting rate above 55 to 60 is unusual.
Decided: Scored bets with the ties taken out of both sides. A bet that landed exactly on the closing price neither beat it nor missed it, so counting it as a miss would drag a flat record under 50 percent for agreeing with the market.
Scored: A bet is scored when its close was priced at the point it took. A close at another point cannot be compared, and a bet level with the close is out of the rate.
CLV coverage: A bet has a close only when a line at its own book was captured before kickoff. A game or a prop never captured has none, and cannot be measured.
Avg CLV: Mean closing line value over the scored bets. At a fair close it is roughly what each dollar staked was worth, before luck.
Median CLV: The middle closing line value: half the scored bets did better, half worse. One big miss moves the average but not this.
CLV interval: The range the true beat-the-close rate most likely sits in, given how few bets are scored. Wide means too few to tell.
CLV distribution: Bets counted by their closing line value, one bar per percent. The end bars hold everything past 10 percent either side.
Rolling CLV: Average closing line value over the most recent bets, moving as new bets grade. A lifetime average can hide a fading edge for months; this turns first.
Line movement: Every price the books posted for this side at the number you took, from the first capture to the close, the sharp book dashed as the reference. The ring is the price you locked; the dot at the end is the close at your book.

Lines board

Best price: The book paying the most for that outcome at the number most books post. Lit only when two or more books can be compared there; a price at another number is not compared.
Fair price: The price with the books' margin taken out, averaged over the playable and the market books posting both sides at the same number. A best price longer than this is one the market as a whole says is too generous.
Hold (vig): The vig, also called the juice: what a book charges on a market, baked into its prices rather than billed. It is how far the two prices add up past 100 percent; −110 both ways holds about 4.5 percent, so a bettor has to clear that, not zero, to make money. Lower is cheaper to play.
Movers: The numbers that have moved furthest since the line opened, a point counting before a price. A move at one book before the others is where steam shows first.
As of: When the newest line on the board was captured. A price captured more than two hours ago is greyed, because it may have moved since.
Market books: DraftKings, FanDuel, BetMGM, Caesars, BetRivers and LowVig, six US books shown for where the market sits and not as books to bet at, so a best price never names one of them.
Trend: Each book's number on the side chosen, from the first capture on record to the newest; the dashed line is the sharp book, the reference the others are read against and never a price to play.
Held vs. now: The number your open ticket holds against the best price at that number among the four playable books now or, when none of them still posts it, your own book's number now. Green when the market has moved past you, red when it has gone the other way.

Hedging

Hedge: A bet against your own position, taken to lock in a result or cap a loss.
Positions: Every open ticket on this game, and what each one nets under each outcome of the market you are on. The total row is where you stand if that side wins.
Hedge stake: How much to put on the other side at the price entered so that both outcomes pay the same. Less than that leaves a lean; more flips it.
Locked result: What you net whichever side wins once the hedge is placed. It is usually below the ticket's to-win and above its stake lost, because the book takes its cut on both sides.

The bankroll

Open exposure: Money currently sitting on open bets at each book.
Available to bet: Cash the book shows, brought forward by what settled since the last snapshot, less any reserve you set aside.
Account total: Everything at the books: what is available to bet, plus the reserve, plus the stake on open bets. What the books would hold if every open bet were refunded.
Share: Each book's part of the account total: its cash now plus the stake on its open bets. The rows add up to the total.
Put in: Every deposit less every withdrawal: your own money that went into the books and has not come out. The account total less this is the net against deposits. Bonuses and free bets are the book's money and are not in it.
Money in, money out: Every deposit above the line and every withdrawal below it, book by book, from the funding ledger. Bonuses and free bets are the book's money and are not in it.
Net vs. deposits: What the books hold now, open stakes and withdrawals included, less every deposit. Bonuses and free bets are the book's money and do not count.
Reserve: Cash you have chosen to keep off the table. Taken out of what is available to bet.
Snapshot: A balance as you read it off the book at a moment. The chart and the table start from the latest one per book.
Estimated (est.): The last snapshot's cash, plus what settled since and less what was staked since, as far as the record knows.

Cuts of the record

ROI by price: Bets grouped by the price taken. Favourites on the left, underdogs on the right.
ROI by stake: Bets grouped by how much was risked, to show whether bigger bets do better or worse than small ones.
ROI by time before start: How long before the start the ticket was placed, from more than a day out to in play. The line moves most in the last hours, so this is where beating the close is won or lost. The second figure under each column is how often tickets placed that far out beat the close.
By tag: Results by the tags on each bet. A bet with two tags counts under both; one with none is left out.
Dates by: Which day of a ticket the dates are read against: Placed is the day it was bought, Game day the day its game is played (the day it was bought when the start is not known), Graded the day the book settled it.

The research desk

Conviction: How strongly the desk rated its own pick. Since desk v2.1 it is set by how far the desk's own line is from the posted one: Significant (2% of the bankroll), Good (1%), then Thin (0.35%) and Micro (0.25%) only when a slate has nothing bigger. Older tickets carry Thin, Medium or A+. It drives the stake, not a measure of the price.
Value: How much better our number is than the book's, on the pick's side: in runs, goals, points or yards for a total, a spread or a player prop. For a moneyline, a run line or a puck line it is our chance of winning against the chance the price implies, in percentage points: 70% against a price that implies 64% is +6%. A bet needs at least the amount the desk set for that game, shown as Bet at on the line under each pick; the line also shows the book's number, ours, and where ours lands after the desk's stress test.
Public split: Where the public's tickets and money sit on this market, and how the line has moved since it opened. A line moving away from the side most tickets are on means the market leans against the public, and the desk says whether its number agrees.
Form: Both teams' season game by game on a football card: the record and the margins by each team's starting quarterback, yards per play on offense and allowed, the ball given away, and the scores a defense made. A 2-2 can be two different teams, 0-2 with one quarterback and 2-0 with another.
Margins: Both teams' results this season with the margin of each, on a football or basketball side. A record built on one-score losses or garbage-time wins is not the team its record says, so the desk prices the margins, not the record.
Player history: The player's number for this prop's stat in his last three seasons, read before the desk prices it: under the lights and by day, home and away, against this opponent, with this season's starting quarterback, and how often it went over the line.
Upset chance: How likely the moneyline makes it that the side this pick is against wins outright. At 15% or more, the desk's scenarios have to include one where it does, weighted near that chance.
Gap to the market: Why the desk's number differs from the market's: what the market would have to be missing for ours to be right, and whether an adjustment was counted twice by laying it on a rating that already held it.
Standing check: A question every card in this sport answers, added by the morning review after a factor the desk had not weighed changed the result of three of its games in sixty days. It comes off when fewer than three such games remain in the last sixty days.
Our line: The number the desk would hang itself, written from role, health, game script and the beat before it looked at the board, against the line the book posted.
Beat: What the team's own beat reporters and named insiders said in the last day or two. A quiet 48 hours reads "no new beat".
Role: The player's job this week: snaps, targets, carries, routes. A committee with no written split is no ticket.
Health: The injury picture the line was written with: out and IR are off the card, doubtful is not a starter, questionable costs the line 15 to 25 percent.
Number: Whose number the desk's line is for, where a card could be read more than one way.
Gate: The worst number or price the desk would still take. The bot places a play only while the board is inside it; a watch waits for the board to reach it.
Kills it: What the desk said would make it drop the pick before the start. It is checked once, when the ticket is placed; a placed ticket rides to settlement.
Pinnacle (print): The price against Pinnacle's de-vigged fair at the same number. Under desk v2.1 it is a print and a veto, not the reason for a ticket; the gap is.
Wipe: The stress test: the desk's number again after 15 points of likelihood move from the scenario most in the pick's favor to the one most against it. A bet has to clear the bar on this number too, or it is a pass.
Scenarios: The ways the desk thinks the game can go, what each gives (a total, or the result for the pick's side) and how likely each is. Our number is their weighted average. The stress test moves 15 points of likelihood from the scenario most in the pick's favor to the one most against it; its column shows each scenario's chance there.
Mix: The scripts behind the desk's line and their weights, and the arithmetic that blends them into one number.
Plan: One line on what each coach calls, the counter, and the weights the desk gave the scripts.
Thesis: Why the desk's number is what it is and why the book's is wrong in this sport, in one paragraph.
Floor: The gap floors the desk applied to this game, in its own words: Good and Significant for the market on the game's posted total.
Split: The ticket and handle split the desk saw on the game and what it did with it: a shade of half a point, a note, or nothing.
Stake tier: What the stake was cut to while the desk is on probation: a desk version bets Good at most (1% of the bankroll) until 30 tickets of its major version, every v3.x together, have graded with an average closing-line value of zero or better. Until then a Significant edge is bet at 1%.
Last read: When the desk last read this game. Each hourly run reads the games that clear its first screen a handful at a time, the ones read longest ago first, so a game it did not get to this hour shows the read before.
Queued: A game the desk has not read yet. It cleared the first screen (the engine's number against the line) and waits its turn: each run reads never-read games first, soonest start first. Nothing has been decided on it.
Edge: The pick's price against the sharp book's fair price at the same number, in cents on the dollar, green from zero (a fair price); where the sharp book was at another number, the desk's anchored model edge instead, green from 3 cents (2 on an MLB or soccer total). A dash means the pick was not priced.
Model: The desk's probability for this side after the market anchor, against the probability the best price implies once the books' margin is taken out. The difference between them is the edge.
Plays: Picks the desk posted as a play or an add, each with a stake and a gate, and the tickets the ledger already holds that it says to keep; the held ones are counted under the figure.
Holds: The desk's word on tickets the ledger already holds: keep riding. A hold names the ticket it is about.
Watch: Picks the desk likes at a number or a price the board does not show. Each waits for its gate, the line or price that would make it a play.
Passes: Priced picks the desk looked at and left alone. A pass at or above the floor has to name its reason, or the post is refused.

The desk's results

Verdict: The morning review's judgement of our number once the game was final: a good estimate landed nearer the final than the closing line or held for the reasons the read gave, a sound one was right and lost to luck, missed news means something knowable before the start was not in the read, and a bad estimate means the method itself erred.
Desk picks: The desk's cards on the game's own markets, oldest first: the last card on each market, the word that stood at the start, and every play, with its number, price and book, and how that side came out. A side the desk played won or lost; a side it passed or only watched would have, since no money rode on it.
Bets placed: The tickets actually placed on the game, whichever card they came from, at their own number, price and book, with what each paid or lost.
Side, ours against the close: Our margin for the game against the market's and the final, each written as the side ahead and by how much, and which of ours and the closing line landed nearer the final. The lean is the side ours sat beyond the market's line when we wrote it, and whether it covered; in baseball and hockey the margin comes from the moneyline, so the lean is a side to win.
Total, ours against the close: Our number for both sides' scores together against the market's total and the final, and which of ours and the closing total landed nearer. The lean is over or under the market's total when we wrote ours, and whether it came in.
Reasons: The things that could have moved the game, from the review's fixed checklist, each with whether our read weighed it before the start and whether it changed the result. One the read missed that changed the result is what the next read has to fix.
Luck: What no read can price, such as a tipped interception, a fumble's bounce or a goalie who stops everything, with roughly how many points each swung. Luck never counts toward a standing check.
Props, our side: Each player prop the desk read on this game, at its newest read: our number, the line, and the player's number from the final box score. Our side was right when the player's number landed on the same side of the line as ours, whatever the card's own pick; a player who did not play counts for neither.
Next step: What the review says to change before the next read: its lesson, and any standing check the missed factor is part of now. A good or a sound estimate leaves nothing to change.
Win rate: Whether the bets win more often than their prices need. At −110 a bet must win 52.4% of the time to break even, at +150 only 40%. Every record says how far luck alone swings a sample its size: a result past twice that swing is more than luck, one inside it is too close to call.
Needed to break even: The win rate the prices taken need to break even: the average chance their odds imply, the bookmaker's margin included.
Against the close: Each ticket's number against where the line closed: a better number than the close beat it, a worse one is behind it. Beating the close is the earliest reliable sign that bets will win over time, long before the record can show it.
Our number on every game: For every game a desk priced, whether the side its number leaned to against the line won, scored as a flat bet at −110. If the desk's edge is real, the games where its number sat further from the line win more often; a floor is the gap the desk needs before it will bet.
Trials: Betting trials run beside the research desk and are judged apart from it, each on its own track. The price trial bets a side a playable book prices at least 1 cent better than Pinnacle's fair line, the bookmaker's margin taken off, at the same number. The football overs trial bets the over in the NFL and college football where the desk's model total sits 4 or more points above the line. Both stake a flat 0.5% of the bankroll, four bets a day at most and one a game market, and skip a side the desk is waiting on news for. Their tickets are kept out of Your bets and counted here alone.
Against Pinnacle's close: Each trial ticket's price against Pinnacle's closing line with the bookmaker's margin taken off: how many cents of win chance the price beat the fair close by, or fell short of it. A ticket whose number Pinnacle closed off counts as beating or trailing the close by the points the number moved, and is left out of the average.
Stop rule: The rule a trial stops itself by. The price trial stops after 50 settled bets if the average ticket did not beat Pinnacle's close, and from about 100 its win rate is judged against what its prices needed. The football overs trial stops after 30 settled bets below break-even. A stopped trial sends nothing more.
Opening-line paper trial: A trial on paper: nothing is placed and no ticket exists. A formula rates every football team's offense and defense from the play by play of the season so far (how often a snap succeeds, how often it goes for a big gain, yards a snap, each adjusted for the opponents faced, garbage time left out) and turns that into a point spread. Each college and NFL game with a spread at one of your books is picked once, more than a day before kickoff, at the best number a book posts for the formula's side, and graded at that number after the final. The two leagues are judged apart. In a replay of five seasons the college picks beat the opening line every season; the NFL picks did not.
Gap to the line: How far the formula's spread sat from the line when the pick was made. A pick 9 or more points off counts as one of the trial's bets; the rest are recorded to see whether wider gaps win more often, as a real edge would.
Against the close: The number each bet took against the same book's number at kickoff. The line moving past the number taken is the earliest sign a pick was right, long before the record can show it.
Units: What the picks would have won or lost at their own prices, each staking one unit: a win at −110 returns 0.91, a loss costs 1. Paper only.
NFL model: A frozen formula over 34 facts about an NFL game (the ratings, the efficiency numbers, luck so far, rest, the line and the moneyline, the division, the passer, who was missing last game, the week), fitted on the 2022-24 seasons and not changed since. It picks a side only when it gives that side 60% or more against the spread. On the games it was fitted to it went 138-74; on 2025-26, which it never saw, 63-37, carried by 2025 and by November and December. It was the best of 48 such formulas tried, so only its record from here on can say whether it is real. Paper only.
College model: A frozen formula over five numbers about a college game: how far the points ratings, the efficiency formula, the matchup formula, the early-season blend and last season's ratings each sit from the opening spread. Fitted on every college game of 2022-24 and not changed since, it is played only on conference games where it gives a side 55% or more. On those games it went 81-50 where it was fitted and 108-75 on 2025-26, which it never saw; but it was one cell of 2,520 searched and chance alone fills as many cells at that level, so only its record from here on can say whether it is real. Its picks are made when a line is first seen, about a week before kickoff. Paper only.
NBA model: A lineup formula for the NBA: each player's worth comes from four seasons of box scores (how the margin moved with his share of the minutes, last season's rating as the starting point), summed over who is dressing tonight once the injury report is in, against the line inside the last hours before the tip. It plays a side only 4 points or more off the line and sits out a game whose injury report has not been posted. In the replay, with the lineup known, that side covered 55% against the opening spread at one setting of the fit and 50% at another, and 50% against the close, so only its live record can say whether the news reaches it before the books do. Its plays go to the placer at 1% each; the record here is at its own numbers.
Confidence: The chance the model gave the side it took, which is what it was banded by. If the model is real, its more confident picks win more often.
Prop parlays: Paper parlays built each morning from 9:30 AM Eastern out of the prop scan's safest rungs: a rung we give 75% or more by Pinnacle's line and priced at or better than that, one leg a game, the surest first, stacked until a ticket pays about even money (+90 or more). Each leg is graded off the box score and each ticket as the books settle one: any lost leg loses it, a pushed leg drops out and the rest pay. A ticket that pays even money needs to come in about half the time, so the record says whether these legs beat their prices. Paper only: nothing is placed.
Our chance: Our chance of every leg of a ticket coming in: the legs' fair chances multiplied, the games being apart. A ticket beats its price only if it comes in more often than its price needs, which this says it should when it sits above the needed figure.
Units: What the tickets would have won or lost at their own prices, each staking one unit: a win at +108 returns 1.08, a loss costs 1, and a pushed leg pays the rest of the ticket. Paper only.

Alerts and keys

Line moves on held games: An open ticket whose game has not started, when the price at your number moves by the threshold or more, or the number itself moves. Once each way per ticket.
Tickets graded: Every ticket as it grades: the result, the profit or loss, the pick and the book. Once per ticket.
Desk play not placed: A PLAY or ADD the desk first called an hour and a quarter ago or more, still with no open ticket of yours on that side of that game: the betting bot reads the desk's plays once an hour, so it has had its pass. Once per play, however many runs repeat it.
Morning prop parlays: The morning's paper prop parlays as they are built, once a day from 9:30 AM Eastern while their first game is still to start: each ticket's legs, what it pays and our chance against what its price needs. Information only; nothing is placed.
Push alerts: A notification on this device, sent through the browser even when the site is closed. On an iPhone, add linegaps to the Home Screen and turn this on from there.
API key: A key lets your own tools record and read your bets through the API. It can only touch your account and is shown once; only a fingerprint is kept, so a lost key is revoked and replaced.