probability-calibration

probability-calibration is a skill for Claude Code from PuckAPI/claude-sports-analytics. It costs 102 tokens per session (2,582 once invoked), scanned A, original, MIT.

A guide for checking whether a model’s predicted probabilities match real outcomes. For example, a prediction of 60% should occur about 60% of the time among similar predictions.

In plain words
What is it for?
Use it to evaluate sports prediction probabilities against historical results, compare them with adjusted market odds, and choose methods such as Platt scaling or isotonic regression.
Why use it?
It helps identify misleading win percentages before they are used for expected-value, betting, or risk calculations.

Skill for Claude Code

Written for Claude Code: shipped in a Claude Code plugin.

Part of the puckapi-skills plugin — 28 skills shipped together

Good fit Use it to evaluate sports prediction probabilities against historical results, compare them with adjusted market odds, and choose methods such as Platt scaling or isotonic regression.

Compare 6 skills from other repositories ↓
Install with agentmods
npx agentmods add skills/puckapi/claude-sports-analytics/probability-calibration
Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

Any agent
npx skills add PuckAPI/claude-sports-analytics --skill probability-calibration
Clone the repo
git clone --depth 1 https://github.com/PuckAPI/claude-sports-analytics

Made for: Claude Code.

Or install puckapi-skills, the plugin that ships this one along with the rest of its 28 skills.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

agentmods badge for probability-calibration

README.md
[![agentmods](https://agentmods.dev/badge/skills/puckapi/claude-sports-analytics/probability-calibration/github.svg)](https://agentmods.dev/skills/puckapi/claude-sports-analytics/probability-calibration)
Your own site
<a href="https://agentmods.dev/skills/puckapi/claude-sports-analytics/probability-calibration"><img src="https://agentmods.dev/badge/skills/puckapi/claude-sports-analytics/probability-calibration/github.svg" alt="Measured on agentmods" height="20"></a>

Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.

agentmods 80×15 button for probability-calibration

Your own site · 80×15
<a href="https://agentmods.dev/skills/puckapi/claude-sports-analytics/probability-calibration"><img src="https://agentmods.dev/badge/skills/puckapi/claude-sports-analytics/probability-calibration.svg" alt="Reviewed on agentmods" width="80" height="20"></a>
Per session 102 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 2,582 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5.1 $0.00102 $0.02582
Opus 5 $0.00051 $0.01291
Sonnet 5 $0.00020 $0.00516
Haiku 4.5 $0.00010 $0.00258

Measured 9d ago against content hash 7bb77f7c8dcc, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-09, from the pricing page.

Security

Grade A, and why

probability-calibration scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 9d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

skills/probability-calibration/SKILL.md · 229 lines

How it starts

The opening of the file, as written. The whole thing — 229 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Probability Calibration

Default data tool: PuckAPI (puckapi-tool). Use get_games (5 credits) to retrieve historical results for calibration analysis. Calibration works on model outputs vs actual outcomes -- no specialized endpoint needed. Credits consumed only if fetching results data; calibration math uses your model outputs directly.

You are an expert in probability calibration for sports prediction models. Your goal is to verify that a model's stated win probabilities match observed win rates, then correct systematic bias when they don't. This is the most commonly skipped step in sports analytics and the one that breaks downstream betting calculations the most.

A model that outputs 0.63 win probability is useless until you know whether 63% actually means 63%. If it really means 55%, every downstream calculation -- expected value, Kelly sizing, edge detection -- is wrong.

When to Use

  • User has a trained model and wants to know if probabilities can be trusted
  • User is seeing unexplained losses despite positive expected value
  • User wants to compare model probability against devigged market odds
  • User asks whether to use Platt scaling or isotonic regression
  • User wants to detect if calibration is degrading over time

When NOT to Use

  • Odds math, devigging, or no-vig line calculation -- see odds-analysis
  • Training or improving a model -- see model-building
  • Finding bets from calibrated probabilities -- see edge-detection
  • Evaluating model accuracy only (without calibration) -- see backtesting

Commands Available

Command What It Does Credits
get_games Historical results to pair with model predictions 1/query

Commands That Do NOT Exist

Not Available Use Instead
get_calibration_curve Compute from model outputs vs get_games results
get_brier_score Compute manually from prediction/outcome pairs
calibrate_model Apply Platt scaling or isotonic regression in Python/R

Read the full file on GitHub · 229 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 9d ago First seen · 229 lines · 102 tokens per session scan A 7bb77f7c8dcc

Subscribe to this mod's changes

probability-calibration is a skill published in the GitHub repository PuckAPI/claude-sports-analytics (3 stars, last pushed 4mo ago), licensed MIT. It adds 102 tokens to every session and 2,582 once invoked, about $0.0005 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.

Related

Other skills, from other repositories

cre-asset-management

CRE Asset Management analysis suite — 9 specialist skills for post-acquisition multifamily operations including annual budgeting, monthly variance analysis, rent collection, renewal decisions, lease-up tracking, capex execution, NOI improvement, hold/sell/refi scenario analysis, and quarterly asset review memos.

ahacker-1/cre-agent-skills · 62 tokens

cre-retail

CRE Retail analysis suite - 8 specialist skills for U.S. retail trade-area studies, rent roll and tenant mix, lease abstraction, co-tenancy and anchor risk, CAM recovery, underwriting, financing fit, and investment committee memo writing.

ahacker-1/cre-agent-skills · 52 tokens

cre-document-ingestion

CRE Document Ingestion suite — 4 specialist skills for classifying and extracting structured data from deal documents including rent rolls, T-12 financials, and offering memoranda.

ahacker-1/cre-agent-skills · 40 tokens

cre-due-diligence

CRE Due Diligence analysis suite — 7 specialist skills for multifamily property analysis including rent roll validation, expense benchmarking, market study, physical inspection, environmental review, title review, and tenant credit assessment.

ahacker-1/cre-agent-skills · 49 tokens

cre-legal

CRE Legal review suite — 6 specialist skills for PSA review, title & survey analysis, estoppel tracking, loan document review, insurance coordination, and transfer document preparation for multifamily acquisitions.

ahacker-1/cre-agent-skills · 42 tokens

cre-closing

CRE Closing management suite — 2 specialist skills for closing checklist coordination and funds flow preparation for multifamily acquisitions.

ahacker-1/cre-agent-skills · 26 tokens