Losing users to RNG
I saw this great video clip illustrating binomial distribution and was reminded of a lengthy discussion I had with an engineer colleague while designing a distribution balance for loot drops. I wanted to implement a chance floor mechanic so players would not get too unlucky and he was miffed because he thought it redundant, as "everything regresses to the mean in the long run anyway". It was difficult to convince him and he had to build a simulator before understanding my point.
If you set something to drop at a 20% rate per attempt, the long term average would be 5 attempts per acquisition but there would be plenty of individuals who had to make 10 attempts or more. At the extreme, given thousands of individuals, there would probably be some whom had to make 20 or more attempts to acquire said drop.
The user population is unlikely to feel this when an "attempt" is very short and frequent, like a 2-3 second kill on an enemy mob with hundreds of instances in a level. However, imagine this drop is mapped against the completion of a level which would take a ~10min run. The discrepancy of the average 5 attempts (50 mins) and an extreme of 20 attempts (200 mins) can break the entire user experience and cause churn. It is worse for new users, whom have not developed enough of an attachment or history to tolerate a sequence of "unlucky events".
If a game has a d0 and d1 experience with extremely varied progression outcomes due to RNG, then the question starts to become -- how much of your installs will churn because of RNG? I don't think any PM should be OK with 10% of new players being high-risk for churn because they have to spend 4x the mean number of attempts to acquire shiny good object X shortly after the FTUE.
Some thoughts:
1. Serve the users with a tighter, lower-variance experience especially in early game
2. Understand where and how RNG is used in every feature and system of your game
3. Low-frequency and/or long-session attempts need a lot more care in RNG design
4. Using a floor/ceiling rubberbanding mechanic may help
5. Simulate outcomes over thousands of players to see what the distribution curve looks like
6. Don't use RNG in systems and experiences that don't benefit from it
One parting thought: the most interesting mechanic that counters RNG is Trading. This is a controversial and sometimes taboo topic for some F2P companies because trading creates a secondary player market that can negatively impact the primary market (storefront) which the developers own, govern and monetize. But it's also a marketplace where unlucky sequences can be neutralized through trade, resulting in a normalization of weird or unlucky individual roll histories.
If you set something to drop at a 20% rate per attempt, the long term average would be 5 attempts per acquisition but there would be plenty of individuals who had to make 10 attempts or more. At the extreme, given thousands of individuals, there would probably be some whom had to make 20 or more attempts to acquire said drop.
The user population is unlikely to feel this when an "attempt" is very short and frequent, like a 2-3 second kill on an enemy mob with hundreds of instances in a level. However, imagine this drop is mapped against the completion of a level which would take a ~10min run. The discrepancy of the average 5 attempts (50 mins) and an extreme of 20 attempts (200 mins) can break the entire user experience and cause churn. It is worse for new users, whom have not developed enough of an attachment or history to tolerate a sequence of "unlucky events".
If a game has a d0 and d1 experience with extremely varied progression outcomes due to RNG, then the question starts to become -- how much of your installs will churn because of RNG? I don't think any PM should be OK with 10% of new players being high-risk for churn because they have to spend 4x the mean number of attempts to acquire shiny good object X shortly after the FTUE.
Some thoughts:
1. Serve the users with a tighter, lower-variance experience especially in early game
2. Understand where and how RNG is used in every feature and system of your game
3. Low-frequency and/or long-session attempts need a lot more care in RNG design
4. Using a floor/ceiling rubberbanding mechanic may help
5. Simulate outcomes over thousands of players to see what the distribution curve looks like
6. Don't use RNG in systems and experiences that don't benefit from it
One parting thought: the most interesting mechanic that counters RNG is Trading. This is a controversial and sometimes taboo topic for some F2P companies because trading creates a secondary player market that can negatively impact the primary market (storefront) which the developers own, govern and monetize. But it's also a marketplace where unlucky sequences can be neutralized through trade, resulting in a normalization of weird or unlucky individual roll histories.
Comments
Post a Comment