Where to Practice These Strategies
Every strategy above assumes accounts at more than one sportsbook — line shopping, arbitrage and middling are impossible without them. These five books from our tested lineup are the practical starting kit: deep markets to shop, and enough price disagreement between them to matter.
-
1
BetOnline
50% up to $250 Free Bets + 100 Free Spins
Bet Now
-
2
Bovada
75% up to $750 Crypto Sports Welcome Bonus
Bet Now
-
3
Everygame
100% up to $500, twice (up to $1,000 total)
Bet Now
-
4
BetUS
100% up to $2,000 (+25% in casino)
Bet Now
-
5
MyBookie
100% Bet Back up to $500 + 50 Free Spins
Bet Now
These are offshore sportsbooks licensed outside the United States; no US state licensing is claimed or implied. Check each book's availability in your state — Bovada excludes 16 states and DC, and MyBookie excludes Michigan.
Betting decisions start with four numbers: implied probability, vig, expected value and unit size. Everything below builds on those four. If you're still choosing where to bet, start with our rankings of the best sports betting sites; this page is about what to do once your account is funded.
- Bankroll management and unit sizing
- Understanding value & implied probability
- Line shopping: the cheapest edge in betting
- Expected value, worked through
- xG-based betting for soccer and hockey
- Arbitrage & middling
- Live betting strategy
- Parlay math: why parlays favor the book
- Prop & futures strategy
- Record-keeping & taxes
- Common mistakes
- Where to practice these strategies
- FAQ
Bankroll Management: The Strategy That Makes Every Other Strategy Possible
A bankroll is money set aside exclusively for betting — an amount whose total loss would annoy you but change nothing about your life. Every stake is then expressed as a percentage of that bankroll, called a unit. The standard rule: one unit equals 1–3% of your bankroll, and most bets are one unit.
Worked example on a $1,000 bankroll at a 2% unit:
- Your unit is $20. A normal play risks $20; a high-confidence play might go to $30 (1.5 units); nothing ever exceeds $50–$60 (the 3% ceiling).
- A brutal 10-bet losing streak — which will happen; at a 52% win rate a 10-bet skid has roughly a 1-in-1,500 chance on any given run, and over thousands of bets you'll hit several — costs $200. You still hold $800 and your unit drops to $16 if you resize.
- Compare the bettor staking 10% per game: the same streak costs them 65% of their roll (compounding $100, $90, $81…), and the psychological wreckage usually finishes the job. Same picks, same luck — only the sizing differed.
Two refinements. Flat staking (every bet the same size) is the right default — it's immune to the "I'll win it back on this one" spiral. Proportional staking (recalculate your unit as the bankroll moves, e.g. always 2% of the current total) compounds gains and softens drawdowns, at the cost of slower recovery. Kelly-style staking, which sizes bets to your estimated edge, is mathematically optimal but punishes overestimated edges brutally — if you use it at all, use a half or quarter Kelly. What matters most isn't which scheme you choose; it's that stake size is decided by a rule, before the game, every time.
Understanding Value & Implied Probability
Every betting line is a probability wearing a disguise. Converting odds to implied probability is the single most useful skill in sports betting, because a bet is only good when your probability beats the book's.
The conversion for American odds: for negative odds, implied probability = odds ÷ (odds + 100), using the absolute value; for positive odds, it's 100 ÷ (odds + 100). Here's the table worth memorizing:
| American odds | Implied probability | $100 bet returns (profit) |
|---|---|---|
| -200 | 66.7% | $50 |
| -175 | 63.6% | $57.14 |
| -150 | 60.0% | $66.67 |
| -125 | 55.6% | $80 |
| -110 | 52.4% | $90.91 |
| +100 | 50.0% | $100 |
| +110 | 47.6% | $110 |
| +125 | 44.4% | $125 |
| +150 | 40.0% | $150 |
| +175 | 36.4% | $175 |
| +200 | 33.3% | $200 |
Where the vig hides
Add up both sides of a standard point spread priced at -110/-110: 52.38% + 52.38% = 104.76%. Real probabilities sum to 100%; the extra 4.76 points is the vig (or juice) — the book's built-in margin. As a share of all money wagered, the book's theoretical hold is 4.76 ÷ 104.76 ≈ 4.5%. That's the tax every spread bettor pays, and it's why you must win 52.38% of -110 bets just to break even. A "value bet" is simply any wager where you believe the true probability exceeds the implied probability by more than that tax.
To find a market's no-vig (fair) price, divide each side's implied probability by the total. A -110/-110 market de-vigs to 50%/50% — fair odds of +100 each side. When you see a book hang +105 on a side whose de-vigged consensus price elsewhere is +100, you've found value without handicapping a single play.
Line Shopping: The Cheapest Edge in Betting
Different sportsbooks hang different prices on the same game — their traders disagree, their exposure differs, and their updates lag each other. Holding accounts at several books and always taking the best available number is called line shopping, and it is the closest thing to free money in this industry.
Worked example. The same NFL side is priced -110 at Book A and -105 at Book B. To win $100 you risk $110 at A but only $105 at B. Trivial? Run it over 100 bets of $100-to-win-$100, winning a solid-but-human 53:
- At -110: 53 wins × $90.91 − 47 losses × $100 = $4,818 − $4,700 = +$118
- At -105: 53 wins × $95.24 − 47 losses × $100 = $5,048 − $4,700 = +$348
Identical picks, identical results — and the -105 bettor made $230 more, nearly tripling the profit. The same mechanism moves your break-even rate: -110 requires 52.38% winners, -105 only 51.22%. Five cents of price bought more edge than most handicapping ever will. This is also the one strategy with zero downside: it requires no skill, no model and no prediction — only the discipline of checking two or three books before every bet. Offshore books rarely limit winners as aggressively as regulated US books do, which makes multi-book line shopping especially practical with the lineup on this site.
Expected Value, Worked Through
Expected value (EV) is the average profit or loss of a bet if you could replay it thousands of times. The formula: EV = (win probability × amount won) − (loss probability × amount staked).
Worked example. A book offers a team at +150 — implied probability 40%. Your analysis (injury news the market hasn't digested, say) makes them 45% to win. Staking $100:
EV = (0.45 × $150) − (0.55 × $100) = $67.50 − $55.00 = +$12.50
This bet loses 55 times out of 100 — and is still an excellent bet, earning an average of $12.50 per $100 staked. That's the mental leap that separates profitable bettors from the rest: individual results don't grade bets; prices do. A losing +EV bet was a good bet; a winning −EV bet was a bad one that got lucky.
Flip it to see the house side. Take +150 when the true probability is its implied 40%: EV = (0.40 × 150) − (0.60 × 100) = $60 − $60 = $0 — fair. Take it when the true chance is only 37%: EV = $55.50 − $63.00 = −$7.50. Every bet you place sits somewhere on this line. Strategy, reduced to one sentence, is: only bet when your number says the price is wrong, and size it so variance can't kill you while you wait to be proven right.
xG-Based Betting for Soccer and Hockey
Expected goals (xG) is a statistic that values every shot by the probability it becomes a goal — a tap-in might be 0.6 xG, a 30-yard screamer 0.03 — based on location, angle, shot type and build-up. Summed over a match or a season, xG measures the quality of chances a team creates and concedes, stripped of finishing luck. Public sources cover major soccer leagues (sites like Understat and FBref) and the NHL (Natural Stat Trick, MoneyPuck), so this is no longer a sharps-only dataset.
Why bettors care: goals are noisy, chances are stable. A soccer team that's won four straight 1-0 games while being out-created 1.8 xG to 0.7 per match is running hot, and the betting market — which still overweights results and league tables — will price them like a strong side. The xG bettor sees a regression candidate and fades them: against the spread (Asian handicap), on the moneyline against mid-table opposition, or via team totals. The same logic runs in reverse for a team losing close games while dominating chance creation — their price will be longer than their performances justify.
A practical workflow:
- Compare each team's goals scored/conceded to their xG for/against over the last 8–10 matches. Flag gaps bigger than about 0.3 goals per game in either direction.
- Check the why before betting the gap: a clinical elite finisher or a top goalkeeper can sustain overperformance; a mid-table team shooting 25% on target-to-goal cannot. In hockey, check goaltender save percentage above expected — a hot backup inflates results fastest.
- Bet the correction where the market is slowest: totals and team totals reprice slower than moneylines, and smaller leagues reprice slower than the Premier League.
xG won't hand you winners weekly — it hands you a systematically mispriced population of teams. Combined with the EV framework above (bet only when the xG-adjusted probability beats the implied probability), it's one of the few genuinely underused edges left in public data: almost no mainstream betting site teaches it, and few bettors at offshore books apply it.
Arbitrage & Middling: Betting Both Sides on Purpose
Arbitrage: the guaranteed-profit case
When two books disagree enough, you can back both sides and profit whichever way the game goes. The test: convert both prices to implied probability; if they sum to less than 100%, an arb exists.
Worked two-book example. Book A hangs Team A +120; Book B hangs Team B -105 on the same market. Implied probabilities: 45.45% + 51.22% = 96.67% — an arb worth about 3.4%. With $1,000 total, split stakes in proportion to each side's implied probability:
- Stake $470 on Team A at +120 → returns $470 × 2.20 = $1,034 if Team A wins
- Stake $530 on Team B at -105 → returns $530 × 1.952 = $1,034 if Team B wins
Either way you collect $1,034 on $1,000 staked: $34 guaranteed profit, no prediction required. (General stake formula: each stake = total × that side's implied probability ÷ combined implied probability.)
Now the honest part — the risks that make "guaranteed" conditional:
- The line moves between your bets. You get +120 down at Book A, and by the time you reach Book B the -105 is -115. The arb has shrunk or inverted and you're holding one naked side. Always place the side likelier to vanish first.
- One leg gets voided. A postponement, an obvious-error ruling or a rule mismatch between books (overtime included vs excluded, for example) can cancel one bet and leave the other live. Read settlement rules before arbing player props especially.
- Limits and account closures. Books' terms allow them to limit stakes or close accounts of suspected arbers. Offshore books are generally slower to limit than regulated ones, but repeated max-bets on off-market prices get noticed everywhere.
- The margin is thin. At 3.4%, a single botched leg can erase ten clean arbs. Most real-world arbs are 1–3%.
Middling: the lottery-ticket version
A middle is an arb's aggressive cousin: instead of locking a small profit, you bet both sides at different numbers and hope the result lands between them. Example: you took Over 44.5 (-110) for $110 early in the week; by Sunday the total is 47.5, and you add Under 47.5 (-110) for $110. If the game scores 45–47, both bets win: +$200. Any other result wins one and loses one: −$10 (the vig). Risking $10 to win $200 means the middle profits long-term if it lands better than once in 21 attempts — and a three-number window straddling key football numbers like 45 and 47 does. Middles come to you via line movement; the skill is betting early numbers at books that post them first, then recognizing when the market has moved far enough to close the trap.
Live Betting Strategy
In-play markets reprice on every snap, which creates the two edges pregame betting can't: overreaction to single plays, and your ability to watch one game while the book's algorithm prices hundreds. The core plays — hedging a winning position into guaranteed profit, fading momentum-driven overshoots against your own pregame number, and middling totals that move mid-game — all run on the same EV math covered above, executed in second-long windows.
Live betting is enough of a discipline that we built it a dedicated page: our guide to the best live betting sites ranks the eight offshore consoles that handle in-play action best, with worked examples of hedging, cash-out pricing and an NFL drive's play-by-play line movement. Strategy note for this page: live betting magnifies whatever you already are. A disciplined value bettor gets more chances per game; a tilting chaser gets more rope. Set your unit rules before kickoff — they're harder to invent at the two-minute warning.
Parlay Math: Why Parlays Favor the Book
Parlays are the book's favorite bet, and the math explains why: every leg carries the full vig, and multiplying legs compounds it.
Worked example with a three-leg parlay, each leg a standard -110 spread. Each -110 leg pays 1.909 (decimal); the parlay multiplies: 1.909 × 1.909 × 1.909 = 6.96, i.e. about +596 (5.96-to-1). Now price it fairly: if each leg is a 50/50 proposition, the parlay wins 0.5³ = 12.5% of the time, and fair odds on a 12.5% event are 7-to-1. The book pays 5.96 for a 7.00 event:
EV = (0.125 × $596) − (0.875 × $100) = $74.50 − $87.50 = −$13 per $100 — a 13% house edge, versus 4.5% on the same bets placed singly.
Each added leg makes it worse: a five-leg parlay of -110 bets carries roughly a 20% edge for the book. That's why books promote parlays relentlessly and why "same-game parlay" products are their highest-margin shelf — the correlations inside them are priced by the book, not by you.
Are parlays ever defensible? Two narrow cases. Genuinely correlated legs that a book misprices (a heavy favorite with the under in bad weather) can turn a parlay +EV — but books now price obvious correlations in. And a small, fixed "entertainment parlay" budget outside your unit system is a hobby expense, not a strategy — fine, as long as it's labeled honestly in your records. What's never defensible is making parlays the core of a plan to profit: the compounding vig guarantees the math runs against you faster than skill can run with you.
Prop & Futures Strategy
Player props and futures are the softest markets a book runs — and the most quietly expensive ones.
Props get a fraction of the modeling attention the main lines get. A sportsbook's NFL side price is sharpened by millions in professional action; its third-string tight end receiving-yards prop is not. That cuts both ways: prop prices are more often wrong, and prop limits are far lower, which caps how much a sharp can extract. Strategy: specialize narrowly (one sport, one stat family), build or borrow a projection, and compare it to the de-vigged prop price using the EV math above. Offshore books also carry prop menus regulated books legally can't — notably props on college athletes plus politics and entertainment markets — where public money is most emotional and prices are softest.
Futures look fun and price terribly. Sum the implied probabilities of every team on a championship futures board and you'll typically find 130–160% — a hold several times the 4.5% on a spread, hidden by the long odds. Add the capital cost: a futures stake is frozen for months while the same bankroll could compound through weekly value bets. If you bet futures anyway, bet them when prices are most dislocated (immediately after a public overreaction — a star injury, an 0-2 start), shop the number across books since futures boards diverge wildly, and remember the hedging play from our live betting guide: a longshot futures ticket that reaches the final can be hedged into a guaranteed profit. A $100 ticket at +1200 facing a -110 final: hedge about $681 on the opponent and you lock in roughly $519 whichever side lifts the trophy.
Record-Keeping & Taxes: The Unglamorous Edge
You cannot improve what you don't measure, and in betting you can't even know whether you're good. Luck over a 50-bet sample is loud enough to make losing strategies look brilliant. A real record is a spreadsheet with one row per bet: date, sport, market, side, odds taken, closing odds, stake, result, profit/loss and book. Two numbers it unlocks:
- Closing line value (CLV). Did the odds you took beat the final pre-game price? If you consistently take +150 on lines that close +130, you are beating the market even before results confirm it — CLV over a few hundred bets is the earliest reliable signal of real skill, far faster than win-loss records.
- Leak detection. Your log will show things your memory hides: that you're profitable on NFL sides and bleeding on NBA totals, or that your live bets after losses are twice your normal unit. Cut what the data condemns.
The same log is your tax record. US law treats all gambling winnings as taxable income — federal and usually state — and offshore books issue no W-2G, so the entire reporting burden is yours. Crypto adds a wrinkle: spending or converting crypto to bet is a disposal with its own capital gain or loss to track. Keep deposits, withdrawals and per-session results documented, and see the tax section of our sports betting pillar guide for the fuller picture. None of this is tax advice; a professional's hour is cheap against an audit.
Common Mistakes That Undo Everything Above
- Chasing losses. Doubling stakes after a bad night is the single fastest bankroll killer. The fix is mechanical, not moral: units are set before the session, and the session ends when the stop-loss hits.
- Betting every game. Edges are scarce. A bettor with 40 positive-EV bets a month and 200 total bets a month is a losing bettor — the 160 coin flips at -110 eat the profit. No number is not a position; it's the default.
- One sportsbook account. As the line-shopping math showed, a single-book bettor donates about $230 per hundred bets against a two-book bettor with identical picks.
- Grading bets by results. A +EV bet that lost was a good bet. Re-grade decisions by the price you got versus the close, or variance will teach you the wrong lessons with total confidence.
- Ignoring bonus terms. A 100% bonus at 40x rollover is $20,000 in forced wagers on a $500 bonus. Sometimes declining the bonus is the +EV move — do the rollover math before you opt in, not after.
- Betting under the influence — of anything. Alcohol, tilt, fandom, a tout's lock of the century. If the stake wasn't justified by a number you wrote down beforehand, it wasn't strategy.
Kelly staking and the cost of overconfidence
Work a small edge before staking
At decimal odds 2.00 and a genuinely estimated 52% win probability, the full Kelly fraction is (2.00 × 0.52 − 1) ÷ (2.00 − 1) = 4% of bankroll. Half Kelly is 2%. If the true probability is 50%, the edge vanishes. Your probability estimate matters more than the formula.
Record the closing line as a check
Log your bet price and the market price just before the event begins. Repeatedly beating the closing price can be evidence that your process finds value, but it does not guarantee a positive result. Record sample size, sport and market; do not promote a ten-bet streak into proof of an edge.
How we compare the sites on this page
- Start with the operators in the relevant lineup and remove any brand excluded in the page's state.
- Compare the same decision points for each operator: odds, stake limits and settlement rules; record an unavailable field as unverified rather than estimate it.
- Calculate any published wagering requirement using the correct bonus base and explain a worked example.
- Check the practical route from registration and deposit through verification and withdrawal; separate published terms from a timed test.
- Recheck time-sensitive offers and restrictions when updating the page, and correct the ranking when evidence changes.