PerfectPlay app icon

Stop guessing.
Start playing perfectly.

PerfectPlay is the strategy trainer that turns casino card games from guesswork into solvable puzzles. 14+ games. Real math. Every decision graded against the optimal play.

No real-money gambling No ads No accounts Local-only data
14+
Games covered
10
Video poker variants
µs
Latency for advice
110+
Engine tests
The catalogue

Fourteen games. One engine.

Every game uses the same advisory engine — exhaustive search where it's tractable, Monte Carlo where it isn't. The same answers a professional player would give, returned in microseconds.

BJ
Strategy + Counting

Blackjack

Eight-step basic-strategy lesson, full Train mode, plus a complete Hi-Lo card counting suite covering 17 lessons and five drills.

VP
10 variants

Video Poker

Jacks or Better, Deuces Wild, Bonus, Double Bonus, Double Double Bonus, and more. Optimal hold mask in microseconds.

3CP
Exhaustive

Three Card Showdown

The 18,000-combination evaluator that drops a 3.4% house edge to its theoretical minimum.

BAC
Bet trainer

Baccarat

Player / Banker / Tie + side bets. Last-10 scoreboard helps you spot when "patterns" lie.

PGW
Monte Carlo

Pai Gow Poker

House-way splits, optimal high/low arrangements, side-bet expectations.

UTH
Backward induction

Hold'em Showdown

4×, 2×, 1×, fold — the sequential bet decision solved with subgame-perfect equilibrium.

5SS
Exhaustive

5th Street Stud

Mississippi Stud's three street-by-street decisions, every paytable, every paytable-edge case.

IS
163K combos

Island Stud

Caribbean Stud's call/fold table, derived from full enumeration of every dealer hand.

CDP
Hybrid solver

Crazy Deal Poker

Crazy 4 with the swap-card decision, solved with combined exhaustive + Monte Carlo passes.

FP
~1M combos

Flop Poker

Casino Hold'em's call/fold solved at the river, full 1M-combination evaluation.

LIR
Exhaustive

Ride Poker

Let It Ride's two pull-back decisions, every hand pre-solved.

4CP
MC + lookup

Four Card Showdown

The 270K-entry hand-rank table powering optimal raise/fold play.

HCF
Same-suit run

High Card Flush

Longest same-suit run from seven cards, with the bet-spread decision graded against true probability.

DJW
Wild deuces

DJ Wild Stud

Mississippi Stud variant with deuces wild — every street decision solved against the wild-rank distribution.

Behind the curtain

Real math. Real evaluators. Real speed.

  • Exhaustive search where it's tractable. Three-Card: 18K combos. Mississippi Stud: full enumeration. Microseconds end-to-end.
  • Monte Carlo where it isn't. Pai Gow at 100K samples. UTH with backward induction across the betting tree.
  • Precomputed hand-rank tables. 22K three-card, 270K four-card, 2.6M five-card entries — built once, looked up forever.
  • Validated by RTP. 110+ engine tests, including return-to-player tests that compare simulated long-run results against published theoretical values.

Two modes

Learn — bite-sized illustrated lessons walk through each game's rules, paytable, and decision tree. No prior knowledge required.

Train — random hands, every decision graded against perfect play. Misplays flagged in plain English with the correct action and the math behind it.

Stats persist per game. Accuracy %, decisions played, longest streak — all there next time you open the app.

The deep dive

The Hi-Lo card counting suite.

A full curriculum, not a single drill. Designed to take a curious player from "what's a count?" to live-shoe-ready, with the math motivated at every step.

17-step lesson plan

01

Why counting works

Tens are good for the player. Low cards are good for the dealer. The shoe drifts.

02

Hi-Lo values

2–6 = +1. 7–9 = 0. T–A = −1. The simplest balanced count there is.

03

Low cards

Drill the +1 group on its own.

04

Neutral cards

7, 8, 9 contribute nothing — recognise instantly.

05

High cards

T, J, Q, K, A all count −1.

06

Short sequences

Walk through the running count on a 5-card stream.

07

Cancel pairs

+1 / −1 cancel. Ignore the noise; count the net.

08

Full deck = 0

The balanced-count sanity check. End-of-shoe must be zero.

09

True count

RC ÷ decks remaining. Why the divisor matters.

10

Estimating decks

Eyeballing the discard tray to within a half-deck.

11

Bet spread

The whole point of counting is to bet bigger when you're favoured.

12

Kelly sizing

Full / half / quarter Kelly explained with a $10K bankroll example.

13

Illustrious 18

The 18 deviation cells worth the most EV.

14

Insurance @ TC +3

The single biggest deviation in basic counting.

15

Fab 4 surrenders

Late-surrender cells that flip with the count.

16

Penetration

Why deep deals matter, and why 0.75 is the practical floor.

17

Ready to drill

The handoff to muscle memory.

Five drills

+1 / 0 / −1

Card Drill

Flash one card; tap its Hi-Lo value as fast as you can. Configurable speeds from 3.0 s down to 0.18 s. Auto-advance or tap-only modes. Tracks reaction time + streak.

Full shoe

Running Count

Burn through a full 52-card deck. Configurable checkpoints — every card, half-deck, or end-only. The balanced count must land on zero.

RC → TC

True Count

Given a running count and a shoe-depth bar, dial in the true count. Shows the precise division in the review.

Half-Kelly ramp

Bet Sizing

Pick the correct unit bet from the canonical 1u → 12u ramp. Per-tier explanation in the review so you know why.

I-18 / Fab 4

Deviation Trainer

Hand + dealer upcard + true count — pick basic strategy or the deviation play. The TC is sampled with bias around each rule's threshold so the answer flips often.

What's in the app

Built for serious practice.

Persistent stats

Per game, per drill. Accuracy, decisions played, best streak, time-on-task. SQLite-backed — survives reboots; never leaves your device.

Surrender, splits, doubles

Multi-row action buttons handle hands with 4+ choices. Splits draw to each hand independently. Cards fan to fit the table.

Plain-English feedback

"Hit. 16 vs 9 is a hit because the dealer's 9 outperforms your stand-on-16 EV by 2.1%." No magic numbers; the reasoning is shown.

Cross-platform

iOS, iPadOS, Android, Mac, Windows. Same engine, same answers, native UI on each.

Offline

Everything runs on-device. No login, no sync, no network calls. Works on a plane.

Common questions

FAQ

Is this real-money gambling?

No. PerfectPlay is a strategy trainer. There is no wagering of real or virtual currency, no purchasing of chips, and no jackpots. You play simulated hands; the engine grades each decision; you build accuracy.

Will using PerfectPlay get me banned from a casino?

Using a strategy trainer at home is not gambling and has no relationship with any casino. Card counting is not illegal in most jurisdictions, but casinos are private property and may ask players to leave for any reason. We don't endorse, recommend, or facilitate live counting — PerfectPlay is for the math.

How accurate are the recommendations?

Where the game's state space is small enough — Three Card, Blackjack, Mississippi Stud, Video Poker — we exhaustively enumerate every possible deal and choose the EV-maximizing action. There is no closer-to-perfect than that. For larger games (Pai Gow, Hold'em) we use Monte Carlo with enough samples to put confidence intervals well below the granularity of any meaningful decision.

Does PerfectPlay collect data?

No. All stats are stored in a local SQLite database on your device. There is no analytics, no telemetry, no advertising SDK, no account, and no network connection. Uninstalling the app removes the data. See our Privacy Policy.

Why are the in-app game names different from the casino?

Several casino games have trademarked names. PerfectPlay uses descriptive non-trademarked names — "Three Card Showdown" rather than the licensed product name, for example. The rules, paytables, and strategy are mathematically equivalent.

What platforms are supported?

iOS and iPadOS via the App Store, Android via Google Play, plus desktop builds for macOS and Windows. Same engine on all platforms — the answers don't change.

How do I report a bug or request a feature?

Email support@perfect-play.app or use the support page. Include the game, the hand, and what you expected vs what you saw — we'll usually have a fix in the next release.