From a phone photo to a GTO answer: how the analyzer works

The fastest way to review a hand is to photograph it and let the tool do the reading. This article walks through what happens between the photo and the answer, and what you need to add so the result is trustworthy.

What the photo can show

Point your camera at the table or your screen and the analyzer reads the cards: your two hole cards and the community board. Computer vision is good at this, but it is not magic, so the tool always shows you what it read and asks you to confirm it.

The confirm step is not busywork. A misread card changes the whole answer, so a quick check that the tool saw your hand correctly is the single most valuable thing you can do. Cards flagged with a low confidence dot are the ones worth a second look.

What the photo cannot show

A photo captures cards, not the story of the hand. Several things matter just as much and have to be filled in:

1. Stack sizes. How deep you and your opponent are decides which lines are even available. A pot sized bet means something different at 20 big blinds than at 200.
2. The pot. The size of the pot sets the price of every call and the value of every bet.
3. Positions. Who acts last after the flop shapes the entire strategy. Position is one of the strongest forces in poker.
4. The action so far. A check is not the same spot as facing a raise. The line that led to this moment determines which decision you are actually solving.

The confirm screen asks for each of these. The action builder in particular lets you record the betting step by step, so the tool can pinpoint the exact spot in the solution rather than guessing.

What the answer means

Once the spot is complete, the tool returns a recommendation: the action, your equity (your share of the pot if the hand went to showdown right now), and how the decision compares to the price you are being offered. The recommendation itself comes from an open-source DCFR solver, run offline on precomputed spot keys and served from the database rather than recomputed for every request.

When a hand sits in a genuine mix, you will see frequencies rather than a single command. A result of "bet 70 percent, check 30 percent" is not the tool hedging. It means a balanced player splits that hand, and either action is defensible. Over many hands, following the mix is what keeps you unreadable. If you want the theory behind that, read what GTO poker actually means.

Getting an accurate answer

Garbage in, garbage out applies here. A few habits make the output reliable:

• Confirm the cards every time, especially any flagged with low confidence.
• Enter the pot and stacks in big blinds, not chips, so the analysis is stack depth aware.
• Record the real betting line in the action builder rather than leaving it blank, or the tool defaults to the opening decision.
• Set both positions. Hero and villain seats tell the tool who is in position, which flips large parts of the strategy.

Spend ten extra seconds on context and the answer goes from a rough guess to a genuine read on the spot.

This is a study tool

The analyzer is built for reviewing hands after the fact and for training, not for live assistance during real-money online play. Using any tool to make decisions mid-hand online breaks site rules and gets accounts banned. Photograph your interesting spots, review them later, and let the patterns sharpen your instincts for next time.

Bottom line

The analyzer reads your cards from a photo, you confirm them and add the context a photo cannot show, and it returns the action plus your equity and the relevant math. The accuracy is entirely in the context you provide, so confirm the cards, use big blinds, record the line, and set the positions. Then treat the whole thing as review, never as live help.