{"kind":"Skill","metadata":{"namespace":"community","name":"thinking-probabilistic","version":"0.1.0"},"spec":{"description":"Express confidence in ranges, update predictions with new information, and track calibration over time. Use for project estimation, risk assessment, and decision making under uncertainty.","files":{"SKILL.md":"---\nname: thinking-probabilistic\ndescription: Express confidence in ranges, update predictions with new information, and track calibration over time. Use for project estimation, risk assessment, and decision making under uncertainty.\n---\n\n# Probabilistic Thinking\n\n## Overview\n\nProbabilistic thinking, informed by the research of Philip Tetlock's \"Superforecasting,\" treats beliefs as probabilities rather than certainties. Good probabilistic thinkers express confidence in ranges, update beliefs when evidence changes, and track their accuracy to improve calibration over time.\n\n**Core Principle:** Express beliefs as probabilities. Track predictions. Update when wrong. Calibrate over time.\n\n## When to Use\n\n- Project timeline estimation\n- Risk assessment\n- Predicting outcomes (launches, decisions, events)\n- Evaluating uncertain technical choices\n- Making decisions without complete information\n- Any forecast or prediction\n\nDecision flow:\n\n```\nMaking a prediction?\n  → Is outcome uncertain? → yes → EXPRESS AS PROBABILITY\n  → Can you track the outcome? → yes → RECORD AND CALIBRATE\n  → New information available? → yes → UPDATE PROBABILITY\n```\n\n## Core Concepts\n\n### Probability as Confidence\n\nConvert vague language to numbers:\n\n| Vague Statement | Probability Range |\n|-----------------|-------------------|\n| \"Certain\" | 99%+ |\n| \"Almost certain\" | 90-99% |\n| \"Very likely\" | 80-90% |\n| \"Likely\" / \"Probable\" | 65-80% |\n| \"Better than even\" | 55-65% |\n| \"Toss-up\" | 45-55% |\n| \"Unlikely\" | 20-35% |\n| \"Very unlikely\" | 10-20% |\n| \"Almost impossible\" | 1-10% |\n| \"Impossible\" | \u003c1% |\n\n### Confidence Intervals\n\nExpress estimates as ranges, not points:\n\n```\nBAD: \"The project will take 6 weeks\"\nGOOD: \"I'm 80% confident the project will take 4-8 weeks\"\nBETTER: \"50% confidence: 5-7 weeks; 90% confidence: 3-10 weeks\"\n```\n\n### Base Rates\n\nStart with how often similar things happen:\n\n```\nQuestion: Will this feature launch on time?\nBase rate: What % of similar features launched on time? ~40%\nAdjustment: This team is experienced (+10%), scope is clear (+10%)\nEstimate: ~60% probability of on-time launch\n```\n\n## The Probabilistic Process\n\n### Step 1: Express Initial Probability\n\nState your belief as a number:\n\n```markdown\n## Prediction: Will we hit Q2 revenue target?\n\nInitial estimate: 65%\nReasoning:\n- Last 4 quarters: Hit 3/4 targets (75% base rate)\n- Current pipeline: Slightly below historical (-10%)\n- New product launching: Uncertain impact\n```\n\n### Step 2: Identify Key Uncertainties\n\nWhat could change the probability?\n\n```markdown\nKey uncertainties:\n1. Will Enterprise deal close? (+15% if yes)\n2. Will new product cannibalize existing? (-10% if significant)\n3. Will competitor launch disrupt? (-20% if aggressive)\n```\n\n### Step 3: Create Probability Tree\n\nFor complex predictions, branch scenarios:\n\n```\nProject success: ?\n├── Technical risk resolves well (60%)\n│   ├── Team stays intact (80%) → 0.60 × 0.80 = 48% → SUCCESS\n│   └── Key person leaves (20%) → 0.60 × 0.20 × 0.50 = 6% → PARTIAL\n├── Technical risk causes delays (30%)\n│   ├── Scope reduced (60%) → 0.30 × 0.60 × 0.70 = 12.6% → SUCCESS\n│   └── Scope maintained (40%) → 0.30 × 0.40 = 12% → FAILURE\n└── Technical risk blocks project (10%) → 10% → FAILURE\n\nP(Success) = 48% + 12.6% = 60.6% ≈ 60%\n```\n\n### Step 4: Update with New Information\n\nWhen new evidence arrives, update:\n\n```markdown\nOriginal estimate: 65% hit revenue target\n\nNew information: Enterprise deal delayed to Q3\nImpact: -15% (was +15% if closed, now neutral)\nUpdated estimate: 50%\n\nNew information: Competitor launch was weak\nImpact: +10% (was -20% if aggressive)\nUpdated estimate: 60%\n```\n\n### Step 5: Record and Track\n\nKeep a prediction log:\n\n```markdown\n## Prediction Log\n\n| Date | Prediction | Probability | Actual | Brier Score |\n|------|------------|-------------|--------|-------------|\n| 2024-01-01 | Q1 launch | 70% | Yes | 0.09 |\n| 2024-01-15 | Deal closes | 60% | No | 0.36 |\n| 2024-02-01 | Bug resolved in 1 week | 80% | Yes | 0.04 |\n```\n\n### Step 6: Calibrate Over Time\n\nReview your accuracy:\n\n```markdown\n## Calibration Review\n\nFor predictions I rated 70%:\n- Total predictions: 20\n- Actual outcomes \"Yes\": 12 (60%)\n- I'm overconfident by 10% at this level\n\nAdjustment: When I feel \"70%\", actual is closer to 60%\n```\n\n## Calibration Techniques\n\n### The Equivalent Bet Test\n\n\"Would I bet at these odds?\"\n\n```\nPrediction: 80% confident project finishes on time\nEquivalent: Would I bet $4 to win $1?\nIf that feels wrong, adjust the probability.\n```\n\n### The Outside View\n\nAlways check base rates:\n\n```\nInside view: \"Our team is great, we'll definitely finish on time\"\nOutside view: \"What % of similar projects finished on time?\"\n\nInside tends toward overconfidence\nOutside provides calibration anchor\n```\n\n### The Pre-Mortem Adjustment\n\nImagine failure, then adjust:\n\n```\nInitial estimate: 85% success\nAfter pre-mortem: Identified 5 failure modes I hadn't considered\nAdjusted estimate: 70%\n```\n\n### The Confidence Interval Check\n\nAre your intervals too narrow?\n\n```\nTest: Of your 90% confidence intervals, do 90% contain the actual?\nCommon finding: Only 60-70% do\nFix: Widen intervals by 50%\n```\n\n## Application Examples\n\n### Project Estimation\n\n```markdown\n## Project: Payment System Rewrite\n\nTimeline estimate:\n- 50% confidence: 8-12 weeks\n- 80% confidence: 6-16 weeks\n- 95% confidence: 4-24 weeks\n\nKey variables:\n- API complexity: High uncertainty (+/- 3 weeks)\n- Team availability: Medium uncertainty (+/- 2 weeks)\n- Integration testing: High uncertainty (+/- 4 weeks)\n\nCommitment: \"We're 80% confident we'll deliver in Q2\"\n```\n\n### Risk Assessment\n\n```markdown\n## Risk: Database migration causes extended downtime\n\nProbability assessment:\n- Base rate for similar migrations: 20% have issues\n- Our preparation level: Above average (-5%)\n- Complexity of our schema: Above average (+5%)\n- Rollback plan quality: Strong (-5%)\n\nEstimate: 15% probability of extended downtime\n\nMitigation value:\n- If issue occurs: 4 hours downtime × $10K/hour = $40K\n- Expected loss: 15% × $40K = $6K\n- Mitigation cost: $3K for additional testing\n- Decision: Mitigation worth it (ROI positive)\n```\n\n### Technical Decision\n\n```markdown\n## Decision: Adopt new framework\n\nSuccess probability factors:\n| Factor | Probability | Weight |\n|--------|-------------|--------|\n| Team learns quickly | 70% | 0.3 |\n| Framework matures | 80% | 0.2 |\n| Performance meets needs | 60% | 0.3 |\n| Integration works | 75% | 0.2 |\n\nCombined probability (simplified):\n0.70 × 0.80 × 0.60 × 0.75 = 25% (if all must succeed)\nOR weighted average: 70% (if partial success acceptable)\n\nDecision: High uncertainty suggests pilot first\n```\n\n## Brier Score for Calibration\n\nTrack prediction accuracy with Brier Score:\n\n```\nBrier Score = (probability - outcome)²\n\nWhere outcome = 1 if happened, 0 if not\n\nExample:\nPredicted 70% (0.70), it happened (1)\nBrier = (0.70 - 1)² = 0.09\n\nPredicted 70% (0.70), it didn't happen (0)\nBrier = (0.70 - 0)² = 0.49\n\nLower is better. Perfect = 0, Random = 0.25\n```\n\n## Probabilistic Thinking Template\n\n```markdown\n# Probabilistic Assessment: [Prediction]\n\n## Prediction\n[Clear, falsifiable statement with timeframe]\n\n## Initial Probability\nEstimate: [X]%\nBase rate: [Similar events: Y%]\nAdjustment rationale: [Why different from base rate]\n\n## Confidence Interval\n- 50% CI: [Range]\n- 80% CI: [Range]\n- 95% CI: [Range]\n\n## Key Uncertainties\n| Uncertainty | If positive | If negative |\n|-------------|-------------|-------------|\n| [Factor 1] | +X% | -Y% |\n| [Factor 2] | +X% | -Y% |\n\n## Update Log\n| Date | New Information | Old P | New P |\n|------|-----------------|-------|-------|\n| | | | |\n\n## Resolution\nDate: [When known]\nOutcome: [What happened]\nBrier Score: [Calculation]\nLessons: [What to learn]\n```\n\n## Verification Checklist\n\n- [ ] Expressed prediction as specific probability\n- [ ] Checked base rate for similar events\n- [ ] Created appropriate confidence intervals\n- [ ] Identified key uncertainties and their impacts\n- [ ] Recorded prediction for future calibration\n- [ ] Applied equivalent bet test for sanity check\n- [ ] Willing to update when new information arrives\n\n## Key Questions\n\n- \"What probability would I assign to this?\"\n- \"What's the base rate for similar things?\"\n- \"What would change my estimate up or down?\"\n- \"Am I being overconfident? (Usually yes)\"\n- \"What's my track record at this confidence level?\"\n- \"Would I bet at these odds?\"\n\n## Tetlock's Superforecaster Traits\n\n1. **Update often:** Change predictions when evidence changes\n2. **Granular probabilities:** Use 65% not \"likely\"\n3. **Outside view:** Start with base rates\n4. **Seek disconfirming evidence:** Look for reasons you're wrong\n5. **Track record:** Keep score, learn from errors\n6. **Intellectual humility:** Know you're often wrong\n\n## Tetlock's Wisdom\n\n\"The fox knows many things, but the hedgehog knows one big thing.\"\n\nSuperforecasters are foxes—they integrate many perspectives, update frequently, and avoid ideological certainty. They're not smarter; they're more calibrated.\n\n\"Beliefs are hypotheses to be tested, not treasures to be protected.\"\n\nYour predictions should change as evidence changes. Holding steady when you should update is a calibration failure.\n"},"import":{"commit_sha":"a31e22d4445ad8fef7cd771d32af537aebb68c49","imported_at":"2026-05-22T21:14:39Z","license_text":"MIT License\n\nCopyright (c) 2025 TJ Boudreaux\n\nPermission is hereby granted, free of charge, to any person obtaining a copy\nof this software and associated documentation files (the \"Software\"), to deal\nin the Software without restriction, including without limitation the rights\nto use, copy, modify, merge, publish, distribute, sublicense, and/or sell\ncopies of the Software, and to permit persons to whom the Software is\nfurnished to do so, subject to the following conditions:\n\nThe above copyright notice and this permission notice shall be included in all\ncopies or substantial portions of the Software.\n\nTHE SOFTWARE IS PROVIDED \"AS IS\", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR\nIMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,\nFITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE\nAUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER\nLIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,\nOUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE\nSOFTWARE.\n","owner":"tjboudreaux","repo":"tjboudreaux/cc-thinking-skills","source_url":"https://github.com/tjboudreaux/cc-thinking-skills/tree/a31e22d4445ad8fef7cd771d32af537aebb68c49/skills/thinking-probabilistic"}},"content_hash":[236,76,211,50,253,222,174,160,98,230,184,175,161,172,164,239,64,201,215,250,230,157,205,25,25,128,39,14,98,159,124,229],"trust_level":"unsigned","yanked":false}