How would you define "indie" for Steam market research? Working with ~63,800 games (2024–2026 cohort)
retained text
Much of the discussion around independent games on Steam relies on anecdotal examples, platform tags, or commercial reports with methodologies that aren't always transparent. Steam's own "Indie" tag isn't very useful for market research on its own: a solo developer releasing a first project with very limited resources can share the same tag with games published by established labels like Devolver Digital, Team17, or Annapurna. I'm analyzing Steam's independent game ecosystem through Indie Census, an open-source data pipeline (MIT license) built to study game engine adoption, pricing patterns, and commercial survival. While developing the classification methodology, I ran into a problem that I think is worth discussing: how can I distinguish independent developers from publisher-backed productions using publicly available Steam metadata? # Why the 2024–2026 cohort? My current dataset contains approximately 63,800 games released between 2024 and October 18, 2026. The 2026 portion is year-to-date, based on the latest metadata snapshot. Dataset extraction: Steam app metadata was extracted on October 18, 2026, while review summaries were extracted on October 28, 2026. This window gives me an opportunity to examine two industry developments: 1. Changes in game engine adoption: Tracking the adoption of open-source engines such as Godot against commercial engines like Unity and Unreal in the years following the late-2023 controversy around Unity's Runtime Fee. The goal is to observe adoption patterns, not to assume that one event caused them. 2. Steam's generative AI content disclosure: Valve introduced a dedicated generative AI disclosure section in its Steamworks Content Survey in January 2024. Its current documentation focuses on AI-generated content that ships with a game and is consumed by players, distinguishing between pre-generated and live-generated content. General efficiency gains from AI-powered development tools are not the focus of this disclosure. My pipeline records publicly available AI-related evidence using four internal labels: declared (the developer's disclosure through Steam), tagged (community tags), self\_reported (developer statements in descriptions or patch notes), and unknown (no public evidence found). These are my own evidence categories, not official Valve classifications. An unknown result does not necessarily mean that a game used no AI. However, before studying how engine choice or AI disclosure relates to commercial survival, I faced a more fundamental problem: I needed a consistent, reproducible way to classify what counts as an "indie" game. # The heuristic: indie_score (0.0 to 1.0) Rather than forcing every game into an unexplained binary label, I developed an experimental scoring heuristic based on five signals extracted from Steam metadata: 1. Tag / Genre Presence (Weight: 0.25): Whether the title has the official or community "Indie" tag. 2. Developer–Publisher Relationship (Weight: 0.25): Full score when the developer and publisher fields match (developer == publisher). When an external publisher is involved, the score is also influenced by its Steam catalog size: full points for catalogs with 5 or fewer games, scaling down to 0.4 for publishers with up to 25 games. 3. Corporate Exclusion & Publisher Scale (Weight: 0.25): A hard exclusion if the publisher matches a curated list of major corporate entities or subsidiaries (EA, Ubisoft, etc.), or if its Steam catalog exceeds 25 games. 4. Base Price at Extraction (Weight: 0.15): Uses Steam's undiscounted base price captured at extraction, not necessarily the price at the game's original launch. Full score for titles priced at $35 or less, decreasing linearly to 0 at $65. 5. Localization Breadth (Weight: 0.10): Full score for 8 or fewer supported languages, decreasing to 0 at 23. I use this as a rough and imperfect proxy for localization scope and potential resources, not as a direct measure of development budget or game quality. A game qualifies as indie if it reaches an indie\_score of at least 0.60 and passes the publisher exclusion check: it must not match a major corporate entity, and its publisher's Steam catalog must not exceed 25 games. Under the current rules, approximately 57,000 of the \~63,800 titles (89.4%) qualify as indie. This percentage describes the output of my current heuristic, not an objective estimate of the actual proportion of indie games on Steam. # The dilemmas & edge cases Before freezing this heuristic for my public Parquet dataset release, I'm evaluating three structural trade-offs and would appreciate feedback from developers and data analysts. 1. The boutique publisher problem & overlapping signals The 25-game catalog cutoff excludes established indie-focused publishers such as Devolver Digital, Annapurna, and Hooded Horse. It can therefore exclude games like Cult of the Lamb or Inscryption, even though many players and developers would consider them indie. There's also a potential overlap between signals 2 and 3: signal 2 already uses publisher catalog size to adjust the developer–publisher relationship score, while signal 3 uses that same catalog size as part of a hard exclusion rule. Question: Would you separate developer self-publishing from publisher size to avoid counting the same factor twice? Or would a third category, such as "Publisher-backed Indie" or "Triple-I," better represent games backed by established indie-focused publishers without grouping them with major corporate releases? 2. Independence vs. production quality (asset flips & spam) A $0.99 asset flip, uploaded by a newly created account and available in only one language, could receive a near-perfect score of 1.0. That's not necessarily a mathematical error: the heuristic is designed to approximate structural independence, not development effort or production quality. But including large amounts of low-effort spam could distort market benchmarks. Question: I don't think I can simply remove games with fewer than 10 reviews without potentially introducing survivorship bias into my survival analysis. Doing so would remove some of the commercial failures I'm trying to study. How would you distinguish low-effort spam from legitimate low-budget indie releases without filtering out games simply because they failed commercially? 3. Community-driven reach vs. budget proxies A solo developer with an active Discord or Crowdin community might support 15 or more languages through volunteer translations, at little or no direct financial cost to the developer. Under my current model, that broader language support lowers the game's indie\_score, even though it may reflect community involvement rather than a larger localization budget. Question: Is language count too noisy to serve as a proxy for localization resources, especially when community translations are involved? Would you reduce its weight, remove it, or look for another signal? If you were using this dataset to benchmark your own game's engine choice, pricing strategy, or release planning, how would you refine these boundaries?
document fetched 2026-10-31 00:25 · canonical https://reddit.com/r/gamedev/comments/1xjlk3n/how_would_you_define_indie_for_steam_market