Breath of the Wild Decomp: 20 Percent Code and a Documentation Mission

By Indie Kings | October 28, 2026

Updated October 28, 2026: zeldaret/botw, the experimental decompilation of The Legend of Zelda: Breath of the Wild version 1.5.0 on Switch, reports 20.264 percent of code and 32.794 percent of functions reconstructed. Its newest commit landed today, October 28, 2026, and changes the build tools to force the linker to use clang hashes for ELF output. The repository dates to August 19, 2020 and has 2,401 stars, 146 forks and 34 contributors.

Image: a Nintendo Switch OLED, the console family Breath of the Wild 1.5.0 targets, via Wikimedia Commons.

Experimental by design

The README uses the word experimental in its first sentence, and it means it. This is a work in progress decompilation of Breath of the Wild version 1.5.0 on Switch, it contains no game assets or RomFS content, and the README states in italics that it cannot be used to play the game. That last point is worth dwelling on, because decompilation projects are often mistaken for ports or for a way to run a game without buying it. This one is explicitly neither. Its stated goal is to better understand the game's internals, to aid glitch hunting, and to document existing knowledge in a permanent, unambiguous form that helps further reverse engineering.

For a game the size of Breath of the Wild, that framing is the only honest one. A full matching decompilation of a modern open world title is a multi year effort involving enormous amounts of proprietary middleware, physics, animation and rendering code, and the result would not be a playable PC build even if it finished. What a project like this can deliver is a readable map of the parts people care about: the systems behind movement, memory layout, event flags, and the physics interactions that glitch hunters exploit. Documentation is the product, and the code is the medium.

Two numbers, two measures

The project publishes two progress badges rather than one. The code measure sits at 20.264 percent, and the function measure at 32.794 percent. The gap is not an error. Functions are counted individually, so a large number of small, simple routines can be matched early and push the function percentage ahead of the byte weighted code percentage. A single sprawling function in the rendering or physics layer can be worth more code than dozens of small accessors, and until it is matched, the code figure lags.

Reading the two together gives a better sense of where the project stands than either alone. Thirty two percent of functions means a substantial portion of the executable has been identified and rewritten, while twenty percent of code means the remaining work is concentrated in larger, harder routines. For glitch researchers, the function count is often the more useful figure, because the routines that matter to a specific trick are usually identifiable even when the surrounding systems are still unknown. For preservation, the code figure is the honest measure of how much of the original binary has actually been reconstructed.

Why linker hashes matter

Today's commit changes the build tools so the linker uses clang hashes for ELF output. In a matching decompilation, the hash of the final binary is the proof of correctness, and that requires every stage of the toolchain to be deterministic and to agree with the original build. Modern linkers can embed or compute hashes for symbol tables and sections, and small differences in how those hashes are produced can change the output even when the machine code is identical. Forcing a specific hashing behavior is the kind of fix that looks obscure and is essential if the goal is a reproducible binary.

The commit is numbered 235, which is a modest counter for a repository created in August 2020, and the contributor count is 34. Both numbers reflect the project's nature: this is not a broad community effort with hundreds of drive by contributors, it is a focused group working on a very large binary with a narrow public interface. The 21 open issues and 2,401 stars suggest strong interest from outside, and the README's pointer to botw.link gives that audience a place to follow progress without expecting a playable result.

Project card

FAQ

Can I play Breath of the Wild with this repository?
No. The README states that it contains no game assets or RomFS content and cannot be used to play the game.

What is the project for?
Understanding the game's internals, supporting glitch hunting, and documenting reverse engineering knowledge in a permanent form.

Why do the two progress numbers differ?
Functions are counted individually while code is weighted by size, so many small matched routines can push the function percentage above the code percentage.

Which version does it target?
Breath of the Wild 1.5.0 on Switch.

What changed on October 28?
The build tools now force the linker to use clang hashes for ELF output, which helps keep the matching build deterministic.

How active is the project?
A commit landed today, and the repository has 34 contributors and 21 open issues.

Bottom Line

Breath of the Wild is the decompilation where the honest framing matters most, because the gap between what fans want and what the project claims is widest. It says experimental, it says not playable, and it points at documentation as the goal, which is the correct ambition for a game of this scale. Twenty percent of code and thirty two percent of functions in six years is slow progress by the standards of a Game Boy title, and it is fast progress for a modern open world engine. The linker hash commit today is a reminder that even the plumbing has to be exact.

Sources

• GitHub zeldaret/botw README, experimental status, version target, asset and playability disclaimers, accessed October 28, 2026
• botw.link progress badge endpoints for code and function measures, accessed October 28, 2026
• GitHub API metadata for zeldaret/botw, stars, forks, contributors, issues and latest commit, accessed October 28, 2026

Related: Nxnative: Nintendo Switch Games Run Native on Apple Silicon | Cubyz: Zig Voxel Sandbox Rebuilt From Java With 3.7k Stars