We measured the gap between reachable checkpoints in every game we have played this year, then deliberately quit at the worst possible moment to see what survived. The median loss was eleven minutes. The worst was forty-one, in a game with no manual save at all and no checkpoint inside a long traversal sequence.
This survey exists because save systems are discussed almost entirely in anecdote. Everybody has a story about losing an hour to a crash, and nobody has a number. We wanted a figure that could be compared between games, so we defined one and then went looking for it forty-four times.
The measurement is simple. For each game we identify the longest stretch during normal play where no checkpoint is reachable and no manual save is permitted, then we force a quit inside it and record what the game gives back. The procedure is described in full on the standards page.
| Worst-case loss | Games | Share | Typical genre |
|---|---|---|---|
| Under 3 minutes | 9 | 20% | Roguelikes, platformers |
| 3–8 minutes | 12 | 27% | Action, adventure |
| 9–15 minutes | 11 | 25% | Narrative, adventure |
| 16–25 minutes | 8 | 18% | Survival, strategy |
| Over 25 minutes | 4 | 9% | Survival, one narrative game |
The median sits at eleven minutes, which we find higher than it should be given how cheap saving has become. Twelve of the forty-four games would lose a quarter of an hour or more, and every one of those twelve is a game people play in the evening, in short sessions, on machines they need to shut down.
The strongest predictor of a low loss figure is not autosave frequency but whether the player is permitted to save deliberately. Games with unrestricted manual saving averaged a four-minute worst case. Games with autosave only averaged sixteen.
| Save model | Games | Median worst-case loss | Notes |
|---|---|---|---|
| Unrestricted manual save | 10 | 4 minutes | Loss is nearly always the player's choice |
| Manual save at rest points | 13 | 9 minutes | Depends entirely on rest point spacing |
| Autosave only, frequent | 14 | 14 minutes | Frequency matters less than completeness |
| Autosave only, chapter-based | 5 | 26 minutes | Worst outcomes in the survey |
| Run-based, no mid-run save | 2 | Not applicable | Loss is the design, and it is stated |
Nine of the forty-four games save some of your progress and not all of it, and this produced our angriest notes. A game that autosaves your position but not your inventory teaches you to trust it and then takes something anyway.
Our clearest example is reviewed in full elsewhere: an autosave every ninety seconds, twenty-six saves during the session, and forty minutes of gathered materials gone because position was recorded and inventory was not. Compare a game with no autosave at all, where a player knows to save manually and does.
| What was preserved | What was lost | Games affected |
|---|---|---|
| Position only | Inventory, task progress | 4 |
| World state only | Current construction, held items | 3 |
| Chapter progress only | Collected items, side objectives | 2 |
A save that keeps half your work is worse than no save, because it persuades you to stop being careful.
Seven survival games got their own treatment because the genre's whole appeal is accumulated work, which makes an unexpected quit unusually expensive. We closed the window without warning during construction, during a resource run and during a day transition.
| Quit moment | Games handling it cleanly | Games losing the task | Games losing a day cycle |
|---|---|---|---|
| Mid-construction | 3 | 2 | 2 |
| Mid-resource run | 4 | 3 | 0 |
| During day transition | 2 | 1 | 4 |
Day transitions are the weak point across the genre. Four of the seven games lost an entire cycle of work when quit at the wrong moment during one, which suggests the transition is treated as an atomic operation that cannot be interrupted. That is understandable technically and unacceptable practically.
Nine roguelikes were audited, and the genre performs better than anything else here for a structural reason: when losing progress is the design, a game has to state exactly what persists. Every one of the nine did so clearly, in its own interface, without our having to test for it.
Compare twelve narrative games in the same survey, of which four gave us no indication at all of what an autosave had captured. The genre that makes loss its subject is the genre most honest about it, which we think says something uncomfortable about the rest.
Three things, drawn from the games that scored best. First, permit a deliberate save at any point, and if the design forbids it, say why in the interface. Second, save the whole state or clearly mark what is provisional. Third, when a quit will cost something, warn the player with a number: "you will lose about six minutes" is more use than a generic confirmation box.
None of these are expensive. The best-performing game in this survey does all three, and its worst-case loss was under two minutes across every test we ran.
