1 The Landscape: Variants, Formats, Live vs Online

By the end of this chapter you will be able to identify which form of No-Limit Hold’em you are sitting down to play and explain how that choice reshapes what counts as correct strategy.

Before you can study hand reading, ranges, or the psychological battle that runs underneath every decision, you have to know which game you are actually sitting down to play. “No-Limit Hold’em” is not one game. It is a family of related games whose surface rules are nearly identical but whose strategic centers of gravity could not be further apart. A move that prints money in a deep-stacked cash game can end your tournament in a single hand. A read that is gold against a live recreational player is worth almost nothing against a pool of online regulars who are all running the same software you are.

This chapter maps that terrain. By the end you should be able to place any seat you sit in along three axes (cash versus tournament, live versus online, and the format and speed within each) and understand how those coordinates change what “good play” even means. We will also pin down the most important and least glamorous question in the whole book: what does it actually mean to be a winning player, and how big an edge can you realistically expect?

1.1 The two great branches: cash and tournament

Everything in No-Limit Hold’em descends from one fork in the road. Either the chips in front of you are money (cash games, also called ring games), or they are tournament chips that only convert to money according to a payout structure (tournaments). This single distinction drives almost every strategic difference that follows.

Cash games: chips are cash

In a cash game, every chip on the table equals a fixed amount of real money. A $1/$2 No-Limit game has a $1 small blind, a $2 big blind, and you typically buy in for somewhere between 50 and 200 big blinds. The defining features:

You can rebuy at any time. Lose your stack, reach into your pocket, reload to the maximum. There is no “elimination.”
Blinds never rise. The $2 big blind at noon is still the $2 big blind at midnight.
You can leave whenever you want. Pick up your chips, cash out, go home. Your stack is your money.
Stack depth is roughly constant. Good players keep their stacks near the maximum (often 100bb), topping up whenever they drop below it (e.g., after losing a pot), so they always cover the other deep stacks.

Because chips equal money linearly, the math is clean: a chip won is worth exactly as much as a chip lost. This means you should make every decision to maximize your expected value in chips, full stop. If a call shows a profit of even a fraction of a big blind on average, you take it. There is no hidden penalty for risking your stack, because you can always reload.

Key idea

In cash games, chip EV is money EV. Your only job is to find spots where, averaged over all the ways the hand can play out, you win more than you put in. Survival has no independent value; a stack lost is simply reloaded. This is why cash strategy is the “purest” form of poker theory and why most solver study starts here.

Tournaments: chips are not cash

A tournament works differently. Everyone buys in for the same amount, receives the same starting stack of non-cashable tournament chips, and plays until one player has them all. You cannot reload after the early “re-entry” or “rebuy” period ends; once your chips are gone, you are out. The blinds, and in almost all modern structures the antes, rise on a timer (e.g., every 15–30 minutes online, every 30–60 minutes live), relentlessly compressing everyone’s stack depth as measured in big blinds. And only a fraction of the field, typically the top 10–15%, gets paid, with the prize pool concentrated heavily at the very top.

This produces several non-negotiable strategic consequences:

Rising blinds and antes force action. A 100bb stack at the start can be a 20bb stack two hours later without your having lost a single chip, simply because the blinds increased fivefold. You cannot wait for premium hands; the structure itself pushes you to accumulate chips or perish. It is not only the blinds that climb: nearly every modern tournament also posts an ante each hand, and most now use a big-blind ante (a single ante paid by the player in the big blind that covers the whole table). Antes inflate the size of every preflop pot before anyone looks at their cards, which raises the reward for stealing and sharpens the incentive to open, re-steal, and apply pressure. They also drain chips faster each orbit, so effective stack depth shrinks more quickly than the blind level alone would suggest.
Stack depth is a moving target. Over a single tournament you will play deep-stacked poker (100bb+), mid-stack poker (30–50bb), short-stack/push-fold poker (under ~15bb), and everything in between. Each regime has its own correct strategy. We’ll return to this.
Survival has real value. Because being eliminated ends your chance at the prize pool, the chips you would lose are worth more than the chips you would win; each additional chip is worth less than the one before it. This is the heart of ICM. The principle is broader than the all-in-for-your-life case: even when your whole stack is not at risk, near a pay jump a marginal chip-EV gain can still be money-EV-negative. ICM pressure shapes every decision near a pay jump (thin value bets, light calls, loose steals, flat calls out of position), not only the spots where you can be eliminated.

ICM: why a tournament chip is not worth a tournament chip

ICM stands for the Independent Chip Model. It is the standard mathematical tool for converting a tournament chip stack into its real-money equity given the remaining payouts. The crucial insight it encodes is non-linearity: in a tournament, chips have diminishing marginal value. Doubling your stack does not double your equity, because the chips you win are worth less (per chip) than the chips you risk to win them.

Consider a classic illustration. Five players remain in a satellite that pays the top four an identical seat; fifth place gets nothing. You have a comfortable stack that covers the short stack’s all-in. Should you call as a slight favorite, say 55%, against that short stack? In chip-EV terms, calling is profitable. In ICM terms it can be a catastrophe. Because you cover the villain, calling does not put your own tournament life at risk on this hand; what it risks is chips you don’t need to reach a seat. Fold, and the shorter stacks are overwhelmingly likely to bust before you, locking up your seat for free. Call and lose, and you convert a near-certain seat into a live contest. You decline the 55% gamble because a bigger stack wins you the same seat, so the chips you’d win are nearly worthless while the chips you’d risk carry far more real-money (ICM) value than the gamble returns. This is also why the standard satellite play is to let the short stacks collide and bust each other while you fold your way in.

Key idea

ICM means chip EV and money EV come apart. Near a pay jump (the bubble, a final-table ladder, a satellite cutoff) the correct play is often to decline chip-profitable gambles to protect your equity, and to attack opponents who themselves cannot afford to gamble. A move that is mandatory in a cash game can be a clear money-losing error in a tournament, and vice versa.

Common mistake

Treating a tournament like a deep cash game and “playing for stacks” with marginal hands on the bubble. Stacking off 100bb-style with top pair when you’re 25bb deep on a money bubble ignores the enormous real-money cost of busting. Conversely, players who grind one format and dabble in the other often import the wrong instincts: cash players gamble too much in ICM spots, and tournament players play far too tight and passive when they sit in a cash game where there is no bubble to fear.

Tournament sub-formats: MTT vs SNG

Within tournaments, two broad shapes matter:

MTTs (Multi-Table Tournaments). Large fields (dozens to tens of thousands of entrants) playing down to a single winner over many hours. Top-heavy payouts (the winner might take 18–25% of a prize pool that pays 15% of the field). MTTs are among the highest-variance of the traditional formats: you can play flawlessly for eight hours and min-cash, or bust 50 in a row and then bink a score that covers all of them. Long-run results are measured over thousands of tournaments, not hundreds.
SNGs (Sit-and-Goes). Tournaments that start when a fixed number of seats fill rather than at a scheduled time. They come in two strategically distinct shapes. The classic single-table SNG (9- or 6-max, paying the top 2–3 spots) has small fields and flat payouts, so its variance is meaningfully lower than an MTT’s. It is ICM-intensive almost from the first hand and rewards disciplined push-fold and bubble play, with thinner edges and faster structures. Modern online play, however, is dominated by spin/lottery SNGs: 3-handed, ultra-fast, and effectively winner-take-all, with a randomized prize pool that occasionally explodes to hundreds or thousands of times the buy-in. That combination of winner-take-all payouts and a jackpot multiplier makes spins very high variance, frequently exceeding even MTT variance.

1.2 Live versus online: same rules, different sport

The cash/tournament axis defines the math. The live/online axis defines the information environment, the pace, and the opposition, and those differences are so large they essentially make live and online into different sports played with the same deck.

Pace and volume

The single biggest practical difference is hands per hour.

Setting	Approx. hands/hour	Notes
Live home game (manual shuffle)	20–25	No dealer; players shuffle, deal, and handle chips themselves
Live casino (dealer + auto-shuffler)	30–35	Best case live; a pro dealer and a pre-shuffled second deck keep play moving
Online, single table	60–100	Auto-dealing, betting timers
Online, multitabling (e.g., 4 tables)	250–400+	Linear scaling per table
Online, fast-fold pools (e.g., Zoom-style)	200–250 per table	You’re moved to a new table the instant you fold

A live player might see 30 hands an hour. An online grinder running a modest four tables sees ten times that. This has enormous downstream effects: online players reach the long run far faster, but they also have far less time per decision and must lean on heuristics and software rather than in-the-moment soul-reading.

Information: physical tells vs HUD data

This is the difference that most directly feeds the two skills you most want to develop.

Live, your information is physical and behavioral. You can see your opponent’s hands shake when they have the nuts, watch how they cut out a bet, hear the change in their breathing, notice that they reach for chips before it’s their turn, or clock that they always glance at their stack when they’re about to bluff. These reads are powerful but probabilistic and player-specific: a tell that screams “weak” from one player means nothing from the next, and the best live players actively reverse and mask their patterns. (We devote an entire later chapter to hand reading and another to the psychological game; this is where those skills earn their keep.)

Online, you get zero physical information but a flood of statistical information. Tracking software (a “HUD,” or heads-up display) overlays each opponent’s historical tendencies on the table in real time, computed from hands you’ve played against them. The core stats:

VPIP (Voluntarily Put $ In Pot): how often a player enters a pot. A typical solid 6-max reg sits around the low-to-mid 20s in percent; recreational players are often 35%+.
PFR (Pre-Flop Raise): how often they raise pre-flop. For a tight-aggressive reg this trails VPIP by only a few points (e.g., 22/19); a large gap (e.g., 40/8) screams a passive, calling-station recreational player.
Other common stats include 3-bet % (how often a player makes the first re-raise before the flop), fold-to-3-bet, c-bet % (continuation-bet frequency: how often the preflop raiser bets the flop), fold-to-c-bet, and aggression frequency, plus dozens more for the dedicated grinder.

These numbers are population- and sample-dependent, so treat the figures above as rough guides, and remember that small samples lie. A “65% fold to 3-bet” over 12 hands tells you almost nothing.

An important 2026 caveat: the rich-HUD environment described above applies only where tracking is actually permitted. Many of today’s major sites now ban third-party HUDs outright, and several, including the largest networks, seat players at anonymous or temporary-alias tables where no persistent opponent history can be built. On those sites a historical HUD simply does not exist. Your reads revert to in-session observation (bet-timing, sizing, bet-sequencing, and whatever basic table stats the site surfaces), much as a live player works from behavior, only without the physical tells. That narrows online’s statistical-information advantage but does not erase it: you still reach a useful read faster than live, you just build it within the session rather than from a lifetime database.

Common mistake

Trusting a HUD stat over a tiny sample. Early in a session your reads on a new opponent are dominated by noise; a 3-bet% computed from 8 hands is essentially random. Equally, live players over-weight a single dramatic tell (“his hand shook, he must be strong”) while ignoring the player’s overall range and the board. Both errors share a root cause: mistaking a noisy signal for a reliable read. Weigh every read as evidence rather than treating it as a verdict.

Rake and the time charge

Poker rooms make money by taking a cut. How they take it differs by venue and quietly determines whether a game is beatable at all.

Online cash and most tournaments use a percentage rake: typically around 5% of each cash pot up to a capped maximum (the cap is what makes higher stakes relatively cheaper), and a tournament fee tacked onto the buy-in (e.g., a “$50+$5” tournament rakes $5).
Live cash is usually either a percentage drop (e.g., 10% up to a $5–$8 cap per pot, much heavier than online) or a time charge (“time”), where each player pays a fixed fee per half hour (say $6–$14) regardless of how many hands are dealt.

The time charge interacts brutally with live’s slow pace. If you pay $10 per half hour and see 15 hands in that time, you’re paying roughly $0.67 per hand just to sit there, before you’ve put a chip in a pot. At low live stakes the rake or time charge can be a larger drag on your winrate than at comparable online stakes, a fact many live players never properly account for.

Key idea

You don’t beat the players; you beat the players by more than the rake. Rake is the house’s guaranteed share of every dollar that crosses the table. Your true edge is your edge over your opponents minus what the house extracts. In a soft but heavily raked low-stakes live game, or a high-rake micro online game, the rake can convert a small skill edge into a long-run loss. Always measure your edge net of rake.

But rake is only one side of the ledger. Most sites and rooms return a portion of what you pay through rakeback, VIP, and loyalty programs, and large guaranteed MTTs occasionally overlay, meaning the guarantee exceeds the prize pool the entries actually funded and the field gets free equity. The figure that truly matters is therefore your rake net of rewards. This is not a rounding error: at micro and low stakes online, where edges are thin and rake is proportionally heavy, a large fraction of grinders are roughly break-even at the tables and profit only from rakeback and loyalty rewards. Fold those rewards into any honest winrate calculation, exactly as you fold in the rake. A player who ignores rakeback will badly misjudge whether their game is actually beatable.

Player-pool softness and multitabling

The reason many serious players still prefer live despite the slow pace and heavy rake is simple: the games are softer. Live poker attracts recreational players (people there for a night out, tourists, gamblers), and the friction of physically traveling to a casino keeps the dedicated, study-hardened grinders relatively scarce, especially at low stakes. Reads are deeper because you face the same opponents for hours.

Online pools are, on average, tougher and getting tougher: the best players multitable, use tracking software, study with solvers, and the population’s baseline competence has risen for two decades. But online offers the decisive advantage of volume. Multitabling lets a winning player turn a thin per-hand edge into a large hourly rate by simply playing far more hands. The trade-off is direct: more tables raise your hourly volume but lower your edge per table, because divided attention means you read situations less deeply and exploit less. The optimal table count is the one that maximizes (winrate per table) × (tables), not the maximum you can physically click.

1.3 What does “winning” actually mean?

If you take one practical lesson from this chapter, make it this: poker results are noisy, and you cannot evaluate yourself (or set realistic goals) without the right yardsticks and a respect for the long run.

Cash: bb/100

The standard cash-game winrate metric is big blinds won per 100 hands, written bb/100. (Many trackers use “bb” to mean the big blind; some older conventions use “BB” for big bets = two big blinds, so know which your software means.) Rough, honest benchmarks for a player who is genuinely beating their game, net of rake:

A small but real winrate is on the order of a few bb/100.
A strong winrate at mid stakes might be 5–10 bb/100.
Anything sustained above ~10 bb/100 over a serious sample usually means you are crushing a very soft game or your stakes are low relative to your skill; it rarely persists as you move up.

The brutal companion fact is variance. The standard deviation in No-Limit Hold’em cash is roughly 80–100 bb/100. That number dwarfs the winrate. The consequence: a genuinely winning player can run below break-even over tens of thousands of hands through nothing but variance. Meaningful confidence in a cash winrate takes hundreds of thousands of hands.

Tournaments: ROI and ITM

For tournaments, two metrics matter:

ITM% (In The Money): the fraction of tournaments in which you reach the paid places. Because typically only ~15% of a field cashes, even excellent MTT players post ITM rates only modestly above the field average, and chasing a high ITM by playing scared actually lowers your profit, because the money is at the top.
ROI (Return On Investment): total profit divided by total buy-ins (including fees), as a percentage. This is the real measure of MTT skill. A solid online MTT reg might run a long-term ROI in the low double digits (e.g., 10–20%); live MTTs, with their softer fields, can support higher ROIs but come with far smaller samples.

Tournament variance makes cash variance look gentle. Top-heavy payouts mean your results are dominated by rare deep runs. Downswings of 100–200+ buy-ins are normal for winning MTT players, and a meaningful read on your ROI requires thousands of tournaments. This is why bankroll requirements for MTTs are measured in hundreds of buy-ins, not the 20–40 typical for cash.

Common mistake

Judging your skill (or your strategy changes) by short-term results. “I switched to a more aggressive 3-betting range and lost four buy-ins, so it must be wrong” is statistically meaningless reasoning. In a game with a standard deviation near 100 bb/100, four buy-ins is noise. Evaluate decisions by their logic and EV, and evaluate results only over samples large enough to overcome variance. Outcome and decision quality are only loosely correlated in the short run.

Where edges come from today, and realistic expectations

The modern game is more studied than ever, so it’s worth being clear-eyed about where a competitive edge actually lives in 2026:

Fundamentals executed consistently. Most players, even “regulars,” leak EV constantly — tilting, mis-sizing, over-folding to aggression, paying off too light. Solid, disciplined, fatigue-resistant fundamentals beat most of the population by themselves.
Exploitation. GTO (Game Theory Optimal) play is unexploitable, but it is not maximally profitable against opponents who make mistakes, and everyone makes mistakes. The largest edges come from correctly reading how a specific opponent or pool deviates from balanced play and deviating in response: value-betting thinner against a station, bluffing more against a nit, declining ICM gambles against players who can’t fold. This is the “play the player” thread that runs through the whole book.
Game selection. The single highest-EV skill in poker is choosing soft games. A mediocre player in a great game beats a great player in a tough game. Table selection, seat selection, and stake selection are not optional extras; they are core strategy.
Mental game and stamina. Over the volume required to realize an edge, the player who tilts least, quits worst games soonest, and keeps deciding clearly in hour six wins. We treat this at length later; for now, simply file it as a genuine, durable source of edge.

Realistic expectations: for a dedicated, studying player, beating low-to-mid stakes for a solid winrate is achievable; turning that into a livelihood requires volume, bankroll discipline, and tolerance for swings that would unnerve most people. Poker is a game of small edges magnified by repetition. You are not trying to win every hand or every session; you are trying to make better decisions than your opponents, again and again, and let the long run pay you.

1.4 A worked example: the same hand in two formats

Let’s make the cash/tournament gulf concrete with a single decision.

The situation. It folds to the player on the button (BTN, the seat one to the right of the blinds, which acts last after the flop), who opens to 2.5bb. You’re in the big blind (BB) with A♠Q♦, and effective stacks (the shorter of the two stacks in play, which caps how much money can actually go in) are 25bb. The BTN is a competent, aggressive regular. You 3-bet to 10bb, the BTN moves all-in for 25bb, and the action is back on you. Calling costs another 15bb to win a final pot of ~50bb (50.5bb once you count the dead small blind), so you need roughly 30% equity to call profitably in chips. Your equity depends heavily on which reg you’re facing. Against a sensible, balanced 25bb shoving range whose jam includes a healthy share of bluffs (broadly, big pairs, AK/AQ, plus suited-ace and broadway semi-bluffs), A♠Q♦ runs roughly 38–43%, comfortably above the ~30% the pot odds demand. Against a value-heavy, nittier reg who jams few or no bluffs, essentially {AA, KK, QQ, AK}, the picture flips. Your blockers help (holding the A♠ removes a chunk of his AA and AK combos), but A♠Q♦ is dominated or a clear underdog across that whole range: combo-weighted, you hold only about 25% equity, and even if he widens to include JJ/TT it climbs to just ~29–32%. That sits at or below the ~30% the pot odds demand, so against a genuine no-bluff nit the call is marginal at best and can be slightly chip-losing. The conclusion is therefore not opponent-independent: the call is robustly +EV in chips against the balanced, bluff-laden jam, but close to losing against the pure-value nit. Even before we add a single drop of ICM, the chip-EV read already depends on the player.

In a cash game, against that aggressive reg’s bluff-laden jam this is an easy call, close to automatic. Chips are money; with ~38–43% equity you sit well over the required ~30%, so you call and reload if it doesn’t go your way. Folding here would be a clear, money-losing leak. (Against a pure value-jamming nit the call is far closer to break-even, but that reflects the opponent, not the cash format.)

In a tournament, it depends entirely on the context:

Early MTT, deep field, no bubble pressure. ICM is negligible this far from the money (your stack’s chip value and money value are nearly linear), so you should play it almost exactly like the cash hand and call.
On the money bubble, or near a steep pay jump, with several shorter stacks at other tables about to bust. Now ICM bites. The chips you’d win are worth less than the chips you’d lose (busting forfeits all your equity right before a pay jump you were likely to reach), so the required real-money equity to call rises above the ~30% the chip odds demand. How far it rises depends on the pay structure. A regular MTT min-cash bubble is a relatively gentle pay jump (the min-cash is often only ~1.5x the buy-in), which typically nudges the required equity into the high-30s to low-40s. Since A♠Q♦’s chip equity sits in that same range, the hand becomes a close, marginal decision: a call against a bluff-laden jam, a fold against a value-heavy one, rather than an obvious fold. It is in satellites (where every survivor wins the same seat, so chips above the bubble are nearly worthless) and in very steep pay-jump spots (a final-table ladder, a super-satellite cutoff) that the required equity climbs well past 45%. There, folding a hand this strong despite good chip odds is the disciplined, clearly correct play, because survival is worth far more than the chips on offer.

Same two cards, same board-less all-in, same pot odds. The cash player calls without thinking. The tournament player first has to ask where in the structure am I, and what is survival worth right now? The answer may point the opposite way. That question, in a thousand forms, is what separates the formats.

Drill

For one week, before every session, write down on a single line: (1) the format (cash / SNG / MTT), (2) live or online, (3) your effective stack in big blinds, and (4) the dominant force on your decisions right now: chip EV (cash, or an MTT far from the money) or ICM/survival (bubble, final table, satellite). Then, after any all-in or near-all-in decision, note whether you reasoned in chip EV or money EV. The goal is to make “which game am I actually playing?” an automatic first question, the foundation every later chapter builds on.

1.5 Summary

No-Limit Hold’em splits first into cash and tournaments. In cash, chip EV equals money EV; in tournaments, ICM makes a chip worth less the more of them you hold, so survival carries its own value near a pay jump.
Live and online share the rules but differ in pace, information, and opposition. Live trades volume for softer games and physical tells; online trades soul-reading for speed and statistics, where tracking is permitted.
Rake, time charges, and rewards decide whether a soft game is actually beatable, so always measure your edge net of both.
Results are noisy. Judge decisions by their logic and EV, judge winrates only over large samples, and treat game selection and mental stamina as core sources of edge.