How XPloit's AI Coaching Works

XPloit is a place to practice poker decisions and understand why a strong line is strong. You play a spot, or you bring a hand you already played, and you say what you were thinking. Then XPloit works through the decision with you: what your opponent's range looks like, what your hand wants to do against it, which sizes pressure the right hands, and what changes on later streets.

The goal is not to hand you an answer and move on. It is to sharpen your own reasoning, so the next time you are in the spot for real, you already know which questions to ask. If you want the thinking it is trying to build, the strategy guide and the manifesto lay out the approach.

XPloit reasons from the same information a thoughtful player at the table would have. The structured facts of the hand:

• •positions, stack depths, and the stack-to-pot ratio,
• •the board, the full betting sequence, and the bet sizes,
• •anything you tell it about the players: reads, archetypes, notes, history,
• •and your own stated reasoning, when you give it.

That is the whole input. It does not know the future, it does not see your hand through a villain's eyes, and it has no secret file on your opponents. Give it more context and it reasons better. Give it less and it tells you what it would assume, and why.

We will say plainly what it is not: XPloit is not a solver running in the background, and we do not present it as trained on a private trove of winning hands. It is poker-specific reasoning applied to the spot in front of you.

XPloit is not a solver. It does not compute an equilibrium, and it does not promise that any single recommendation is the mathematically perfect play or that following it will be profitable.

A solver answers a different question: how to play unexploitably against an opponent who never makes a mistake. Most of the games you actually sit in are full of mistakes, and that is exactly where the money is. XPloit is built for the human game, not the perfect one.

So treat a recommendation as a strong argument you are free to push back on, not a verdict. It is a sparring partner, not an oracle, and it will sometimes be wrong.

Feedback is coaching, not a scoreboard. Read it for the reasoning: the range interaction, the sizing logic, the exploit it is targeting, and the plan for the next street. Not just the letter grade or the action it landed on.

In close spots, more than one line can be defensible. When that is true, XPloit should show you the tradeoff rather than pretend the margin does not exist. If you disagree, say so. The most useful sessions are the ones where you argue back and one of you learns something.

An AI that reasons in words can still say something that sounds right and is not. So XPloit does not just generate feedback and ship it.

• •Decisions run through a structured grading process with a consistent rubric, so a spot is judged the same way across the app instead of graded on feel.
• •A separate checking pass looks for the kinds of mistakes language models make, claims that do not match the actual cards, math, or action, and flags or corrects them before you see them.
• •A library of regression tests pins down behavior we care about, so spots we have already gotten right stay right as we keep building.

None of this makes the AI infallible. It makes it accountable, and it lets us catch and fix patterns instead of guessing.

Worth being blunt about. XPloit:

• •only knows what it is given. A thin or wrong read on a player poisons the reasoning built on it.
• •does not feel the table. Tells, timing, and the way someone has been tilting for an hour are invisible unless you describe them.
• •works from simplified assumptions about ranges that real opponents will happily violate.
• •can be confidently wrong, like any model, and can land on a different reasonable line than another strong player would in a close spot.

Use it the way you would use a sharp friend who studies a lot. Worth listening to, worth arguing with, never the final word.