There is advice every Steam Workshop creator gets once they have a mod that works: do it again, in a game that is growing. There is other advice they get once a game dies: take your audience with you.
Both sentences assume the same thing — that whatever made the first mod work is a property of the creator, and that it travels. Nobody has checked whether it does.
So we checked, on the whole platform. This is a study of 1,898,060 creator-and-game pairs built from 6,005,516 active Steam Workshop mods carrying at least one subscription, published by 1,692,288 distinct creators across the live catalogue ModSignal tracks. For every creator who appears in more than one game, we ordered their games by when they first published in each, and asked a single question: does what happened in game #1 predict what happens in game #2?
The 9.35% who move
Start with how rare the situation is. Of 1,692,288 creators with at least one active, subscribed mod:
| Games published in | Creators | Share |
|---|---|---|
| 1 game | 1,534,091 | 90.65% |
| 2 games | 125,510 | 7.42% |
| 3 games | 23,453 | 1.39% |
| 4 games | 6,007 | 0.36% |
| 5–9 games | 3,165 | 0.19% |
| 10+ games | 62 | 0.00% |
| 2 or more | 158,197 | 9.35% |
158,197 creators — 9.35% — publish in more than one game. The other nine in ten never leave their first ecosystem.
That minority nevertheless accounts for 764,448 active mods — 10.38% of the entire 7.34-million-item live catalogue — holding 1.88 billion subscriptions, or 15.11% of every subscription on the platform. More than one mod in ten, and more than one install in seven, belongs to a creator who has crossed games.
The question is whether crossing helped them.
The split is whether you banked a hit
Take every creator's second game — 205,772 creator-game pairs, where the second game is the chronologically second ecosystem they published into. Then split them by one fact: had this creator, in any of their earlier games, already landed a mod that reached 1,000 subscribers?
| Prior game produced a 1k+ hit | Creators | Mods in the new game | Reached 1k | Rate | Avg best mod | Median best mod |
|---|---|---|---|---|---|---|
| Yes | 44,964 | 208,595 | 62,056 | 29.75% | 10,709 | 288 |
| No | 160,808 | 555,853 | 59,831 | 10.76% | 1,960 | 24 |
That is a 2.77x difference in the odds that any given mod in the new game reaches 1,000 subscribers — and a 3.18x difference in the odds that the new game produces at least one such mod at all (33.77% against 10.63%).
Look at the median. A creator who arrives with a banked hit has a median best mod of 288 subscriptions in the new game. A creator who doesn't has a median best of 24. Both are modest — which is the honest framing. The typical cross-game move does not produce a hit; it produces a near-zero. The difference is that the creators carrying a hit have a real chance of landing one, and the others mostly do not.
The ladder is monotonic in how much of a hit they banked:
| Prior games with a 1k+ hit | Creators | Mods | Rate |
|---|---|---|---|
| 0 (all flops) | 160,808 | 555,853 | 10.76% |
| 1 | 38,780 | 178,247 | 29.56% |
| 2 | 4,730 | 23,188 | 30.54% |
| 3 | 984 | 4,859 | 29.57% |
| 4 or more | 470 | 2,301 | 37.16% |
One banked hit does almost all the work. Two, three and four add a little; the jump from zero to one is the entire story, which is a pattern this series has now seen three times — in the tag cliff, in the short-title problem, and in the debut signal.
It is not the game
The obvious objection is that creators with a prior hit simply pick better second games — that they move into ecosystems with higher base rates, and the whole table is a story about target choice. So we controlled for it.
First, inside the target game. For the twelve most common destinations, here is the same split measured only among creators arriving at that game:
| Target game | No prior hit | Prior hit | Multiple |
|---|---|---|---|
| Garry's Mod | 41.60% | 55.09% | 1.32x |
| Left 4 Dead 2 | 19.54% | 39.57% | 2.02x |
| Planet Zoo | 11.80% | 31.03% | 2.63x |
| Brick Rigs | 8.50% | 33.91% | 3.99x |
| Stormworks | 8.52% | 26.17% | 3.07x |
| Space Engineers | 2.54% | 13.67% | 5.38x |
| Arma 3 | 4.17% | 12.33% | 2.96x |
| Trailmakers | 1.09% | 8.80% | 8.07x |
| Scrap Mechanic | 2.84% | 5.66% | 1.99x |
| Barotrauma | 4.03% | 13.25% | 3.29x |
| Portal 2 | 0.78% | 3.62% | 4.64x |
| Kerbal Space Program | 0.70% | 1.97% | 2.81x |
Every row moves the same way, including the ones where the target game is easy (Garry's Mod, where 41.6% of everyone hits) and the ones where it is nearly impossible (Kerbal Space Program, where the platform-wide rate is 1.08%). The prior hit is not a proxy for picking a good game, because within a single game it still separates the arrivals by 2x to 8x.
Second, inside the creator's own catalogue size.
It is not the catalogue
The next objection: creators with a prior hit are simply bigger publishers, and size is the real variable. So hold the first game's catalogue size fixed and vary only whether it produced a hit:
| Mods in first game | No prior hit | Prior hit | Creators (no / yes) |
|---|---|---|---|
| 1 mod | 12.56% | 29.76% | 69,587 / 7,208 |
| 2–4 | 10.80% | 33.95% | 40,398 / 8,613 |
| 5–19 | 9.60% | 31.94% | 17,772 / 8,193 |
| 20–99 | 6.43% | 33.97% | 2,910 / 2,905 |
| 100+ | 33.47% | 31.20% | 177 / 434 |
The 20–99 band is the cleanest test in the whole study: 2,910 creators without a prior hit and 2,905 with one — near-perfectly matched samples — and the rates are 6.43% against 33.97%. Two groups of the same size, publishing the same volume in their first game, differing by 5.3x on a single fact about whether the first game produced a hit.
The 100+ band is the one place the effect vanishes (33.47% vs 31.20%), and it is worth saying why: at that scale the "no hit" group is 177 creators who published a hundred-plus mods without ever reaching 1,000 subscribers. That is not a control group, it is a set of near-misses, and their second game performs like a hit creator's. Scale substitutes for a hit — the same substitution the portfolio effect found in the opposite direction.
One more cut, the tightest in the study. Restrict to creator-game pairs where the creator published exactly one mod in the new game — no catalogue to hide behind:
| Prior hit | Creators | Reached 1k | Rate |
|---|---|---|---|
| No | 74,587 | 4,611 | 6.18% |
| Yes | 13,964 | 3,261 | 23.35% |
A creator with a banked hit who ships a single mod into a new game succeeds 23.35% of the time. A creator without one, shipping a single mod, succeeds 6.18% of the time. Neither has a second attempt in that game. The gap is 3.78x on a sample of 88,551.
It is the size of the old hit
Here is the finding that makes this more than a binary. Break the prior result into five magnitude classes instead of asking yes/no:
| Best mod in an earlier game | Creators | Mods in new game | Rate |
|---|---|---|---|
| 1–999 subs (no hit) | 160,808 | 555,853 | 10.76% |
| 1,000–9,999 | 31,103 | 134,946 | 24.56% |
| 10,000–99,999 | 11,730 | 60,235 | 36.86% |
| 100,000–999,999 | 1,989 | 12,673 | 49.22% |
| 1,000,000+ | 142 | 741 | 64.64% |
Read the rate column alone: 11%, 25%, 37%, 49%, 65%. It is close to linear in the log of the old hit's size, and it does not saturate at the top — the 142 creators who arrived with a million-subscriber mod behind them hit 1,000 again 64.64% of the time.
That is the shape of a transferable quantity. If a prior hit were merely a proxy for "this creator is skilled," the ladder would flatten after the first step, the way the debut signal flattened at 10,000 subscribers. It doesn't flatten. The bigger the thing that travelled, the more of it arrives.
The corollary is uncomfortable for anyone reading this as advice: creating a small hit in game #1 is worth roughly a quarter of what a large hit in game #1 is worth, in game #2. A single 800-subscriber mod — the outcome most creators actually achieve — contributes nothing measurable to the next game.
And it is not the clock
The last mechanical explanation: creators who have a hit are older accounts, so they have more time to accumulate anything, including a second-game hit. We tested it by splitting every second-game arrival by the time gap between when the creator started in game #1 and when they started in game #2:
| Gap between first and second game | No prior hit | Prior hit | Multiple | Creators (no / yes) |
|---|---|---|---|---|
| Under 90 days | 7.98% | 37.56% | 4.71x | 17,218 / 2,960 |
| 90–364 days | 10.40% | 34.52% | 3.32x | 33,108 / 5,847 |
| 1–2 years | 12.23% | 33.02% | 2.70x | 28,043 / 5,552 |
| Over 2 years | 12.95% | 29.25% | 2.26x | 52,485 / 12,942 |
Time in the platform is a resource both groups spend equally, and the separation persists at every gap. It narrows with age — 4.71x for creators who move quickly against 2.26x for veterans — but it never closes, and it never reverses.
Read the two columns separately, because they say different things. The prior-hit column barely moves: 37.56%, 34.52%, 33.02%, 29.25% — a slow decay that is really just the whole platform's difficulty gradient over time. The no-prior-hit column climbs steadily: 7.98%, 10.40%, 12.23%, 12.95%. Creators who took years to make a second move and hadn't banked a hit are measurably better at it than creators who moved fast without one. The fast movers with no hit — a mod, a flop, and off to the next game within 90 days — are the worst-performing group in the entire study at 7.98%.
That is a fact about impatience, and it is the one place in this study where a decision at the publishing screen does seem to matter.
The paired test: moving is not the lever
Everything above compares creators to each other. The stronger test is to compare each creator to themselves — their first game against all their later games:
| Mods | Reached 1k | Rate | |
|---|---|---|---|
| First game | 740,906 | 125,489 | 16.94% |
| All later games | 725,186 | 118,844 | 16.39% |
Across 118,935 creators with at least three total mods in two or more games, the first game and the later games are indistinguishable — a 0.55-point gap on 1.47 million mods. The per-creator distribution is the same story: mean difference +0.0066, median difference exactly 0.0000, with 81,665 of 118,935 creators perfectly tied and the remaining 37,270 split almost evenly (18,439 first-better, 18,831 later-better).
Moving games is not a penalty and not a bonus. There is no first-game advantage to preserve, and no diversification discount to pay. Whatever the difference between the two groups in the table above is, it lives entirely between creators — in whether they banked a hit — and not in the act of moving.
That distinction matters, because it removes the two readings most people would reach for. "Your first game is your best" is false. "Going wide dilutes you" is false. What is true is narrower and more useful: the thing that changes your second game is the first game's outcome, not the decision to have a second game.
The peak usually stays home
One number to calibrate expectations. For every creator in multiple games, which game is their single biggest — the one holding the most subscriptions?
| Games published in | Creators | Peak game is game #1 | If it were random |
|---|---|---|---|
| 2 | 125,510 | 55.16% | 50.00% |
| 3 | 23,453 | 38.19% | 33.33% |
| 4 | 6,007 | 29.82% | 25.00% |
| 5 | 1,952 | 24.74% | 20.00% |
The first game is the peak more often than chance at every size — but only modestly. Even at two games it is a coin flip weighted 55/45. And across all 158,197 multi-game creators, 49.03% of their cumulative subscriptions sit in the first game. Half. The other half arrived afterwards.
So the honest reading of the whole study, assembled: cross-game creators are rare; their later games are not worse than their first; the first game is not a fortress; and if you want to know whether the next one will work, the only field on the record that tells you is whether the last one did — and how big it was.
The corridors, and who actually did it
The transitions are not random, and they reveal the shape of the platform's creator flows. All 15 routes travelled by at least 400 creators — first game to second game — with the share of arrivals who land a hit:
| From | To | Creators | Landed a 1k+ hit |
|---|---|---|---|
| Planet Coaster | Planet Zoo | 1,275 | 34.90% |
| Scrap Mechanic | Trailmakers | 1,031 | 2.81% |
| Portal 2 | Garry's Mod | 805 | 45.47% |
| Euro Truck Simulator 2 | American Truck Simulator | 696 | 56.03% |
| Garry's Mod | Left 4 Dead 2 | 684 | 44.74% |
| Empyrion | Space Engineers | 647 | 4.02% |
| Besiege | Kerbal Space Program | 633 | 2.37% |
| Kerbal Space Program | Stormworks | 629 | 11.76% |
| Scrap Mechanic | Stormworks | 586 | 9.39% |
| Space Engineers | Space Engineers 2 | 581 | 7.23% |
| Left 4 Dead 2 | Garry's Mod | 563 | 62.70% |
| Besiege | Trailmakers | 497 | 5.43% |
| Stronghold: Definitive Edition | Stronghold Crusader: Definitive Edition | 488 | 1.23% |
| Total War: WARHAMMER II | Total War: WARHAMMER III | 487 | 56.67% |
| Planet Zoo | Planet Coaster | 417 | 4.08% |
The corridors with the highest hit rates are the ones where the destination is the same audience — Planet Coaster→Planet Zoo (34.90%), ETS2→ATS (56.03%), L4D2→Garry's Mod and back (44.74% and 62.70%), Warhammer II→III (56.67%). The corridors with the lowest are the engineering-sandbox web — Scrap Mechanic→Trailmakers at 2.81%, Besiege→Kerbal at 2.37%, Besiege→Trailmakers at 5.43%, Stronghold DE→Crusader DE at 1.23% — where creators move between games whose audiences overlap barely and whose base rates are under 3%.
That led to one more test. If audience continuity is the mechanism, then moving within a franchise should be worth something on its own. We flagged every second game whose name shares a ten-character prefix with any earlier game the creator had worked in:
| Creators | Mods | Rate | |
|---|---|---|---|
| Same franchise | 4,454 | 18,369 | 27.91% |
| Different franchise | 153,743 | 565,852 | 15.73% |
And then crossed it with the prior hit — the decisive table of the study:
| Same franchise | Prior hit | Creators | Mods | Rate |
|---|---|---|---|---|
| No | No | 128,054 | 443,994 | 11.48% |
| No | Yes | 25,689 | 121,858 | 31.22% |
| Yes | No | 2,790 | 9,783 | 9.94% |
| Yes | Yes | 1,664 | 8,586 | 48.39% |
Read the top two rows against the bottom two. Staying inside a franchise without a banked hit scores 9.94% — lower than leaving it with no hit (11.48%). Shipping a sequel in the game's own series buys you nothing if the first one flopped. But a prior hit inside a franchise reaches 48.39%, nearly double the rate of a prior hit outside one (31.22%).
Neither lever works alone, and one of them is worthless alone. A franchise move is only worth anything as an amplifier of an audience you already proved you could hold.
The people behind the split
The names in the data are the argument. Here are the largest cross-game hits that came after a hit in a different game:
| Creator | First game (best mod) | Second game (best mod) |
|---|---|---|
| Mr. uLLeticaL™-S | Garry's Mod — 59,236 | Counter-Strike 2 — 24,361,342 |
| rezecib | Don't Starve — 1,886,896 | Don't Starve Together — 10,342,279 |
| squeek. | Don't Starve — 1,501,830 | Don't Starve Together — 8,349,425 |
| Yesber | Portal 2 — 13,021 | Counter-Strike 2 — 5,724,469 |
| IvanX | Don't Starve — 720,095 | Don't Starve Together — 5,407,005 |
| yolokas | Garry's Mod — 178,584 | Counter-Strike 2 — 5,317,619 |
| Skolas | Wallpaper Engine — 2,236 | MECCHA CHAMELEON — 5,309,453 |
| Dustino | Garry's Mod — 5,298 | Dota 2 — 3,773,775 |
| RareKiwi | ARK: Survival Evolved — 36,676 | MECCHA CHAMELEON — 2,996,757 |
| IAmChris_ | House Flipper — 3,376 | MECCHA CHAMELEON — 2,862,313 |
Now the other half — creators whose first game produced nothing and whose second produced a giant:
| Creator | First game (best mod) | Second game (best mod) |
|---|---|---|
| (anonymous) | Tabletop Simulator — 479 | MECCHA CHAMELEON — 6,709,290 |
| kowalski7cc | Portal 2 — 17 | Garry's Mod — 6,185,812 |
| Empyrean | KovaaK's — 218 | Aimlabs — 4,278,811 |
| computerkiwi | Portal 2 — 8 | Dota 2 — 4,030,082 |
| (anonymous) | KovaaK's — 57 | Aimlabs — 3,487,122 |
| (anonymous) | Arma 3 — 162 | DayZ — 3,438,059 |
| BRITA | This War of Mine — 23 | Project Zomboid — 3,324,375 |
| ChthonicKnight | Tabletop Simulator — 154 | Project Zomboid — 3,314,076 |
| bigwig | Portal 2 — 151 | Garry's Mod — 3,253,527 |
| zac | KovaaK's — 117 | Aimlabs — 2,192,997 |
Both lists are real, and the second is the reason this article is not a rule. KovaaK's→Aimlabs appears five times in the lower table: a cohort of creators who got nothing out of an aiming trainer and then owned an aiming trainer. Portal 2→Garry's Mod appears three times. The flop-to-giant path exists and is not rare — 160,808 creator-game pairs sit in the no-prior-hit group, and they produced 59,831 hits between them.
The correct claim is not "a flop condemns you." It is that across 205,772 cross-game moves, the observed rate is 29.75% for creators with a banked hit and 10.76% for creators without — and that this is the strongest single predictor available in the data, not a guarantee and not a ceiling.
Why this is hard to disentangle
We can show the association survives controls for target game, catalogue size, publish year, time gap, burst structure and franchise membership. We cannot tell you which of three mechanisms is doing the work, and they predict the same numbers:
- Skill persistence. Creators who made one mod work have abilities — thumbnail sense, title craft, update discipline, community handling — that apply in any game. Our own earlier work found each of those individually worth 2x to 3x, which compounds to roughly this size.
- Audience continuity. Some fraction of the people who installed the first mod follow the creator's profile and install the next one wherever it lands. The franchise table is direct evidence for this: hit-within-franchise (48.39%) nearly doubles hit-within-nothing (31.22%).
- Platform treatment. A creator with a proven track record may be surfaced differently — followed, featured, recommended to friends-of-subscribers. Steam's own systems are not visible to us.
The franchise result argues audience continuity is real. The catalogue-size result, where the lift holds even at 20–99 mods, argues skill is real. We cannot size them against each other from this data.
A note on what "later game" means here
Two definitional limits, stated plainly.
Games are ordered by first publication, not by importance. A creator's "first game" is whichever ecosystem they published into earliest in our records — and the live catalogue only reaches back to what was active as of 2026-08-21, when ModSignal began continuous snapshots. A creator who worked in three games from 2015 to 2019 and is active today has only their surviving mods counted, and the ordering can differ from their true history.
1,000 subscribers is a low bar, deliberately. It is the threshold the rest of this series uses because it is the first rung where a mod is unambiguously "in use" rather than a curiosity. Rates at 10,000 subscribers run roughly a quarter as high and rank the same groups the same way, so the split is not an artefact of the threshold.
The integrity section
Four tests the headline has to survive, reported as they came out.
The paired within-creator test — first game 16.94% against later games 16.39% on 1.47 million mods (median per-creator difference 0.0000). This is the test that would have killed the article's framing if it had gone the other way. It did not: moving is neutral, so the between-creator split in the main table is about the creator, not about the move.
The matched-catalogue test — at 20–99 mods in the first game, 2,910 no-hit creators against 2,905 hit creators, 6.43% against 33.97%. Equal-sized samples, equal publishing volume, 5.3x apart.
The per-target-game test — the lift holds in all twelve major destination games, from Garry's Mod (1.32x) to Trailmakers (8.07x).
The ordering sanity check — we also tried to split second-game arrivals by whether they had already hit in the destination game itself, and the group is structurally empty: all 158,155 second-game arrivals fall into "no prior hit here, maybe a hit elsewhere" (27,301) or "no hit anywhere" (130,854). That is the expected result and it confirms the ordering is sound — a creator's "first game" is by definition the earliest ecosystem in the records, so no one can have a result in their second game before their first. The finding is not an artefact of mis-sequenced history.
The time-gap test — the separation persists at every interval between first and second game: 4.71x under 90 days, 3.32x at 90–364 days, 2.70x at 1–2 years, 2.26x past two years. It is not an accumulation artefact of account age, and it is not concentrated at one horizon.
And the one that didn't cleanly — the 100+ mods band, where the two groups converge (33.47% vs 31.20%). We have reported it rather than dropped it. A 177-creator control group that published more than a hundred mods without a single hit is a near-miss cohort, and it behaves like the hit group in its next game. Scale is a substitute for a banked hit, exactly as the portfolio effect found when it watched the breadth premium evaporate past 1,000 items.
For the creators behind these numbers — the most influential creators and the breakout factories the rest of this series covers — the practical content of this study is a single sentence. Your next game will perform like your last one, not like your next idea. And unlike most of the levers this series has tested, that is one you cannot decide at the publishing screen.
Data: ModSignal's live Steam Workshop dataset — 7,343,084 active items, 6,005,516 of them carrying at least one subscription, published by 1,692,288 creators, snapshotted continuously since 2026-08-21. 1,898,060 creator-game pairs; 205,772 second-game arrivals; 118,935 creators in the paired first-vs-later test. Per-game ecosystems are browsable at ModSignal.