In chapter 11 of How To Read Hands At No-Limit Hold’em, Ed Miller explains that accurate hand reading depends on profiling opponents, but the usual “wait for showdowns” approach fails because most hands never reach showdown—and the ones that do often reveal little. So you must learn from frequency and pattern data: how often someone enters pots, raises, barrels, gives up, etc.
The Core Problem: Humans Infer Badly
Miller argues people mis-profile opponents because they:
-
Overweight recent events (a few dramatic hands feel like “proof”).
-
Ignore how common or rare a trait is in the player pool (the “population”).
A key idea: when you observe something unusual, it’s often more likely your opponent simply ran hot than that they possess a rare, elite tendency. In small-stakes games, “world-class hand reader” is treated as an extremely rare event—so you should default to more ordinary explanations unless evidence becomes overwhelming.
Hypotheses, Then Updates: The Bayesian Mindset
He frames profiling as a two-step loop:
-
Form a hypothesis (e.g., “calling station,” “nit,” “overaggressive,” “trappy”).
-
Update the probability of that hypothesis with every new observation.
Observations can support or contradict your hypothesis, and you should adjust gradually, with bigger adjustments for more informative evidence (like revealed cards) and smaller ones for weak evidence (like a single fold).
Uncertainty Attached to Most Poker “Data”
Miller emphasizes that poker observations are noisy:
-
A turn shove with no showdown could be value, a draw, or a bluff.
-
Many “signals” have high false-positive rates.
So you should treat many actions as ambiguous evidence, not certainty—similar to how medical tests can be “pretty accurate” yet still misleading when the condition is rare.
Rarity and Base Rates: Why “Rare Reads” Need More Proof
A central takeaway is the importance of base rates (how often traits appear in the population you’re playing in). Some traits are common at small stakes (calling too much, bluffing too little). Others are rare (frequent high-quality multi-street bluffs, sophisticated 3-bet bluffing, turn bluff-raises).
Because rare traits are rare, you need many more confirming observations before you should believe them. Early on, assume you’re seeing a common player type experiencing uncommon cards, not the other way around.
A Practical Table Application: Profiling Preflop Reraisers
Miller gives a concrete Bayesian-style example using preflop reraises and proposes three broad reraiser profiles:
-
Ultra-nit reraiser: only the very top premiums.
-
Tight reraiser: premiums plus a few near-premiums.
-
Loose reraiser: wider value range plus situational bluffs.
He then shows the correct approach isn’t “who reraises most?” but:
How common is each type in this stake/venue, and how much does one observed reraise move the needle?
The punchline: even if a loose reraiser reraises more frequently, if loose reraisers are rare in your $1–$2 pool, a single reraise is still more likely to come from the common tight profile than from the rare loose one.
How to Use This Without Doing Math at the Table
Miller’s point isn’t that you should calculate exact probabilities mid-hand. It’s that you should adopt the process:
-
Start with a reasonable baseline for the game you’re in.
-
Treat early reads as probabilities, not labels.
-
Update slowly unless you get high-quality evidence (especially showdown info).
-
Prefer common explanations over exotic ones until the evidence truly stacks up.
Training Method: One-Player Deep Tracking
He recommends an exercise: pick one opponent for a session, write down a few initial hypotheses with an estimated confidence level, then track every action as confirming/contradicting—and note whether each data point is weak or strong. This combats the common bias of noticing only evidence that supports your first impression.
Key Takeaways
-
Profiling is essential because showdowns are scarce and often uninformative.
-
Most poker observations are noisy; treat them as uncertain evidence.
-
Always incorporate base rates: rare traits require lots of proof.
-
Use a Bayesian mindset: hypothesize, observe, update—incrementally and objectively.
-
Structured note-taking and deliberate tracking can dramatically improve your accuracy.