Crash-free rate calculator
Crash-free rate is the share of sessions, or of players, with no crash. Crash-free sessions = 1 minus crashed sessions divided by all sessions; crash-free players use players instead. Enter your counts to get both rates, why they differ, and how many crashes you must fix to reach a target such as 99.5%.
Crash-free sessions are 3.75 points higher than crash-free players. A player counts as crashed after one crash in any of their sessions, and each player had 8 sessions on average, so the player rate is lower. Each player who crashed had 1.33 crashed sessions on average, so some players crash again and again: look for a crash tied to one device, save file or level.
Against published benchmarks
- Below Bugnet (game bug and crash reporting), target 99.5% crash-free sessions: "Percentage of sessions without a crash. Target: 99.5%+." Games, any platform. A target in a vendor's documentation, not a store rule. Bugnet does not say over which period. Source
- Below Google Play Android vitals (Android apps, games included), bad-behavior threshold 98.91% crash-free players: "Overall bad behavior threshold: At least 1.09% of daily active users experience a user-perceived crash, across all device models." Mobile (Android). Counted per day: the share of daily active users with a crash while using the app. Above it, Google says the app is likely to be less discoverable. Compare only with a one-day player number. Source
- At or above Google Play Android vitals (Android apps, games included), bad-behavior threshold 92% crash-free players: "Per-device bad behavior threshold: At least 8% of daily active users experience a user-perceived crash, for a single device model." Mobile (Android). Per day and per phone model. Compare with a one-day player number for one device row. Source
| Row | Sessions | Crashed sessions | Crash-free sessions | Crash-free players | Crashed sessions to remove |
|---|---|---|---|---|---|
| Windows | 30,000 | 150 | 99.5% | n/a | 0 |
| Steam Deck (worst) | 10,000 | 170 | 98.3% (below target) | n/a | 120 |
| macOS | 8,000 | 40 | 99.5% | n/a | 0 |
| All rows | 48,000 | 360 | 99.25% | Not summed: one player can play several rows |
Crash-free sessions = 1 - crashed sessions ÷ sessions. Crash-free players = 1 - players with a crash ÷ players. Crashes allowed = the total × (100 - target) ÷ 100, rounded down. Crashes to remove = crashed - crashes allowed.
How to use it
- Pick one date range and export, for that range, the number of game sessions and the sessions that ended in a crash. Add players and players with a crash if you have them.
- Enter the counts. Leave a pair empty if you only have sessions or only players.
- Set the target crash-free rate. It starts at 99.5%, the target in Bugnet's documentation.
- Optional: paste one line per build or platform as "name, sessions, crashed sessions", with players and crashed players at the end if you have them.
- Read both rates, the gap between them, the crashes to remove and the rows flagged as worst or below target.
How it works
The calculator uses the same formulas as the Game health board in Manalens, and as Firebase Crashlytics and Sentry. A session counts once if it had at least one crash. A player counts once if any of their sessions crashed, so a player who crashes three times is still one crashed player.
In Manalens a crash is an app_unclean_exit event, sent at the next launch when the last run did not exit normally while the game had focus, or an error with severity critical and handled false. A player is one install. Only game events count, not server events.
Crashes to remove keeps the total fixed and asks how many crashed sessions, or crashed players, must go. The most crashes allowed is the total times (100 minus the target) divided by 100, rounded down. Fixing a crash that hits one player in many sessions removes many crashed sessions but only one crashed player.
Benchmarks are quoted from their sources: Google Play counts the share of daily active users with a crash, so 1.09% is 98.91% crash-free players in one day; Bugnet gives a target of 99.5% crash-free sessions.
crash-free sessions = 1 - crashed sessions / all sessions
crash-free players = 1 - players with a crash / all players
crashes allowed = floor(total × (100 - target %) / 100)
crashes to remove = max(0, crashed - crashes allowed)
gap = crash-free sessions % - crash-free players %Worked example
Example with made-up inputs: one week of a PC game on three platforms
The game had 48,000 sessions in the week and 360 ended in a crash. Crash-free sessions = 1 - 360 / 48,000 = 99.25%. It had 6,000 players and 270 of them crashed at least once. Crash-free players = 1 - 270 / 6,000 = 95.5%.
The gap is 3.75 percentage points. Each player had 8 sessions, and each player who crashed had 1.33 crashed sessions on average, so one crash in any session makes the whole player count as crashed.
For 99.5% crash-free sessions, at most 240 of 48,000 sessions may crash (48,000 × 0.5%), so 120 crashed sessions must go. For 99.5% crash-free players, at most 30 players may crash, so 240 must go.
By platform: Windows 150 of 30,000 (99.5%), Steam Deck 170 of 10,000 (98.3%), macOS 40 of 8,000 (99.5%). Steam Deck is the worst row and holds all 120 crashed sessions to remove; fixing its crash alone reaches the target.
How to read the result
Read both rates for the same date range. Crash-free sessions show how often the game crashes; crash-free players show how many people had a crash. The player rate is usually lower and falls as the range gets longer, because each player has more sessions. Compare builds over ranges of the same length.
A benchmark only fits a number measured the same way. Google Play's thresholds (1.09% of daily users overall, 8% for one phone model) are for Android apps and count one day of users. Bugnet's 99.5% is a target for sessions. We found no PC or console store that publishes a crash threshold.
Start with the worst row: one build or platform often holds most of the crashes. A high number of crashed sessions per crashed player points to a crash that repeats for the same players, such as one device, save or level.
Crash-free benchmarks and where they come from
Google Play Android vitals (Android apps, games included): "Overall bad behavior threshold: At least 1.09% of daily active users experience a user-perceived crash, across all device models." That is 98.91% crash-free players. Mobile (Android). Counted per day: the share of daily active users with a crash while using the app. Above it, Google says the app is likely to be less discoverable. Compare only with a one-day player number. Google Play Android vitals (Android apps, games included): "Per-device bad behavior threshold: At least 8% of daily active users experience a user-perceived crash, for a single device model." That is 92% crash-free players. Mobile (Android). Per day and per phone model. Compare with a one-day player number for one device row. Bugnet (game bug and crash reporting): "Percentage of sessions without a crash. Target: 99.5%+." That is 99.5% crash-free sessions. Games, any platform. A target in a vendor's documentation, not a store rule. Bugnet does not say over which period.
How to see crash-free rate in Manalens
Add the Godot or Unity SDK, or send error and app_unclean_exit events from any engine to the events endpoint. Open Dashboards, pick the Game health template and create the board. It shows crash-free sessions, crash-free players, errored sessions, errors per session by build and the top errors. The cards are SQL charts: to split crash-free rate by build or platform, edit the query, because the top bar filters do not apply to SQL charts. Manalens does not store crash dumps or symbols, so keep a crash reporter for stack traces.
Limits
It counts what you enter. It does not know which crash reporter or SDK produced the counts, and tools differ on what a crash is, for example whether a hang or a crash in the background counts.
Players cannot be added up across rows, because one player can play several builds or platforms, so the rows show a session total only.
Crashes to remove assumes the totals stay the same. It does not say which bug to fix; a crash reporter with stack traces does that.
Frequently asked questions
What is a good crash-free rate for a game?
No PC or console store publishes one that we found. Bugnet's documentation gives a target of 99.5% crash-free sessions. For Android, Google Play treats 1.09% of daily users with a crash as bad behavior, which is 98.91% crash-free users in a day. Use these as reference points and compare your own builds with each other.
Why is crash-free players lower than crash-free sessions?
A player counts as crashed after one crash in any of their sessions. A player with ten sessions and one crash is one crashed player but only one crashed session in ten. The more sessions each player has, the wider the gap. The calculator shows the sessions per player and the crashed sessions per crashed player.
Should I track crash-free sessions or crash-free players?
Both. Firebase says crash-free sessions suit apps with short, intense use such as real-time games, and crash-free users suit long, casual use where players continue later. Sessions tell you how often the build crashes. Players tell you how many people had a bad time.
Can I compare crash-free rates across date ranges?
No. Firebase says the crash-free users value should not be compared over different time periods, because each player has more chances to crash in a longer range. Compare builds or platforms over ranges of the same length, such as the first seven days after each release, and use the same definition of a crash.
Does Manalens show crash-free rate?
Yes. The Game health dashboard template shows crash-free sessions and crash-free players, with the same formulas as this calculator. A crash is an app_unclean_exit event or an unhandled critical error, sent by the Godot and Unity SDKs or by your own code. It does not store crash dumps or symbols, so keep a crash reporter next to it.
Sources
- Firebase Crashlytics: Understand crash-free metrics (Google), read
- Sentry documentation: Release Health, read
- Backtrace for Unity documentation: Stability Metrics Configuration (Sauce Labs), read
- Android Developers: Crashes in Android vitals (Google), read
- Bugnet documentation: Analytics, read
Last reviewed . The numbers are worked out in your browser. Nothing you enter is sent to our server.
Next step
The Game health dashboard in Manalens shows crash-free sessions and crash-free players from your game events, next to retention and the campaign that brought each player. See custom dashboards →