Shove·Lab

An MTT poker training tool with Nash equilibrium push/fold ranges, ICM-aware decisions, and built-in coaching that explains the "why" before and after every action.

Inspired by GTO Wizard, but with a focus on teaching the thought process rather than just showing the solver's answer. Every spot can be drilled with or without coaching, and the session review tracks where you're leaking.

Features

• Three training modes

  • Open Shove — folded to you, decide whether to jam (6–15bb)
  • Call — villain shoves, decide whether to call all-in
  • Reshove — villain min-raises, decide whether to jam over (10–25bb)

• Learn — 8 foundation lessons covering position, pot odds, required equity, hand-vs-hand equity, outs counting, fold equity, ICM, and how it all comes together in short-stack strategy. Each lesson links to the relevant drill.

• Drills — six fast-paced math drills with per-drill accuracy and average-time tracking:

  • Pot odds (free input)
  • Required equity (multiple choice)
  • Hand-vs-hand equity (50+ matchups, multiple choice)
  • Outs counting + rule-of-4-and-2 application
  • Fold equity / breakeven calculations
  • Hand-rankings A/B comparison

• Tournament stage selector — Chip EV, Mid (ITM), Bubble, Final Table, or Custom ICM

  • Stage-based heuristic multipliers for quick play
  • Custom ICM uses the Malmuth-Harville algorithm with your stacks + payouts

• Range Viewer — 13×13 hand grid color-coded by Nash threshold

• Session Review — accuracy breakdown by mode, position, stage; identifies your worst leaks

• Coaching toggle — turn explanations on for learning, off for testing

• 4-color GGPoker deck with Triton-broadcast-style card visuals

• Flexible table — Heads-Up, 3-max, 6-max, or 9-max

Tech Stack

• React 18 (functional components + hooks)
• Vite for dev/build
• Pure CSS-in-JS (no styling library — keeps the bundle tiny)
• No backend yet (state lives in-memory; session resets on refresh)

Setup

npm install
npm run dev          # local dev server at http://localhost:5173
npm run build        # production build into ./dist
npm run preview      # preview the production build

Project Structure

src/
├── App.jsx                       # thin wrapper around the trainer
├── main.jsx                      # React entry point
├── index.css                     # global reset
├── components/
│   ├── PushFoldTrainer.jsx       # main orchestrator (tabs + state)
│   ├── TritonCard.jsx            # GGPoker 4-color card visual
│   ├── MiniTable.jsx             # interactive seat-count + position SVG table
│   ├── RangeViewer.jsx           # 13×13 hand grid, color-coded
│   ├── SessionReview.jsx         # trainer-hand stats + leak detection
│   ├── ICMSetup.jsx              # custom stacks + payouts editor
│   ├── LearnView.jsx             # lesson list + reader
│   ├── DrillsView.jsx            # drill picker + per-drill stats
│   ├── ChipRow.jsx               # reusable chip-style picker
│   └── drills/
│       ├── DrillShared.jsx       # frame, choice button, number input, feedback
│       ├── PotOddsDrill.jsx
│       ├── RequiredEquityDrill.jsx
│       ├── HandVsHandDrill.jsx
│       ├── OutsDrill.jsx
│       ├── FoldEquityDrill.jsx
│       └── HandRankingDrill.jsx
├── data/
│   ├── pushRanges.js             # Nash open-shove ranges by position
│   ├── callRanges.js             # call ranges vs different shover positions
│   ├── reshoveRanges.js          # reshove ranges over a min-raise
│   ├── tableConfigs.js           # HU / 3-max / 6-max / 9-max definitions
│   ├── icmStages.js              # stage-based ICM multipliers
│   ├── equityMatchups.js         # precomputed preflop heads-up equities
│   └── lessons.js                # lesson content (data, not JSX)
└── lib/
    ├── handUtils.js              # rank constants, hand-code, random dealing
    ├── decisionLogic.js          # optimal action + threshold lookups
    ├── reasoning.js              # template-based coaching text generation
    ├── oddsCalc.js               # pot odds, outs-to-equity, fold equity math
    └── icm.js                    # Malmuth-Harville ICM equity calculator

The split is deliberate: data/ holds raw lookup tables (no logic), lib/ holds pure functions (no React), components/ holds the UI (no math). Adding a new training mode means a new file in data/, a few functions in lib/, and either extending PushFoldTrainer.jsx or adding a new view component.

Roadmap

• LLM-generated explanations (currently template-based — the hooks in lib/reasoning.js are ready for the swap)
• Persistent storage so session history survives a refresh
• User accounts + cloud sync
• 3-bet defense module (facing a 2.5-3x raise)
• Postflop play (requires a solver dataset)
• Mobile-responsive layout
• Custom range import (paste your own solver ranges)

Data Sources

Push/call/reshove ranges are approximated from publicly published Nash equilibrium charts (HoldemResources, ICMIZER, SnGWiz). They are accurate to the strategic concept and within a couple of bb of true solver outputs for the spots covered. ICM in Custom mode uses the standard Malmuth-Harville recursive algorithm — accurate for short-handed final tables.

For a launched product we'd want to replace the approximated thresholds with a fresh solver run, but the architecture is the same: JSON lookup keyed by (position, hand_code) → threshold_bb.

License

Personal/educational use. Not affiliated with GTO Wizard, GGPoker, Triton Poker, or any commercial poker tool.