Patreon for speedrunning creators — 2026 edition

Route documentation, RNG manipulation, frame data, glitch mechanics, split analysis, and the Apple Tax.

Speedrunning Patreons retain when they deliver the technical substrate that GDQ runs and clip highlights compress away — the frame data tables, RNG state manipulation sequences, glitch geometry explanations, and split analysis that convert a viewer who watches someone beat a game in 47 minutes into a runner who understands exactly why each second was saved and how to reproduce it.

What speedrunning Patreon creators document

Frame data. At 60fps, one frame is 16.67ms. A frame-perfect input requires consistent execution within a ±16.67ms window — one frame early or late and the game registers a different state. Games that render at 30fps use a doubled frame clock (33.33ms per frame), but inputs are still processed on the 30fps tick, meaning the effective execution window is 33.33ms for 30fps titles. Input lag — the delay from controller signal to game response — typically runs 3–8 frames on modern consoles (50–133ms); characterizing the exact input lag of a specific console, capture card, and display chain is prerequisite work before practicing frame-perfect sequences, because an attempt calibrated for 3 frames of lag will be consistently late on a setup with 5 frames. Tool-assisted superplay (TAS) demonstrates the theoretical optimal by playing frame-by-frame with rewind, establishing the upper bound on what is achievable; human-achievable consistency on a frame-perfect input varies from around 40% success rate for highly practiced runners on difficult setups to near-100% for lenient 2–3 frame windows, and documenting this distribution (not just the input itself) is what separates a useful guide from a clip.

RNG manipulation. Most games use pseudo-random number generators (PRNGs) — deterministic algorithms that produce a sequence of values from a seed. Because the PRNG is deterministic, a runner who knows the current internal state can predict every future RNG output and, critically, can steer the state to a known favorable index through frame-counted inputs, menu cycles, or specific enemy interactions. The documentation format is: starting state (typically the PRNG state at title screen, characterized by the number of frames elapsed before pressing start), the input sequence to reach the favorable index, and the resulting outcome (specific item drop, enemy movement pattern, room layout, or warp destination). RNG state tables are high-value Patreon deliverables because they require significant data collection (recording many runs at known frame counts) and become outdated when a new category route changes the state at the point of manipulation — which means they need maintenance, making ongoing patronage valuable.

Glitch categories. Out-of-bounds (OoB) via geometry clipping: the player character's collision box must be positioned at a precise angle and velocity relative to a collision polygon such that the game's physics step moves the character past the polygon boundary before the collision response fires; the setup typically requires a specific position measured in sub-pixel units (the game stores position as a fixed-point number with fractional bits below the visible pixel grid), a precise input sequence to generate the required velocity vector, and sometimes a specific RNG state that affects the collision geometry. Wrong warp: touching an out-of-bounds tile that is linked in the map's memory to a different room or area destination, effectively transporting the player to a part of the game that normal progression would require hours to reach; the trigger tile is typically invisible in normal play and accessible only via an OoB route. Arbitrary code execution (ACE): precise in-game inputs overflow an in-game variable (inventory count, character name string, dialogue index) whose value is interpreted as executable memory; by controlling the overflow value through specific input sequences, the runner writes arbitrary instructions into the program counter and executes them, allowing direct manipulation of the game state — warping to credits, setting completion flags, or modifying in-game values without the game's intended logic. Sequence breaks via alternate routes: bypassing a required trigger (door key, NPC interaction, flag set) by reaching the trigger's effect through a route the game's collision and logic designers did not anticipate, typically by combining movement glitches with knowledge of the game's event graph.

Split analysis with LiveSplit. Splits are checkpoint timestamps recorded during a run, each marking the end of a defined game segment. LiveSplit software records these in real time and displays comparison data: the current run's split time vs the PB split time (showing whether this run is ahead or behind the personal best at each checkpoint), and the current run's split time vs the gold split (the best time ever achieved for that individual segment across all runs). The sum-of-best (SOB) is the sum of all gold splits — the theoretical minimum a runner could achieve if every segment went as fast as its respective gold; because gold splits from different runs are unlikely to all occur in the same run (a gold early segment often leaves the runner in a suboptimal state for later segments), SOB is an unreachable lower bound, not a target. A run is considered "on pace" when its deficit vs SOB is within the runner's historical save range — if a runner typically saves 15–30 seconds in the final third of a run, then being 20 seconds behind SOB at the two-thirds mark is still on pace for a PB. Timing standard: LLRT (load-removed leaderboard timing) subtracts all measured load screen durations from the wall-clock time, allowing fair comparison across hardware with different disk or cartridge speeds — a PS4 runner and a PC runner with an SSD compete on the same metric. RTA (real-time attack) includes loads in the clock, making hardware choice part of the category meta.

Patreon tier structure for speedrunning creators

Speedrunning Patreon tiers convert when they offer the reference materials and analysis that make a viewer into a runner — route documents with current strats, RNG tables, glitch setup diagrams, and access to the runner's training process. The tier names should reflect speedrunning vocabulary rather than generic membership language.

• Any% ($5–8/mo): monthly route update notes documenting new or revised strategies for the primary category (new glitch discovered, strat updated after community testing, comparison of time save vs execution risk), written strategy documents with glitch explanations and setup diagrams for major time-saves, and PB highlight video breakdowns annotating the key decisions — the splits that went well, the mistakes that were recovered, and what the run needed to be a world record attempt.
• All Flags ($15–18/mo): everything in Any% plus full commentary VOD archives of training and grind sessions with timestamp annotations (so patrons can find the moment a specific strat is practiced or discussed without watching seven hours of grinding), RNG manipulation guides with complete state tables for supported games, and split history analysis comparing the runner's current split distribution across recent runs against gold splits and the SOB target — showing which segments are the primary time-loss sources and what the practice priority should be.
• World Record Attempt ($30–40/mo): live training stream access including unannounced grind sessions so patrons see the unedited practice process rather than curated highlight content, practice ROM or ISO files for supported games where redistribution is legally permitted, and direct coaching on specific route segments via Discord voice or recorded feedback on patron-submitted runs — reviewing a patron's LiveSplit data and VOD to diagnose where time is being lost and what to practice next.

Community Discord access belongs at every tier — the real-time discussion of new strats, route changes, and glitch discoveries is half the value of a speedrunning Patreon. Tier-gated channels (a private strat-discussion channel for All Flags and above, a coaching channel for World Record Attempt patrons) enforce the tier structure without cutting off the core community.

The Apple Tax on speedrunning Patreon revenue

Speedrunning audiences on Twitch and YouTube run 72–88% iOS — the gaming-adjacent demographic that follows GDQ events, watches category highlight compilations, and subscribes to runner Patreons is heavily iPhone-dominant despite the desktop-native nature of the activity itself. English-speaking speedrunning viewership concentrates in North America and Europe where iPhone market share is highest. From November 1, 2026, Apple's 30% IAP commission applies to all Patreon iOS subscription renewals — every patron who subscribed through the Patreon iOS app generates 30% less revenue than a patron who subscribed through the website.

The calculation: Apple cut = monthly revenue × iOS share × 0.30. At $200/mo with 78% iOS: $200 × 0.78 × 0.30 = $46.80/month, $561.60/year. These are dollars that leave the creator's revenue permanently once iOS-billed subscribers renew under the post-November 1 rules.

The fix is web-only billing. Patreon subscriptions completed at patreon.com/[name] in a mobile browser bypass Apple's in-app purchase system entirely — the web checkout routes through Stripe, not the App Store. Adding a direct patreon.com/[name] link in YouTube video descriptions, Twitch panel links, and stream chat commands (with one sentence of context: "subscribe on the website, not the app, to avoid the Apple fee") captures the iOS majority through web billing. For a full implementation guide see web-only Patreon billing and the Apple Tax explainer.

Part of the KeepTier explainer series — receipts-first coverage of the Patreon Apple Tax and what speedrunning, any%, and gaming creators can do about it before November 1, 2026.