About Nodal
How Nodal Recommends Games
A map for discovering your next favorite game.
Nodal is a solo-built project that blends a recommender system with a visual explorer. I wanted discovery to feel less like browsing a catalog and more like exploring a world. Search a game (or a whole library), and Nodal shows nearby games and why they are nearby.
Why I built it
I love machine learning and statistics, and I am also a heavy gamer. I kept running into the same problem: discovery is either too generic (top sellers, trending, ads) or too narrow (only games that look identical on paper). I wanted a tool that surfaces both obvious neighbors and the surprisingly good "people who liked this also got weirdly into that" picks.
The spark was honestly mundane. One night I was watching Survivor with my mom and the "similar" row popped up. Then the question came to mind: are those picks based on content (more Survivor variants, like Survivor Australia), or based on viewer behavior (the overlap between people who watch Survivor and people who watch something else, like The Amazing Race)? That question maps cleanly to games: content similarity is useful, behavior similarity is useful, and the best results usually come from blending signals instead of betting everything on one idea.
What Nodal is trying to be
- A fast way to answer "what should I play after this?"
- A visual map of the catalog so you can explore clusters and niches
- A place where smaller, older, and stranger games still have a shot at being found
The interface is inspired by node-based research tools like Connected Papers, adapted for exploring games instead of papers.
How recommendations work
Nodal combines a few signals. Think of them as different lenses on the same question: "what's close to this?"
1) Collaborative filtering
This is behavior-based similarity. Games that are frequently owned or played by the same people tend to move closer together. This often works extremely well. For example, players who enjoy The Witcher 3 often also play Red Dead Redemption 2 or Cyberpunk 2077, even if the games are not identical on the surface.
Sometimes this reflects shared player habits more than true similarity. For example, Portal 2 may appear near Half-Life 2 or Garry's Mod because many players tend to own or play those classics together.
2) Tags and metadata similarity
This is content-based similarity. Steam tags are surprisingly expressive and can help balance out the community model. Tag similarity helps when behavior is sparse, and it helps explain recommendations in human language.
3) Blended ranking
The final list is usually a blend. Behavior can pull in "popular but not actually similar" games. Tags can keep results coherent, but tags alone can be too literal or inaccurate, especially for smaller or newer titles. Blending aims to balance both signals for the best results.
4) Re-ranking controls
You can re-rank results using broader signals like popularity, quality, or recency. This does not change which games are considered similar, only how the final list is ordered.
What the map is
The map is a 2D projection of a much higher-dimensional representation of games. Nodal uses a dimensionality reduction algorithm called UMAP to compress the similarity structure into a plane so you can see clusters form. Nearby points often share themes and tags, but the map is still an approximation.
The map is meant for exploration, not as a scientific claim that distance equals an exact numeric truth. It is simply a navigational tool.
Data and what I do not collect
Nodal uses publicly available Steam information and public library data for Steam IDs you choose to enter.
- I do not ask for your Steam password
- I do not sell personal data
- I do not run ad targeting profiles
If your Steam profile is private, Nodal cannot fetch your library and personalized results may be limited.
Roadmap
- Better controls for blending and reranking signals
- Clearer "why this was recommended" explanations
- Stronger filters (budget, session length, etc.)
- Performance work so the explorer stays snappy as the catalog grows
Disclaimer
Nodal is an independent project and is not affiliated with Valve or Steam. Trademarks, logos, and game assets belong to their respective owners and are used for identification. Recommendations are informational only. Pricing, discounts, and availability can change without notice.
Support
Nodal is free to use with no ads or paywalls. If you find it useful and want to support continued development, you can buy me a coffee. Totally optional.
Built by Nicholas Amirsoleimani. If you have feedback, ideas, or see bugs, the fastest way to reach me is the Contact page.