--- title: Your Model’s 0.9 Probably Isn’t 90% | daily.dev description: Probability scores from ML models like random forests or gradient boosted trees are often not true probabilities. A model scoring 0.9 doesn't necessarily mean... canonical: https://daily.dev/posts/your-model-s-0-9-probably-isn-t-90--fhtgyspvh twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: Your Model’s 0.9 Probably Isn’t 90% | daily.dev og:description: Probability scores from ML models like random forests or gradient boosted trees are often not true probabilities. A model scoring 0.9 doesn't necessarily mean... og:url: https://daily.dev/posts/your-model-s-0-9-probably-isn-t-90--fhtgyspvh og:image: https://api.daily.dev/og/posts/FHTgySPVh.png og:image:alt: Your Model’s 0.9 Probably Isn’t 90% og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # Your Model’s 0.9 Probably Isn’t 90% **[Medium](https://daily.dev/sources/medium_js)** · 12 min read · 0 upvotes · 0 comments ## Summary Probability scores from ML models like random forests or gradient boosted trees are often not true probabilities. A model scoring 0.9 doesn't necessarily mean 90% of those samples are positive. This piece explains model calibration: when it matters (cost-sensitive decisions, thresholds), why different model families miscalibrate differently, and how to fix it using scikit-learn's three calibration methods — sigmoid (Platt scaling), isotonic regression, and temperature scaling. Results are validated on the IBM Telco churn dataset. Key findings: Naive Bayes benefits enormously from calibration; logistic regression is already well-calibrated; isotonic regression overfits with small calibration sets (374 positives proved insufficient); temperature scaling is best for multiclass problems. Practical guidance covers disjoint data splits, using FrozenEstimator for pre-trained models, and monitoring calibration drift in production. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Similar posts on daily.dev - [The AI Model Confidence Trap](https://daily.dev/posts/the-ai-model-confidence-trap-93to0xhgy) · Towards Data Science · 5 upvotes · 1 comments - [Why 90% Accuracy Doesn’t Mean You Should Trust an AI Model](https://daily.dev/posts/why-90-accuracy-doesn-t-mean-you-should-trust-an-ai-model-vttihmbg7) · Medium · 0 upvotes · 0 comments - [Part 3: Knowing when your agent doesn’t know: the confidence layer](https://daily.dev/posts/part-3-knowing-when-your-agent-doesn-t-know-the-confidence-layer-7vx9rreeu) · Stack Overflow Blog · 0 upvotes · 0 comments - [Consistent security model deployment with FPR calibration](https://daily.dev/posts/consistent-security-model-deployment-with-fpr-calibration-jhnpk6vcj) · Cisco · 0 upvotes · 0 comments --- Tags: [#machine-learning](https://daily.dev/tags/machine-learning), [#python](https://daily.dev/tags/python), [#classification](https://daily.dev/tags/classification), [#scikit](https://daily.dev/tags/scikit) [View this post on daily.dev](https://daily.dev/posts/your-model-s-0-9-probably-isn-t-90--fhtgyspvh) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"Your Model’s 0.9 Probably Isn’t 90%","url":"https://daily.dev/posts/your-model-s-0-9-probably-isn-t-90--fhtgyspvh","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/your-model-s-0-9-probably-isn-t-90--fhtgyspvh"},"datePublished":"2026-07-18T06:47:44.321Z","dateModified":"2026-07-18T06:54:30.355Z","description":"Probability scores from ML models like random forests or gradient boosted trees are often not true probabilities. A model scoring 0.9 doesn't necessarily mean...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/a2132eec096728d40bdf612759fa27ef?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/a2132eec096728d40bdf612759fa27ef?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"Medium","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"Medium","logo":"https://media.daily.dev/image/upload/t_logo,f_auto/v1/logos/medium","url":"https://daily.dev/sources/medium_js"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/your-model-s-0-9-probably-isn-t-90--fhtgyspvh","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"machine-learning,python,classification,scikit","timeRequired":"PT12M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Medium","item":"https://daily.dev/sources/medium_js"},{"@type":"ListItem","position":3,"name":"Your Model’s 0.9 Probably Isn’t 90%"}]} ```