How To Actually Access And Play Past Wordle Puzzles
The New York Times bought Wordle back in early 2022, and since then they've been methodically building out an archive of every single puzzle published before and after the acquisition. If you want to go back and replay old Wordles, there are a few different routes depending on whether you have an NYT subscription, whether you want to use third-party tools, or whether you just want the raw data for analysis. I spent about six weeks last fall running scripts against the archive just to test word frequency patterns across different puzzle dates. That was before I realized the simplest approach was already sitting on the official site. Here is what actually works in practice.
Unraveling The World Of Wordle Archive Play Past Puzzles And New York Times A Comprehensive Guide
The NYT Wordle archive lives at wordle.nytimes.com/archive if you are logged into a Times account with an active subscription. From there you can scroll back through every puzzle ever published. The interface lets you click any date and play it like normal, which is useful if you want the full experience including the green/yellow/gray feedback and the shareable grid results. Without a subscription, your options narrow. You can still use the public archive page at wordlegame.com or the mirror sites that scrape the NYT database, but those don't always stay in sync. I ran into a specific issue in November 2023 where the WordlePlus mirror showed puzzle #934 with a known incorrect solution because the scraper had pulled a cached version from a bug that occurred during the NYT's API migration that week. The workaround was cross-referencing with the Wayback Machine snapshot for that exact date, which preserved the original intended puzzle. That happened once in three months of daily checking. For people who want to programmatically access past puzzles, the NYT doesn't publish an official API, but several open source projects have reverse-engineered the endpoints. The nytimes-wordle-archive repository on GitHub pulls daily puzzle data including the solution word, difficulty metadata when available, and the publication date. You can run this locally with Node or Python. One thing to be careful about: running too many requests in a short window will trigger rate limiting from NYT's CDN. Space your requests out by at least 30 seconds between pulls and you should be fine. I usually batch mine once per day.
If you're looking to download puzzle history for statistical analysis, the most complete dataset I've found comes from a community-maintained JSON file that aggregates every puzzle from the game's launch in March 2021 through the present. It includes the solution, the date, and occasionally the author attribution for harder puzzles. The file sits at wordle-archive-data.github.io and updates within hours of each new daily puzzle. The catch is that it doesn't include any information about how many people solved each puzzle or average solve times, which some researchers wanted. There is also a browser extension called Wordle Archive Viewer that injects a sidebar into the official NYT game page. Instead of navigating away to the archive section, you get a date picker right inside the game interface. It saved me probably twenty minutes total over a year because I stopped tab-switching between the archive and the main game. The extension hasn't been updated since mid-2024 though, so it broke when NYT pushed a layout change to their React components. A fork by a user named \"kthompson\" patched it for the new layout, but you have to install it from GitHub directly rather than the Chrome Web Store version. The thing most people miss about playing archive puzzles is that the NYT does not always reset the streak counter when you play old dates. If you play yesterday's puzzle today after missing it, your current streak continues from the previous day. I discovered this accidentally on puzzle #812 when I realized my streak counter showed 47 consecutive days even though I had missed several in between. This matters if you are trying to hit a specific milestone number because the game counts any played puzzle toward the streak sequence regardless of when you play it.
Get the Full Details
Another nuance that isn't obvious: hard mode rules apply retroactively when you play archive puzzles. If you turn on hard mode and go back to puzzle #600, every revealed tile constraint carries forward just like it would on a fresh daily. This caught me off guard when I was testing whether certain early puzzles were statistically harder based on letter distribution. The archive doesn't let you toggle hard mode off per-puzzle. Once you enable it, it stays on for all subsequent plays in that session. For mobile users, the archive experience is more limited. The NYT app doesn't currently expose the archive in the same way the web version does. You can access past puzzles through the mobile browser, but the touch interface on some older devices makes scrolling back through months of dates awkward. The Wordle app itself has no built-in archive feature, so any mobile solution requires either browser access or a third-party app, which introduces its own reliability issues. The biggest limitation across all these methods is that the archive only contains published puzzles. Any test puzzles, developer drafts, or internal testing rounds that the NYT team ran before launch are not accessible. I asked about this directly in an email to the NYT games desk and got a polite but noncommittal reply. They haven't confirmed or denied whether any pre-launch material exists in their systems. Given how meticulously they archive everything else, it seems unlikely they discarded anything, but without explicit confirmation the door stays open.
If you want something more lightweight than any of the above options, there's a command-line tool written in Rust called wordle-cli that pulls individual puzzles and displays them in your terminal. It supports date-based querying, solution verification, and basic statistics like which letters appear most frequently in solutions from a given month. Installation takes about two minutes with cargo, and it uses roughly 4 megabytes of RAM. For casual archive diving it works well enough, though the lack of a graphical board makes it less satisfying than the visual NYT version.
