Balancing Fruit - Fruit Ninja Mini Game Analytics

9 Apr 2023

This article is about some analysis work I did back when I worked at Halfbrick Studios. I originally presented this at an IAPA meeting in Brisbane in 2016.

Halfbrick is a global games development company which was first established in 2001. Based in Brisbane, Halfbrick has developed and released various successful games on multiple platforms.

Whilst I was at Halfbrick one of my roles was working as a data analyst on Fruit Ninja. This game is now pretty much a cultural icon and has more than 1 billion downloads across all platforms.

When the 5th anniversary of Fruit Ninja came around we did lots of things to celebrate it. From cross-overs with other Halfbrick games to a documentary about the making of the game itself it was a really exciting time to be at Halfbrick; and that’s saying something as Halfbrick was pretty exciting even on a normal day. 

One of the main things we did in celebration of the 5th year anniversary was to release 7 new game modes in Fruit Ninja itself. Out of these 7 new games modes, 6 were mini games, and 1 was the 5th anniversary Tournament. I’m going to be talking about some analysis we did for the 6 mini games.

The 6 mini games were part of the new game loop we designed:

  1. Play the 3 Original modes (Arcade, Zen, Classic) to earn Starfruit
  2. Pay Starfruit to play mini games and earn Golden Apples
  3. Pay Golden Apples to play the Tournament

Mini games were unique variations on the Original game modes; but were still the core slicing mechanic. Depending on their scores, players could win 1, 3, 5, or 10 golden apples.

  • 1 golden apple was given to beginners who were still learning a mini game.
  • 3 golden apples were given to those who had learned the mini game but weren’t good yet.
  • 5 golden apples were given to those who were good at the mini game, making only a few mistakes. 
  • 10 golden apples were given to those who had mastered the mini game.

Regardless of how low their score was, players would always get 1 golden apple. This is because players had to pay 100 starfruits to play a mini game so we thought it was only fair that they get something for their trouble. We also wanted all players to get a chance to play in the tournament at least once.
After shipping any major update we always checked in to see how the players were reacting to the changes. After we shipped this update we ran the numbers for the mini games. 
The graph below shows the total number of games played soon after the launch of the update. The X-axis is in millions. What we predicted was what we were seeing in this graph: Players play all the games and play their favourites a little more often. 
No cause for concern. Or so we thought.

The graph below shows % of total games for each mini game broken down by what reward levels (golden apples) players achieved in each game. 1 being the lightest and 10 being the darkest. This is where we started seeing things we had not predicted or designed for. 

Two of the mini games looked as we expected.
Time Attack: Conformed to the difficulty curve we set for it. It’s the most familiar mini game to Fruit Ninja players as it’s basically Arcade with a time extension mechanic in the form of lemon waves. It’s also one of the easiest mini games. Players spend some time learning the mode but once they’ve learnt it they are mostly getting 3’s. Less games are 5’s and then less games again are 10’s, and this roughly reflects number of games played and the spread of skill in the player base we had observed previously.
Cherry Bomb: This also conformed to the difficulty curve we set. Cherry Bomb is a much less familiar play style to the existing Fruit Ninja and it’s also a much harder mini game. We still see the same difficulty curve only it’s skewed towards more 1’s, as players take longer to learn the game and have a higher chance of having an unlucky game. Again we see a good chunk of 3’s, then we see a lot less 5’s and 10’s because it’s harder to achieve these. 

Now things got weird….

Fortune, Swarm and Quick Draw: All these were way out from our intended design, 80% of the games are 1’s. 
This surprised us because:

  • Swarm & Quick draw have similar play styles and difficulty to Cherry Bomb
  • Fortune is a completely different play style and difficulty to Swarm & Quick Draw. It’s supposed to be easier and much closer to the difficulty of Time Attack

Now things get crazy...

Juggle: 80% of scores are 5’s and 10’s. This really blew us away. Juggle is not any easier than Time Attack and it is probably the least familiar play style out of all the mini games for Fruit Ninja players. It should be hard yet we see players killing it pretty much straight away. 
What was going on?

Theory 1: Players are just really bad at some games, so don’t play them much.
Answer: No. We ran the numbers and that just wasn’t the case. In the graph below the y-axis shows average games per player per day. You can see on this where each of the mini games sits. Juggle is highest as you would expect. We have lots of returning players each day so if players really weren’t playing the mini games they were bad at, then we would see games like Fortune, Swarm and Quick Draw with averages well below 1 but instead they are all well above 1.  

Theory 2: Players are “grinding” for Golden Apples
Answer: Most likely yes; for a few reasons.
Grinding is a term in gaming for when players repeatedly engage in the same gameplay to earn rewards. It is very often much less enjoyable than the intended game and feels a lot like work. If you incentivise grinding enough, even inadvertently, you’ll burn players out.   
Juggle had a known exploit where players could swipe continuously at the bottom of the screen. This is a lot of work for the player and it requires very little skill. It’s easy to get rewards, but it’s not a rewarding experience: It’s a grind. 
We didn’t know if many players would find and abuse this. They did.

In the case of Swarm, Fortune and Quick Draw these were the only games with kill bombs. Hit a kill bomb and you end the game instantly. You can end a game with a score of 0 and still collect 1 Golden Apple. 

Players were choosing to convert 100 Starfruit to 1 Golden Apple this way. We already knew there were large existing Starfruit balances in the Fruit Ninja player base prior to the 5th anniversary release so this made a lot of sense once we thought about it.
Again though: This is a grind and is not fun! In this instance players were actively avoiding playing the game. That’s not good at all. It’s even worse than what players were doing in Juggle!

We reacted by changing the game design in the following ways. 

  • Replaced all kill bombs with score bombs. Score bombs take points off your score and break your momentum but it’s still possible to recover and have a good game. 
  • Added time limits to all the games that previously had no time limits. They had previously had kill bombs.
  • Rebalanced the difficulty settings & reward levels of each mini game

Here’s what happened after we made those changes. 
We are looking at the same chart from the beginning that looks at the percentage of games played in each mini game broken down by reward level achieved. We are looking at the data before we changed the game vs the data after we released the changes. 

  • Cherry Bomb & Time Attack: Stayed pretty much the same
  • Fortune: Now looks a lot closer to Time Attack in terms of difficulty; as it was originally designed to be.
  • Juggle: Players still get good scores in this mode; but they are getting them for skilful play and so the score distribution reflects this.
  • Quick Draw & Swarm: These are still two of the hardest mini games, but now they look more like Cherry Bomb in terms how players are able to learn and achieve in them.

We had achieved our goal of happier and more skillful ninjas!

We used analytics to see what was happening and our understanding of our games to work out what it meant in practice. When you add qualitative information to quantitative analysis often great things happen. 

Follow me on SolMedia


Enjoy this blog? Subscribe to andrewsaul


No comments yet.
Most relevant comments are displayed, so some may have been filtered out.