Use Quick Play in ICM Trainer when you want the app to choose a short varied practice run. Use the Train tab when you already know the tournament situation you want to work on. Capture a surprising decision while its feedback is visible, then use the result screen to choose a focused follow-up.
The useful distinction is between a hand and a decision. Quick Play ends after ten hands, but a postflop hand can contain several graded decisions. The result screen summarizes both. Its hand list highlights the worst grade within each hand. That summary will not preserve everything you need to reconstruct a decision.
GTO Solutions makes ICM Trainer and publishes this site. This walkthrough was checked on September 30, 2026 against the app’s version 1.2.0 source and local component tests. The US App Store currently lists 1.2.0. We did not test an installed store build. The examples below are authored test data, not customer results or solved poker hands.
Start a run, or choose a situation first
On Home, Start Training launches Quick Play. On an active dashboard the card can instead say CONTINUE; this starts a new Quick Play run, not a saved run at its previous hand. For a selected scenario, open the bottom Train tab rather than relying on that Home card.
Quick Play chooses scenarios for you. When postflop is available, the normal plan contains five preflop and five postflop hands in a shuffled order. Without postflop access, it plans ten preflop hands. If one trainer has no usable hand to deal, the checked implementation can fall back to the other type. The run is not a fixed alternating sequence, and a network error is not a promise of offline fallback. The current Google Play description introduces the ten-hand run and its summary.
For focused practice, the Train tab offers separate PREFLOP and POSTFLOP paths. Preflop selection starts with the tournament stage, then the available player-count and stack-profile choices. Postflop selection includes stage, stack profile and hero/villain seat choices. Choose from what the app offers; these controls do not create an arbitrary new tournament solve.
Read the feedback before you continue
When an action surprises you, pause at its feedback. Record the tournament stage, stack profile, hero/villain seats, hole cards, board and action sequence while they are visible. Add the action you chose and the action the feedback identifies as best. Mark anything you did not capture as unknown.
This matters because the final hand list is shorter than a hand history. It displays a setup label, a hand label, your action and, where relevant, the solver’s best action for the hand’s worst-graded decision. A postflop row does not expose the entire board and betting line. Tapping that row does not reopen a replay in the checked implementation.
Add one question to the note: for example, “Why does the feedback prefer this action in this stage and stack setup?” This keeps the next session tied to an identifiable problem. Our hand-notes guide explains how to keep an incomplete record honest.
What ten hands and twelve decisions look like
We supplied an authored fixture to the app’s results component: five preflop hands with one decision each, four postflop hands with one decision each, and one postflop hand with three decisions. That is ten hands and twelve decisions. The fixture alternates hand types; the table groups them to make the counting visible. Grades and scores are supplied test inputs, not solver recommendations.
On narrow screens, scroll the table sideways. Keyboard: focus the table region, then use the arrow keys.
| Group | Hands | Decisions | Decision grades and scores |
|---|---|---|---|
| Preflop hands | 5 | 5 | 5 Best Move × 100 |
| Four postflop hands | 4 | 4 | 4 Best Move × 100 |
| Last postflop hand | 1 | 3 | Best Move: 100 Correct: 80 Mistake: −20 |
| Total | 10 | 12 | Score sum: 1,060 |
11 of 12 decisions
1,060 ÷ 12, rounded
The ten Best Move decisions and one Correct decision count toward accuracy. That gives 11 ÷ 12 = 91.67%, displayed as 92%. The average score is (10 × 100 + 80 − 20) ÷ 12 = 88.33, displayed as 88. It is not “88% of hands correct.”
The final postflop hand contains Best Move, Correct and Mistake decisions. Its row says Mistake, while the GRADES section counts all three decisions separately. Nine other hand rows are Best Move. Counting the ten hand labels therefore does not reproduce the accuracy tile.
The summary also contains an EQ LOST tile. Keep this as app feedback, not a cash result or a prediction of tournament winnings. We did not supply or validate EV values in this fixture; it tests counting and score display only.
Turn the run into a focused next session
Before selecting PLAY AGAIN, choose one question from the feedback you captured. For example, if you want to revisit a postflop decision, note its stage, stack profile and seats, then use Done → Train → POSTFLOP to choose the available matching context. Check the cards and action path in the next dealt hand; matching a profile does not guarantee the identical hand or decision.
For a preflop question, use Done → Train → PREFLOP, select the relevant stage and available stack setup, then start the ICM session. If the missing piece is the meaning of ICM itself, read the ICM foundations before trying to explain a particular action.
PLAY AGAIN starts a newly selected run. It is a reasonable choice when you want more variety. It is not a button for repeating the exact same ten hands. For deliberate review, keep the recorded question beside you and explain the next relevant decision before reading its feedback.
A new run can contain different scenarios and a different number of graded decisions. Do not treat a higher score on the next run as proof of improvement. The separate trainer-accuracy guide covers comparisons across changing practice sets.
What we checked
We inspected the current Quick Play selection, result aggregation, Home navigation and targeted training controls. We ran the app’s existing Quick Play and targeted-picker tests, then rendered its results component with our own ten-hand fixture. An independent Python calculation checks the displayed counts and arithmetic. Download the fixture, expected results, independent checker and method and manual check steps into one folder. A checksum manifest is also available.
These checks establish the behavior of the inspected source and synthetic result screen. They do not certify a downloaded mobile release, test real purchases or network availability, establish solver accuracy, or measure learning. Interface labels and access can change after updates.
For this study workflow, use ICM Trainer on the App Store or its Google Play listing. The app provides precomputed tournament study and practice. Work away from live play: choose variety when you need a starting point, and a selected training context when you have a specific question.
Product sources accessed September 30, 2026. Header: original AI-generated editorial artwork, not an app screenshot.