# Developer portfolios: what carries across releases? September 5, 2026 review snapshot; analysis completed September 6. All quantitative work used the saved data. A bounded official-source lookup corrected one Steam release date. Source tables were not changed. ## Main reading A developer's largest game is often a poor benchmark for judging the rest of their work. Currently strong catalogs are usually uneven. Yet strong reception elsewhere in a catalog remains associated with a later game's response, even when the immediately preceding game has few reviews. Long gaps among developers who return do not show a simple monotonic disappearance of that association. Automated tag distance is too unreliable to turn into a claim about creative departure. All review totals are observed now. This is **not** an estimate of release-time audience or causal audience transfer. Earlier games can gain reviews after later games release, even because of them. Skills, reputation, resources, scope, publishers and selection cannot be separated by this snapshot. Thresholds measure response scale, not sales, profitability or artistic value. ## Data and identity The study starts with 102,394 currently paid, valid dated released game records outside explicit-content descriptors 3/4. Full developer credits, not just the first listed developer, determine eligibility. Excluding joint/missing credits, ambiguous names and explicitly named demos/prologues/playtests/previews leaves 93,013 games credited to 59,259 normalized developer names. An identity audit found that Steam creator-page IDs can merge many different developers: one page represented 93 names in the preliminary eligible table. The main study groups by normalized credit name instead. Unverified name changes are not merged. This avoids manufactured careers but leaves incomplete catalogs and possible homonyms. Credits may identify individuals or organizations of any size; the sample is not a verified indie/solo population. Catalog shapes use 7,427 developers with 2–10 included games through 2025, first observed release in 2010–2023 and latest in 2020–2025. Follow-up comparisons use 6,721 developers with a qualifying consecutive paid-release pair ending in 2023–2025, at a gap of 30–3,650 days. Only the latest qualifying pair per developer gets a vote. Other periods and alternative credit/history filters are retained. ## A dominant title can coexist with a substantial body of work Among the 1,248 small catalogs containing at least one game with 1,000 reviews: - The biggest title contributes a median **76.9%** of total catalog reviews. - **565/1,248 (45.3%)** have at least 80% of reviews in that one title. - **1,065/1,248 (85.3%)** have another game with at least 100 reviews. - **568/1,248 (45.5%)** have another game with at least 1,000 reviews. - The median second-largest game has **847 reviews**. The best game outside the peak has fewer than 100 reviews for 183 catalogs (14.7%), 100–999 for 497 (39.8%), and 1,000+ for 568 (45.5%). Concentration in one title and meaningful response elsewhere are compatible. For the 307 catalogs with a 10,000-review peak, median peak share is 79.6%, 286 (93.2%) have another 100-review game, 240 (78.2%) have another 1,000-review game, and the median second-largest title has 4,595 reviews. Catalog size matters: among 1,000-review-peak catalogs, median peak share is 80.3% for 2–5 titles and 53.4% for 6–10. Half-share concentration is automatic in a two-title catalog, so that threshold alone is not interesting. The biggest current title is the first observed release in 573/1,248 catalogs (45.9%), leaving 675 (54.1%) whose largest title comes later. Using the highest release-quarter review percentile instead of raw count reduces the first-title share to 32.9%. This demonstrates the influence of accumulation time; it does not establish a learning effect or describe developers who did not return. ## Current reception is strongly associated across consecutive games | Previous game's current reviews | Developers | Follow-up median | Follow-up >=100 | Follow-up >=1,000 | Follow-up >=100 and >=80% positive | |---|---:|---:|---:|---:|---:| | <10 | 2,668 | 2 | 99 (3.7%) | 15 (0.6%) | 55 (2.1%) | | 10–99 | 2,345 | 17 | 389 (16.6%) | 87 (3.7%) | 268 (11.4%) | | 100–999 | 1,125 | 116 | 609 (54.1%) | 136 (12.1%) | 426 (37.9%) | | 1,000–9,999 | 458 | 742 | 378 (82.5%) | 187 (40.8%) | 264 (57.6%) | | 10,000+ | 125 | 3,148 | 113 (90.4%) | 89 (71.2%) | 67 (53.6%) | The raw reference for first observed eligible paid releases in 2023–2025 is 2,842/22,388 (12.7%) reaching 100 reviews. Returning developers overall have 1,588/6,721 (23.6%). First observed is not inexperienced; returning developers are selected. The broad returning average hides a far stronger relationship with the reception elsewhere in their catalogs. After release-quarter/current-price standardization, the 100-review observed / expected ratios for the five predecessor bands are 0.21, 0.74, 1.73, 2.10 and 2.02 relative to the full returning-developer reference. Sparse cells fall back; this is descriptive standardization, not a causal model. The strong-predecessor relationship survives useful restrictions: | Restriction | Earlier game 1,000–9,999: follow-up >=100 | Earlier game 10,000+: follow-up >=100 | |---|---:|---:| | Main study | 378/458 (82.5%) | 113/125 (90.4%) | | Current included catalog has 2–10 games | 363/438 (82.9%) | 104/116 (89.7%) | | No recorded shared franchise | 223/278 (80.2%) | 65/73 (89.0%) | | Unshared creator-page identities | 303/364 (83.2%) | 87/97 (89.7%) | | Include all developer credits, including joint credits | 453/558 (81.2%) | 163/183 (89.1%) | History-exclusion results are retained in `date_history_sensitivity.csv`. Missing franchise records are not proof of a new IP, and metadata cannot verify team size. ![Current reception across consecutive games](followup_reception.png) ## Smaller than the previous game is common Among 583 pairs whose predecessor currently has 1,000+ reviews: - 486 (83.4%) have fewer current reviews than their predecessor. - The median follow-up/predecessor count ratio is 24.2%. - Nevertheless, 491 (84.2%) reach 100 reviews and 276 (47.3%) reach 1,000. - 185 (31.7%) exceed the predecessor's percentile within its release-quarter pool. Older games have had longer to accumulate reviews, and selecting substantial predecessors induces regression toward less extreme outcomes. These are not estimates of audience retention or loss. They do show why a successor can look small beside an earlier game while still having substantial response on its own. ## Older catalog strength matters even after a quiet immediate predecessor Among 2023–2025 follow-ups with a predecessor currently below 100 reviews, compare only third-or-later eligible releases: | Earlier catalog | Developers | Follow-up >=100 | Follow-up >=1,000 | Median follow-up reviews | |---|---:|---:|---:|---:| | No older 1,000-review game | 2,100 | 189 (9.0%) | 36 (1.7%) | 7 | | An older 1,000-review game | 103 | 52 (50.5%) | 22 (21.4%) | 107 | Among catalogs with at most ten included titles, the 100-review rates are 182/1,843 (9.9%) versus 44/84 (52.4%). Restricting current follow-up prices to $5–$20 gives 128/943 (13.6%) versus 26/54 (48.1%). These remain observational and do not control all differences in skill, resources, scope or prior reach. The association also appears in the other periods: among the broader low-predecessor comparison, 34/74 in 2019–2022 and 20/53 in Jan–Aug 2026 reach 100 reviews when an older 1,000-review game exists. Exposure times differ, so the falling raw rates must not be read as a trend. The immediately preceding release is an incomplete summary of a catalog. This does not prove that releasing a quiet game leaves a reputation untouched. ## Long gaps among returners do not show a simple monotonic penalty Restrict to predecessors currently at 1,000+ reviews: | Recorded release gap | Developers | Follow-up >=100 | >=1,000 | Median follow-up reviews | |---|---:|---:|---:|---:| | 30–180 days | 24 | 17 (70.8%) | 10 (41.7%) | 574 | | 181–365 days | 44 | 39 (88.6%) | 23 (52.3%) | 1,021.5 | | 1–2 years | 111 | 88 (79.3%) | 55 (49.5%) | 959 | | 2–4 years | 173 | 148 (85.5%) | 84 (48.6%) | 914 | | 4–10 years | 231 | 199 (86.1%) | 104 (45.0%) | 854 | Across all predecessor bands, after date/current-price and predecessor-band standardization, 100-review ratios for those gaps are approximately 0.80, 0.96, 1.06, 1.01 and 1.04. The impressive raw long-gap results largely flatten after accounting for the kinds of predecessors and current prices in each group. These gaps are between recorded eligible Steam releases. They are not development times or measures of absence from public life. They omit people who did not return, and some are distorted by ports, editions, changed credits and missing dates. The data do not establish that waiting is beneficial or costless. They do not provide evidence for a simple career expiry clock either. ![Older catalog context and release gaps](catalog_context.png) For old first-release cohorts (2010–2020), 480/1,117 names whose first included game now has 1,000–9,999 reviews have no later eligible paid game recorded; the same is true for 161/347 with 10,000+. Those are 43.0% and 46.4%. This is not retirement: free work, co-development, other platforms and changed names can all be missing. It demonstrates the selection in a study of returning developers. ## Artistic departure cannot be reduced to tag overlap Among strong-predecessor pairs, top-10 tag-overlap categories show 100-review rates of 68/72 (94.4%) for close, 144/165 (87.3%) for middle and 279/346 (80.6%) for distant. Top-20 tags retain the same broad ordering. These are not defensible estimates of the cost of changing artistic direction. The metric labels BYTEPATH → SNKRX distant (top-10 Jaccard 0.059), despite their known continuity. Audits also identify sequels and closely related activities as distant. Of the 23 valid pairs in an additional outcome-hidden pilot of the distant, strong-predecessor group, 15 have a clear change of advertised activity, six are related variations and two are unclear. Some clear changes still retain a fictional universe, characters, tone or audience. One of the original 24 pairs was invalidated by the official date correction and was not silently replaced. Selected examples, with current review totals: | Developer | Earlier → later game | Earlier → later reviews | Later positive | |---|---|---:|---:| | System Erasure | ZeroRanger → Void Stranger | 2,030 → 3,280 | 91% | | RocketPunch Games | HARDCORE MECHA → Anomaly Collapse | 2,329 → 567 | 85% | | Steel Mantis | Valfaris → Valfaris: Mecha Therion | 1,771 → 300 | 94% | | Tributary Games | King Of The Castle → Occlude | 1,401 → 341 | 87% | | Ludosity | Slap City → Heartwood Heroes | 2,409 → 32 | 96% | | Failbetter Games | Sunless Skies → Mask of the Rose | 2,656 → 260 | 76% | A familiar identity can persist through a new activity, but substantial prior reception does not guarantee substantial reception for every departure. Public descriptions and counts do not tell us whether the same players followed. ## Whole catalogs illustrate the alternatives more clearly These examples are deliberately selected, not representative samples. Counts are current Steam filtered reviews, not sales. Paid sequences omit other work unless explicitly noted. - **Suspicious Developments:** Gunpoint (10,274) → Morphblade (264) → Heat Signature (6,851) → Tactical Breach Wizards (11,457). The free Floating Point (3,854) is also credited and is excluded from the paid sequence. A quiet smaller release coexists with later major releases; it cannot fairly summarize the career. - **Daniel Mullins Games:** Pony Island (14,761) → The Hex (4,480) → Inscryption (134,744). Inscryption's scale makes the other games look small in a catalog share, but their own reception is substantial. - **Terry Cavanagh:** VVVVVV (6,648), Super Hexagon (18,357), Dicey Dungeons (10,418), Terry's Other Games (61). Different kinds of release exist under the same credit; the last collection is not interchangeable with a new major game. - **Hopoo Games:** Risk of Rain (20,767) → DEADBOLT (4,851) → Risk of Rain 2 (239,710) → Risk of Rain Returns (24,781). Original departures and returns to a recognizable property both appear; shared-world and maker identity are not mutually exclusive explanations. - **increpare games:** English Country Tune (185), Stephen's Sausage Roll (1,583), Hypnocult (42), jointly credited Mask Quest (43), Oeuf (89). The joint title is excluded from the sole-credit aggregates but retained here. An established credit does not make every later work widely reviewed. - **Sokpop Collective:** 102 included paid apps, with Stacklands contributing 29,141 of their 37,101 reviews (78.5%). The catalogue also includes many much quieter games. Numerous recorded releases share one January 2020 day, showing why Steam release spacing cannot be interpreted as actual production cadence. - **a327ex:** BYTEPATH has 306 reviews at 92% positive; SNKRX has 4,197 at 94%. Their recorded release gap is approximately 3.2 years. This locates the observed catalog; it does not create a personal probability for the next game. The personal implication is about choosing the appropriate context. An existing body of work is relevant to interpreting a new release, and its largest entry need not become the required scale of every future work. The data support taking the entire catalog seriously while leaving room for smaller experiments. They cannot quantify how much of that room comes from audience loyalty versus enduring skills, resources, taste or other advantages. ## Verification and the next evidence boundary The analysis passed 163 checks after the date correction, including identity, counts, source outcomes, chronological pair accounting, expected-count calibration, 100 sampled earlier-peak reconstructions and frozen-label checks. Both figures were visually inspected. Mechanical consistency is not evidence that every source date or credit describes real-world history perfectly. The King of Fighters XV is a confirmed example: the snapshot date was December 12, 2024, but SNK's official announcement states Steam availability on February 17, 2022: https://www.snk-corp.co.jp/us/press/2022/021701/ . That date was corrected locally and all tables regenerated. No exhaustive date audit was performed. The next useful collection would reconstruct what older games had accumulated *before* a later game launched, on a bounded set of independently verified catalogs. That would begin separating a developer's pre-existing position from the current catalog associations shown here. Even then, measuring actual returning players would require different evidence.