The first version of BangOn had a scoring system where lower was better.
I know.
The logic made sense at the time, in a "this makes sense if you don't think about it" kind of way. You're predicting a score. The further you are from the actual score, the worse you've done. So your score is your total error. Zero is perfect. Higher is worse.
This is mathematically coherent and completely wrong for a game. It's not golf, though that comment made me think of something :-) This is exactly where your mates come in, just right when you were charging confidently down the wrong road.
WHY IT DIDN'T WORK
The problem became obvious after a few people played it.
They predicted 1-0. The actual score was 3-2. Their score went up by 4. They thought they'd done well. They hadn't.
The game was telling people they were doing better when they were doing worse. The leaderboard ranked people in the wrong direction. Explaining it required a paragraph of text and a reference to golf. Any game that requires a paragraph of text to explain what a good score looks like has already lost.
The deeper problem was psychological. Watching your score go up feels like winning. Watching it go down feels like losing. Under the old system, every mistake felt like progress and every correct prediction felt like nothing happened. It was completely backwards.
THE NEW SYSTEM
The redesign started with one question: what should it feel like to get a BANG ON?
The answer was obvious once I asked it properly. It should feel like a big deal.
So we went positive. A BANG ON earns you points, not nothing. Getting the result right without the exact score earns a smaller amount. Getting it wrong costs you. The leaderboard now shows who has the most points, not who has made the least errors, which is what leaderboards are supposed to show.
We also added a streak bonus. Consecutive BANG ONs earn progressively more. Three in a row pays out more than three separate ones. This rewards the player who's genuinely in form, not just the player who got lucky once. We started by having a BIG bonus, but winning then became like winning the pools, more of a random game than a real one. So we changed the scoring for successive BANG ONs from +5, +10, +15... to +5, +6, +7.... This means that other players (who didn't get it at all right this week) didn't think they could ever catch up. To retain players for weeks is what I was looking for.
THE MIGRATION PROBLEM
Here's where it got complicated. BangOn already had users with existing scores under the old system. You can't just flip the sign on everyone's data and call it done. Well you can, but I retained the old BANG ON bonus system. Someone who was top of their group under the old system might not be under the new one, through no fault of their own. That felt unfair.
The solution was a clean break at the game window level. Old game windows kept their old scores for historical record, just turned positive. New game windows used the new scoring. The leaderboard rolled forward from that point.
In hindsight I'd have got the scoring right before launching to real users. But you don't always know what you don't know until someone is standing in front of you looking confused at a zero.
WHAT I LEARNED
The scoring system is the game. Everything else — the design, the groups, the live updates — is infrastructure. If the core loop doesn't feel right, none of the rest of it matters.
I spent a lot of time early on thinking about infrastructure and not enough time playing the game myself and asking whether it felt fun. That's easy to say now. At the time I was too close to it.
The fix was right. The game is better. But the lesson is to listen to people. It's a skill in itself. Take off the dev blinkers.
Next post: the cron job that kept failing at 10:30pm. Debugging a live sync system during an actual live game window, from a sofa, with a phone.