---
title: "Comment by OnlineProxy on \"How I Built a Free Daily Price Scraper with Playwright, GitHub Actions and Google Sheets\" | Hashnode"
description: "so the read_pairs thing, the first match is basically random cause set order isnt fixed, so id rank the candidates and only trust the best one if it clearly bea"
url: "https://hashnode.com/posts/how-i-built-a-free-daily-price-scraper-with-playwright-github-actions-and-google-sheets/6af7fcf2fe28fc546e49ccb0/comment/6af9a74eb0bfe49a3cef5410"
---

[Discussion](https://hashnode.com/posts/how-i-built-a-free-daily-price-scraper-with-playwright-github-actions-and-google-sheets/6af7fcf2fe28fc546e49ccb0)

Comment [How I Built a Free Daily Price Scraper with Playwright, GitHub Actions and Google Sheets](https://hashnode.com/posts/how-i-built-a-free-daily-price-scraper-with-playwright-github-actions-and-google-sheets/6af7fcf2fe28fc546e49ccb0)

[OnlineProxy](https://hashnode.com/@onlineproxyio)

[2d ago](https://hashnode.com/posts/how-i-built-a-free-daily-price-scraper-with-playwright-github-actions-and-google-sheets/6af7fcf2fe28fc546e49ccb0/comment/6af9a74eb0bfe49a3cef5410)

so the read\_pairs thing, the first match is basically random cause set order isnt fixed, so id rank the candidates and only trust the best one if it clearly beats the runner up, otherwise just flag it. taxes or shipping can totally wreck the total math too, so keep writing flagged rows with their status, dont drop them

the stability check is kinda useless tbh, if the configurator keeps showing the old price two identical reads just agree on the wrong number, comparing against yesterday's price for the same site catches that way better

[Shijith Pulikkal](https://hashnode.com/@shijithpulikkal)

[2d ago](https://hashnode.com/posts/how-i-built-a-free-daily-price-scraper-with-playwright-github-actions-and-google-sheets/6af7fcf2fe28fc546e49ccb0/comment/6af9a74eb0bfe49a3cef5410#reply-6af9ba814a838c8865a88704)

Thanks, this is useful. You’re right that read\_pairs can return an arbitrary match when more than one pair fits the maths. I’m changing it to collect all candidates and only trust the best one when it clearly beats the next, otherwise flag it. Rows that fail the total check are already saved with a CHECK status, not dropped, so shipping or tax surprises stay visible. And agreed that the stability check can agree on a stale number. I’m adding a comparison against the previous day’s price per site, flagged but not discarded, since real promotions can move prices.
