How to Read Poker Players

Reading poker players is baseline work: watch every showdown, learn what each opponent's normal looks like — sizes, tempo, posture, talk — and treat deviations from that normal as information. Bet-sizing and timing patterns carry most of the signal; physical tells are a thin final layer, useful only on top of sound fundamentals. That's the whole method in one paragraph. Everything else in this essay is the evidence for it, the mechanics of it, and the honest accounting of what it can and cannot do.

Notice what the method is not. It is not eyebrow-twitches. It is not the movie scene where the hero stares into a man's pupils and announces his hand. The popular version of "reading people" is the least reliable version, and the sooner you demote it, the sooner the real skill starts compounding. But — and this matters just as much — the fashionable modern dismissal, the one that says reads are folklore and only the frequencies are real, is also wrong, and it's wrong in a way that costs money. The greatest players who ever lived built careers on reading specific human beings, and the skill they used was never magic. It was accumulated observation, cashed in at the right moment. That's trainable. Here is how.

The Honest Hierarchy: What Actually Carries Signal

Not all reads are equal, and most of what's written on this subject fails by flattening the hierarchy — a listicle of eleven tells, the shaking hand next to the bet-sizing pattern as if they were peers. They are not peers. Rank them by reliability and the order is stark.

At the top: betting patterns. How a player sizes his bets, in which spots, against which textures — this is the densest, most durable information at any table, live or online. Sizing is a decision, repeated hundreds of times a session, and decisions leak style. The man who bets two-thirds pot with his strong hands and half pot when he wants you gone is telling you his hand every time he reaches for chips, and he will keep telling you, because sizing habits are wired deeper than faces. A player can compose his expression. Almost nobody composes his sizing under pressure.

Just below that: timing. The tempo of a decision is a read on the decision itself. Snap actions tend to mean the choice was easy — which is information about what kinds of hands make that choice easy. Long tanks followed by aggression mean something different from long tanks followed by calls. Timing is harder to fake than a face because faking it requires holding a constant tempo across every hand, forever, and hardly anyone does.

At the bottom — genuinely useful, genuinely last: physical and verbal tells. They exist. They are real. They are also rare among competent players, strongest among recreational ones, and worth acting on mostly at the extremes and in big pots.

This is not just my ordering. It is, engaged honestly, the ordering of the person who wrote the standard texts on the subject. Zachary Elwood — a former professional whose trilogy Reading Poker Tells (2012), Verbal Poker Tells (2014), and Exploiting Poker Tells (2017) is the serious literature here — is at pains to tell his own readers that tells are a supplement to fundamentally sound strategy, not a replacement for it, and that reliable tells cluster among weaker players and in specific, bounded situations. The leading authority on physical tells ranks physical tells below the fundamentals. Sit with that before you buy a course on face-reading.

The academic record points the same direction, with caveats worth naming. The broad psychology of deception is humbling: Bond and DePaulo's sweeping 2006 meta-analysis of lie-detection studies found people distinguish truth from lies at about 54 percent — barely above the coin flip. Faces, it turns out, are where humans are best defended. A small 2013 study in Psychological Science by Michael Slepian and colleagues made the point almost too neatly: observers shown two-second clips of World Series of Poker players judged hand strength below chance from faces alone — professionals' faces actively misled — but above chance from arm motions alone. The smoothness of the reach for chips carried signal the face had been trained to hide. Lab studies like this have drawn fair criticism from practitioners, Elwood among them, and I wouldn't build a strategy on two-second clips. But the shape of the finding matches what serious players report from decades at the table: the money is in the motion, the sizing, the tempo — the things a player does rather than the things he shows.

So the hierarchy stands: patterns first, timing second, body last. Now the method that turns the hierarchy into money.

The Baseline Method: Build the Person's Normal

A deviation is only visible against a normal, and this is the discipline almost everyone skips: you cannot read a player you haven't watched. The single highest-value habit in live poker costs nothing and is practiced by almost no one — watch every showdown, whether or not you were in the hand.

A showdown is the answer key. The hand gets turned over, and suddenly every decision that player made on every street is annotated: that was his flop call with second pair, that was his turn sizing with a draw, that was the speech he gave with the nuts. One showdown is an anecdote. Ten showdowns from the same player is a baseline — his opening sizes, his tempo when comfortable, his posture when not, what he does with his hands, whether he talks, how fast he reloads. You are not looking for tells yet. You are building the person's normal, the way Elwood's books insist and the way every strong live player does without being told: first learn what relaxed looks like on this specific human, because pressure only shows as a departure from relaxed.

Then you read the departures. The reg who has bet 30 dollars into 45 all night and suddenly makes it 55. The talker who goes quiet. The quiet one who starts chatting at you mid-hand. The instant call from the player whose calls usually take five seconds. None of these mean anything in the abstract — the internet's great sin is publishing dictionaries of tells as if "hand over mouth" translates to the same word on every human. On one player a fast bet is a bluff; on another it's the nuts; on a third it's nothing but impatience. The dictionary doesn't exist. What exists is this person's dictionary, which you write yourself, showdown by showdown, and which nobody else at the table owns unless they did the same work. That's the entire appeal, and it's the reason this skill still pays in a way memorized strategy no longer does: an edge everyone can download stops being an edge. The history of how solvers raised the floor and flattened the field is its own essay, but the conclusion lands here — the shared answers became the ante, and the unshared work, the person-specific file you build with your own attention, became the last thing the field can't copy.

One more discipline, because baselines rot: people change within a session. The same man is different stuck two buy-ins than fresh, different at hand four hundred than hand forty. A baseline is a living document. The read that was true at nine o'clock is a hypothesis again at midnight.

Reading Online: The Underpriced Half

The standing assumption is that online poker killed the read — no faces, no chips, no shoulders, nothing to see. The assumption is wrong, and its wrongness is an opportunity, because most of your pool believes it and therefore isn't looking.

Online, the baseline method transfers intact; only the inputs change. Instead of posture you have timing — and timing online is cleaner than timing live, because the software strips away everything else. Players fall into tempo grooves: the auto-piloted snap c-bet, the tank that precedes the check-raise, the full-time-bank river shove. When a groove breaks, something happened in that player's head, and you were told about it as plainly as if he'd sighed. Instead of chip handling you have sizing to the pixel — no rounding to green chips, every bet an exact number, and exact numbers form patterns faster. The player who bets 66 percent pot as a default and suddenly produces 83 percent has said something. Instead of table talk you have the chat box and the meta-behavior around it — who's typing angry after a cooler, who instantly reloads to exactly 100 big blinds, who sits out for a hand after losing a big pot, who has three tables going and is therefore not watching you at all. Tilt is louder online than live; it just types instead of talks.

The reason online reads are underpriced is structural. The online generation was raised on databases and study tools, trained to read populations instead of people — this pool folds to three-bets X percent, so I do Y. Population reads are the floor; they're fine; everyone serious has them. But the player two seats over is not a population. He is one person, on one night, in one mood, and the regs mass-tabling around him have rationed their attention too thin to notice. If you play fewer tables and spend the surplus attention on specific humans, you are mining a seam the rest of the room has agreed doesn't exist.

What Ivey Actually Did in Monte Carlo

Every conversation about reading players arrives eventually at the famous hands, so let's take the most famous one seriously instead of reverently.

A televised pot in Monte Carlo, two decades ago now. Phil Ivey, holding nothing, in a raising war with a man who also held nothing — the board had missed them both, and the betting had become the oldest conversation the game knows: who is willing to keep lying. The other man committed a mountain of chips to the lie. And Ivey, with air, moved all-in on top of him and took the pot without a showdown. Barry Greenstein — a mathematician's temperament, nobody's romantic — watched Ivey do this kind of thing for years at the biggest stakes on earth and reached for the only phrase that fit: soul reading.

Here is the part the highlight reel cuts. That moment was not produced in that moment. Ivey had been reading that specific opponent all night — the tempo of his raises, the shape of his aggression, what his bets looked like when they were walls and what they looked like when they were prayers. The all-in wasn't a psychic flash; it was a withdrawal from an account he'd been depositing into for hours. That is what every legendary read turns out to be when you slow the film down: accumulated observation, cashed in one hand. Doyle Brunson, who stayed dangerous into his eighties against opponents who had studied a thousand hours more theory than he ever would, said the quiet part plainly: "I've always said a poker player should be a psychologist. If you can tune into your opponent, a lot of that other stuff goes away."

So is there still magic in this game, or is it all spreadsheets now? Both answers are half right, and the half matters. The feeling of a great read — knowing, in your body, before you could say why, that the man across from you is empty — is real, and it is the best thing this game gives anyone. But the feeling is the output of a trained system, not a gift. It's what expertise feels like from the inside in every domain where expertise is real: the mechanic who hears the engine and knows, the doctor who reads the walk before the chart. It looks like magic from the rail and like seeing from the chair, and it is neither. It is reps. Which means the romance survives rigor — you don't have to choose between loving the mystery and doing the work. The work is where the mystery comes from.

Two guardrails, before the romance gets expensive. First, the read sits on top of fundamentals, never in place of them — Ivey's read is worth something because his baseline game doesn't leak, and a "read" layered over a broken foundation is just a name for the leaks. Second, every deviation you make on a read opens a hole in your own game shaped exactly like that read, so you price reads like risks, not lottery tickets: deviate wide against the oblivious, narrow against anyone reading you back, and return to the solid baseline the moment the information runs thin.

Drills That Build the Muscle

The read is tacit skill — built by reps, not by reading about it, this essay included. So here is practice, concrete enough to start tonight.

One thing per session. Attention is the scarce resource, and unstructured attention doesn't build skill. So constrain it: pick a single channel for the whole session. Tonight, preflop raise sizes only — every player, every raise, is it always the same, when does it vary. Next session, timing only. Then showdown review only: every hand that gets shown, walk the hand backward and annotate what that player's earlier actions actually meant. One channel, fully attended, beats five channels skimmed, and after a month of this the channels start integrating on their own.

Predict before the reveal. At every showdown you're about to see, call the hand — out loud in your head, committed, before the cards flip. Range first, then narrow. You will be wrong constantly at the start, and the wrongness is the mechanism: prediction plus immediate feedback is how tacit pattern recognition gets wired, and poker offers that loop faster than almost any human-reading domain that exists.

Cut tables, keep attention. If you play online, every table you add halves the attention you can spend per human. The mass-tabler is doing kata — executing baseline against nobody in particular. One or two tables, played with full attention on the specific people at them, builds more read per hour than eight tables of autopilot build per week.

Make the floor automatic so attention is spendable. You cannot watch a man's tempo while recomputing your own preflop ranges. The mechanical layer has to run itself — that's what drilling baseline decisions in the trainer is for, and it's the honest division of labor: the machine work in the lab, so the human work can happen at the table. A fuller system for structuring all of this — people-study instead of node-study — is laid out in how to study poker without a solver.

Rep against something that reads back. The muscle grows fastest against an opponent who punishes inattention. The Read exists for exactly this — an opponent that learns your patterns in real time, so the watching runs both directions and you find out, immediately and a little uncomfortably, how much of your own play is legible. Most players are shocked. The shock is the diagnostic.

The uncertainty I'll admit plainly: nobody has run the controlled trial. There is no peer-reviewed study proving that baseline-and-deviation reading adds X big blinds per hundred, and anyone who quotes you a number is inventing it. What exists is the convergent testimony of the best players across every era, a research literature that points at patterns-over-faces, and the structural argument — in a game where the memorizable part is now shared by everyone serious, the only edges left are the ones that can't be shared. Reading a specific human being is the last of those. The evidence says the skill is real. The economics say it's the one still worth building.

The larger argument — what died when the answers arrived, what survived, and why the person across the table was the game the whole time — is a book called The Church of GTO. The whole thing is on the shelf.