A Deviation Framework: Evidence Before Action
Builds a formal pre-deviation checklist: a minimum evidence threshold scaled to how extreme the deviation is, the smallest-sufficient-deviation principle of moving only as far from baseline as the read justifies, and a reversibility check on how chea
Assumptions: 100bb 6-max online cash at low/mid stakes, no rake unless stated. Quantitative claims use the repo tools logged in content/mathlogs/exploitative-play for this lesson. Population and solver-facing statements are framed as practical low- and mid-stakes heuristics unless an exact calculation is explicitly logged.
A Deviation Framework: Evidence Before Action
Builds a formal pre-deviation checklist: a minimum evidence threshold scaled to how extreme the deviation is, the smallest-sufficient-deviation principle of moving only as far from baseline as the read justifies, and a reversibility check on how cheaply you can undo the adjustment if wrong. Quantifies evidence tiers from a single showdown through 200-hand stat samples and maps each tier to a permitted deviation size. Works through declining to fold-to-river-raise exploitatively after one ambiguous hand versus an unknown at $0.50/$1...
The job of this lesson is to make deviation framework executable in a real $0.25/$0.50 or $0.50/$1 online pool. Exploitative poker is not a license to make random heroic reads. It is a disciplined process: identify a repeated opponent error, choose the part of your strategy that profits from that error, size the deviation to the evidence, and keep a reversion point ready. That process matters more than the particular hand used in the example, because the same habit transfers across sites, stakes, and table lineups.
The baseline remains the anchor. A competent default protects you from being run over when the read is weak, while the exploit earns extra EV when the read is strong. In this module the frame is dynamic adjustment. That means your first question is not, 'what do I feel he has?' The first question is, 'what behavior has this player or pool already shown, and which frequency should I change because of it?' A range can be loose, tight, passive, aggressive, capped, polar, overfolding, or under-bluffing. Once you name that property, the hand stops being a guessing game and becomes a decision about value, bluffing, defense, or note quality.
Use numbers as guardrails, not decoration. The local odds tool says a half-pot pure bluff needs 33.3% folds, a 75% pot bluff needs 42.9% folds, and a pot-sized bluff needs 50.0% folds. Those are not poker vibes; they are break-even thresholds. If a calling station is folding only one time in four, a pot-sized bluff is torching money before card removal or blockers even enter the discussion. If a regular is folding nearly half their range in a node where your story is credible, the same bet can become a clean exploit. The formula does not pick the opponent. Observation does.
Preflop ranges create the same discipline. The checked range file gives a 17% UTG opening range, 26.1% cutoff range, and 42.1% button range for the local TAG baseline. A player who voluntarily enters 45% of hands from all seats is not merely 'a little loose'; they are playing a different game tree from the baseline. A player whose 4-bet range is only QQ plus and AK has 34 live combos before blockers, only 2.6% of all starting combinations. Those counts explain why folding attractive but dominated hands can be mandatory against honest aggression, while widening isolation raises against limpers can be mandatory against people who call too much.
The practical exploit in this lesson can be summarized in three actions: fold strong bluff-catchers to under-bluffed raises, choose blocker-heavy river bluffs, and record the line after showdown. Do not treat these as slogans. Each action needs a trigger. The trigger might be a HUD sample, a showdown, a repeated sizing, a first-orbit limp, a pool default, or a session note. The trigger also has a confidence level. One odd min-raise is a clue; three min-raises followed by a showdown is a read. Thirty hands can tell you VPIP direction, but they cannot prove a rare 3-bet tendency. Strong exploiters scale the adjustment to the reliability of the evidence.
A good in-game script is short. First, name the node: unopened pot, limp isolation, 3-bet pot, c-bet response, turn barrel, river raise, blind defense, or showdown note. Second, name the villain feature: station, nit, maniac, TAG, solver-trained regular, anonymous pool default, or changing table state. Third, choose the exploit direction: more value, less bluffing, more bluffing, wider defense, tighter defense, larger sizing, smaller sizing, or rebase to default. Fourth, ask what evidence would prove you wrong. If you cannot answer the fourth question, your read is probably too emotional.
Worked hand 1
In the first hand, Hero uses the lesson read before thinking about the exact combo. The range question is whether Villain arrives with too many weak hands, too many traps, too many folds, or too many bluff attempts. If Villain is loose-passive, thin value rises and pure bluffing falls. If Villain is a tight regular with a capped check-back range, delayed aggression becomes attractive. If Villain is a maniac, medium-strength hands gain bluff-catching value and strong hands can call rather than raise. The cards matter, but the opponent's error decides which card classes make money.
The pot geometry matters too. Suppose Hero is considering a 75% pot bet. The logged odds command says that bet needs 42.9% folds as a pure bluff. If this line is for value, the same size asks a different question: will worse hands call often enough? Against a station, sizing up with top pair or second pair can be correct because their calling range is too wide and too inelastic. Against a nit, the value threshold moves upward because their continues are stronger. Against a reg who overfolds, the bluff portion expands first through hands with blockers, backdoor equity, or poor showdown value.
Worked hand 2
The second hand is the evidence check. Many losing exploit attempts fail here. Hero sees one behavior, gives it too much weight, and makes a massive deviation. That is backward. A single showdown can be high quality, but only if the line was unmistakable and relevant to the current node. A single fold to a c-bet is low quality because everyone folds sometimes. A recurring note such as 'called three streets with third pair in single-raised pot' is high value because it immediately changes your river value threshold. A recurring note such as 'river check-raise showed nuts twice' changes your bluff-catching threshold.
This is why the smallest-sufficient-deviation rule is central. If the read says a player overfolds blinds, start by stealing wider in the profitable seats, not by opening any two cards from everywhere. If the read says a station calls too much, value bet thinner and cut the low-equity bluffs, not every semi-bluff with strong equity. If the read says a reg attacks your button steals, tighten the steal range temporarily and add value 4-bets, not a permanent retreat from every late-position pot. Exploitation is dynamic portfolio management, not a personality change.
Applying the read
For A Deviation Framework: Evidence Before Action, the most common profitable mistake is usually not subtle. Low- and mid-stakes players reveal themselves with limps, sizing shortcuts, check-backs, timing, showdown quality, and river honesty. Your job is to convert that evidence into a future decision. The note should read like an instruction: 'limp-calls any suited hand, iso larger and value bet Kx,' or 'turn raise from passive player is two pair plus, fold one pair,' or 'folds big blind to steals repeatedly, open button aggressively until resistance appears.' A note that only says 'bad' or 'aggro' is not enough, because it does not tell you what to do with A9s, second pair, or a river bluff-catcher.
There is also a visibility cost. Quiet exploits, such as overfolding to a passive player's river raise, are hard for the table to punish because they rarely see the folded hand. Loud exploits, such as 3-betting every button open, are easy for attentive regs to notice and counter. Before leaving baseline, ask who can see the deviation and whether they are capable of fighting back. Against an isolated recreational player, the exploit can be larger. At a table with two sharp regs watching, the same exploit may need to be narrower, timed, or abandoned after one orbit.
Common failure modes
The first failure mode is stereotyping too early. Buying in short, limping once, or using an odd size gives a prior, not a verdict. The second failure mode is confusing aggression with strength. A maniac's barrel is not the same object as a nit's barrel, even if the bet size is identical. The third failure mode is ignoring stake context. A line that is wildly under-bluffed at 10NL may contain real bluffs at 200NL. The fourth failure mode is keeping stale notes after the player changes. Tilt, stack depth, table composition, and history can all move a player temporarily away from their usual type.
The antidote is a review loop. Mark hands where the exploit was decisive, then write the claim before checking the result: 'I bet thin because Villain calls too wide,' 'I folded because this pool under-bluffs river raises,' 'I 3-bet bluff because Villain overfolds to 3-bets and the blinds were not fighting back.' After the session, audit whether the evidence actually supported the claim. If the hand won but the evidence was weak, do not reward the process. If the hand lost but the read and price were sound, keep the play in the toolkit.
Drill
For the next session, pick one target read from this lesson and track it deliberately for two orbits. Do not try to profile everyone at once. Record the first relevant action, the confidence level, and the exploit you would use if the spot repeats. When the spot repeats, execute the prewritten action unless new evidence has appeared. This removes ego from the decision. You are not proving that you are clever; you are testing whether a specific opponent error exists and whether a specific counter earns money.
The final standard is simple: every exploit should be explainable in one sentence and reversible in one sentence. Explanation: 'Villain folds too much to turn barrels, so I double barrel hands with equity and blockers more often.' Reversal: 'If Villain check-calls two marginal hands or starts check-raising, I return to baseline.' That pair of sentences keeps exploitative play profitable without letting it become reckless.
Quantitative anchors used here
The mathlog for this lesson records the exact local tool outputs behind the numeric anchors. The important values are: half-pot bluff alpha 33.3%, 75% pot bluff alpha 42.9%, pot-sized bluff alpha 50.0%, a $10 call into a $30 pot requiring 25.0% equity, QQ plus and AK containing 34 starting-hand combos, and the local baseline RFI sizes of 17% UTG, 26.1% CO, and 42.1% BTN. These are teaching anchors, not claims that every pool or solver tree uses the same strategy.
If you can connect those numbers to the opponent type, you have the lesson. Bluff when folds exceed the threshold. Value bet when worse calls exceed the threshold. Fold when a passive range is too value dense. Widen preflop when the caller enters too many dominated hands. Rebase when the evidence disappears. That is exploitative play without guesswork.