--- title: "How Poker Sense Grades Your Decisions" description: "GREAT through BLUNDER: how Poker Sense grades poker decisions by solver frequency instead of EV loss, and what a genuinely close spot really means." date: "2026-08-10" author: "The Poker Sense Team" category: "The app and your training" tags: ["gto training", "poker verdicts", "mixed strategy", "expected value", "poker feedback", "home game"] canonical: "https://pokersense.app/blog/how-poker-sense-grades-your-decisions" --- # How Poker Sense Grades Your Decisions GREAT through BLUNDER: how Poker Sense grades poker decisions by solver frequency instead of EV loss, and what a genuinely close spot really means. - The Poker Sense Team - Aug 10, 2026 - Reading time: 9 min You just finished a training hand in [Poker Sense](https://app.pokersense.app). You were in the big blind -- the seat that posts the larger of the two forced bets before any cards are dealt -- and you called a preflop raise holding the Jack and Ten of spades. The flop, the first three shared cards, came Queen of spades, Nine of spades, Four of hearts. You checked, your opponent bet about half the pot, and you took stock: an open-ended straight draw (any King or Eight completes a straight) plus a flush draw (one more spade makes a flush). Calling felt fine. You called. The grade came back: GOOD. Under it, a single line: "Raise is the main play (68%)." Good? Not great? You had a monster draw, you took the careful line, and apparently the computer wanted you to raise -- but only 68 percent of the time, whatever that means. Which raises the real question. What does a grade on one poker decision even mean? Poker is a game of luck, and the math itself often says to play the same hand two different ways. Can a single decision be graded at all? It can, but not the way most people assume. Walking through how Poker Sense actually does it turns out to be a pretty good tour of how to think about poker decisions in general -- starting with why the obvious way to grade them fails. ## The Obvious Way to Grade, and Why It Fails The obvious way to grade a poker decision is to measure what it cost. Modern poker strategy comes from solvers: programs that play a situation against themselves millions of times until they converge on a Game Theory Optimal strategy -- GTO for short -- meaning a strategy that cannot be exploited no matter what the opponent tries. For any spot, a solver can tell you the [expected value](/blog/what-is-expected-value) of every action: the average amount that action wins or loses over the long run, with the luck of any single hand averaged out. So just compare the play you chose against the best available play. The gap in average profit is your grade. Simple. Here is the problem, and it is the mathematical heart of this whole post. On most flops, the solver does not pick one action. It plays a [mixed strategy](/blog/understanding-mixed-strategies): with the same hand in the same spot, it raises sometimes and calls sometimes, at specific frequencies. And at equilibrium -- the state where both players are playing perfectly and neither can gain by changing anything -- every action the solver actually uses in a mixed spot has exactly the same expected value. That is not a coincidence. It is the reason mixing exists at all. If raising made even slightly more money than calling, the solver would not raise 68 percent of the time; it would raise 100 percent of the time, because solvers are ruthless about profit. A mix can only survive when the actions in it are precisely tied. Game theorists call this the indifference principle: at equilibrium you are indifferent between the actions in the mix, because they earn identically. Now watch what that does to the obvious grading scheme. Your call in the Jack-Ten hand? Zero expected value lost. The raise the app called the main play? Also zero. Folding -- the one action the solver never takes with that hand -- is the only choice that costs anything. Grade by EV loss alone and nearly every postflop decision that is even roughly reasonable comes back "perfect." A report card where almost everything is an A is not feedback. It is wallpaper. ## The Signal EV Loss Can't See So Poker Sense grades against something EV loss is blind to: the solver's frequencies. When the solver raises 68 percent and calls 32 percent, both actions are fine -- but they are not identical as habits. Your opponents never play against one decision; they play against your whole [range](/blog/introduction-to-ranges), meaning every hand you could be holding when you take a given action. A player who calls with every big draw in spots like this is, one free decision at a time, building a different range than the solver built. Their raises contain only strong made hands, which makes those raises easy to play against. Their calls are stuffed with draws, which makes their later bets easy to read the moment a straight card or flush card lands. Each individual call cost nothing at equilibrium. The habit quietly moved them onto a different strategy -- one the "costs nothing" guarantee was never issued for. That is the insight the whole verdict system is built on. Frequency alignment is the signal EV loss cannot see. When your action is inside the solver's strategy, the useful question is not "what did this cost" (nothing) but "how central is this action to the strategy." When your action is outside the solver's strategy, cost becomes exactly the right question, because now there genuinely is one. ## The Six Verdicts Poker Sense sorts every decision into one of six verdicts. The first three mean you were inside the solver's strategy, and they differ only in how often the solver does what you did. The last three mean you were outside it, and they differ only in how expensive the detour was. Cost is measured in big blinds -- the size of that forced bet you posted, poker's standard measuring unit, usually shortened to bb. **GREAT** means you chose the primary play, the action the solver takes most often -- or an action the solver plays within 10 percentage points as often as the primary one. That second clause matters. When the solver raises 52 percent and calls 48, calling is not a lesser choice; it is half of a genuine coin flip, and it grades like one. **GOOD** means you chose an action the solver takes a meaningful share of the time: 15 percent or more. Your Jack-Ten call lands here. The solver raises that hand most of the time and calls most of the rest, so calling is a real part of the strategy. Just not the headline. **OKAY** means your action is in the solver's strategy but rare: under 15 percent. Nothing lost, nothing to apologize for. But if you keep landing here in the same kind of spot, the strategy's center of gravity is somewhere you are not. **IFFY** means you left the strategy, but cheaply: less than 1.5 big blinds of expected value lost. Real money over time, not an emergency today. **MISTAKE** means the detour cost 1.5 to 3 big blinds. These are the decisions where focused study pays off fastest. **BLUNDER** means 3 or more big blinds lost. In the Jack-Ten hand, folding that fifteen-out monster draw to a half-pot bet would burn several big blinds and land here. Everyone has blunders in them; the point of finding yours in training is that training chips are free. One more number ties this together: your session accuracy is simply the percentage of your decisions graded GREAT or GOOD. It is a quick read on how often you stayed near the center of the strategy. ## What "Close Spot" Really Means Some spots are not close at all -- the solver takes one action nearly always, and anything else is a leak. But plenty of spots are genuinely mixed, with two or more actions that are near-equals. If the solver bets 45 percent and checks 55, those plays are practically twins. The indifference principle says they earn the same, and the frequencies say the solver barely leans either way. Imagine being told you "failed" for betting in that spot. That would not be teaching; it would be noise. Worse, it would train you to hunt for a single right answer in situations where the entire mathematical point is that there is not one. So when a spot is close, Poker Sense says so. The 10-point window inside GREAT exists precisely for this, and the insight line comes right out with it: "Close spot -- both plays are solid." That line is the app telling you where not to spend your worry. The hands worth replaying on the drive home are the ones where you and the solver genuinely diverged -- not the coin flips. ## Why the Grade Is a Word, Not a Number Some training tools lead with the raw output: bet is worth 2.31 big blinds, check is worth 2.28. Poker Sense computes all of that too -- it is exactly how the IFFY, MISTAKE, and BLUNDER boundaries get drawn, big deviations graded by precisely how many big blinds they burn -- but the grade itself never arrives as a decimal. Two reasons. First, false precision. Solver outputs are converged approximations: the solver runs until its strategies stop moving by more than a tolerance, then stops. The difference between reasonable lines is often thousandths of a big blind, down in the noise of the approximation itself. A third decimal place looks like information. It is not. Second, words teach better than numbers here. "You lost 0.03bb" makes you squint and wonder whether that is a lot. A sentence tells you what to do with what just happened. So every decision gets a one-line, plain-language insight instead: "Raise is the main play (68%)" when there is a clear primary. "Close spot -- both plays are solid" when there is not. "Big miss -- Check was the play here" when you wandered somewhere expensive. One line, and you know whether to move on or lean in. None of the detail is hidden, to be clear. After a decision you can open the range view and see the full solver breakdown -- every hand you could have held, each action's frequency, and yes, the EVs. The difference is what leads: the grade hands you a sentence first and keeps the decimals where the curious can find them, instead of making you do the interpreting yourself. ## What a Grade Cannot Tell You Here is the honest limit of any grading system: a grade can locate a divergence, but it cannot explain one. GOOD with "Raise is the main play (68%)" tells you where your instincts and the strategy split. It does not tell you why the solver loves raising that draw. The why is where the learning lives, and it is what the Ask Why coaching is for. Tap it after the Jack-Ten hand and the AI coach explains the reasoning behind the recommended play: your draw is strong enough that you are close to a coin flip even against hands that continue, so raising gives you two ways to win -- your opponent folds now, or one of your fifteen cards arrives -- and building the pot while your hand has this much potential means the times you get there, you win more. That reasoning transfers to the next big draw you hold, on a board you have never seen. The grade found the gap; the conversation fills it. On the free tier you get 20 practice hands and 3 coaching conversations a day, which is enough to make asking why a habit on the hands that surprise you. ## The Bottom Line A grade is a mirror, not a verdict on you as a player. It compares one decision against a strategy: inside the strategy, it tells you how central your action was; outside, it tells you what the detour cost in big blinds. That is all it claims to know -- and because of the indifference principle, that is all any honest grade can claim. Your accuracy number deserves the same honesty. The percentage of decisions graded GREAT or GOOD is a practice metric, not your win rate. Variance will still decide your Friday night; you can play at 85 percent accuracy and lose three buy-ins anyway, while the guy who calls everything drags the biggest pot of the year. So yes -- one decision can be graded, even in a game of luck and mixed strategies. Not by whether it won, and not by a decimal pretending to more precision than the math holds, but by where it sits relative to a strategy that cannot be pushed around. The grade shows you where you and that strategy diverge, hand after hand. What you do with that is the actual training. Tags - gto training - poker verdicts - mixed strategy - expected value - poker feedback - home game ## Twenty hands a day, free Start training with proven strategies and AI coaching today. Your first 20 hands are free, every day. [Start training free](https://app.pokersense.app/signup) No credit card required. Free forever.