Category Archives: Game Thoughts

The Freemium Funnel

Free to play’s been featuring heavily on my mind lately, and I figured I could shed some light on how freemium games make their money. This applies to all free to play games. This is not exclusively a casual concept. Remember that Team Fortress 2, League of Legends, and DotA2 are all free to play games. Many of these lessons, I learned from Design Rules for Free-To-Play Games. Specifically, the extra chapters you get for buying the eBook version.

To understand free to play games, you need to understand three major concepts: Acquisition, retention, and monetization. These are stages of a player’s life cycle, and the company wants to make sure that as many players get to that final stage as possible.


This is the process of acquiring a new player. Getting new players usually comes from spending money on marketing and from organic growth when users get their friends to play. A key concept here is the Cost Per Acquisition (CPA. Also called CPI or Cost Per Install, but let’s stick with CPA), which is how much it costs, on average, to get a player to start playing your game.


Retention is keeping those players you just spent money to acquire. This is often described as seven day retention, thirty day retention, or something along those lines. A seven day retention rate of 50% means that half of the people who sign up for your game are still playing it a week later. A game with a high retention rate is one that players keep coming back to. Eventually, though, every player stops playing a game, and this is called churn, the opposite of retention. A game with a high churn rate is a game that players aren’t sticking with.


When a player decides to spend money on your game, that’s when they’re monetized. If you’re lucky, you’ll see a 10% conversion rate (a term to describe how many players decide to convert into paying players). This is what pays the bills and keeps the game running. The important concepts in monetization are Average Revenue Per User (ARPU), Average Revenue Per Paying User (ARPPU), and Lifetime Value(LTV). ARPU is how much money you get on average from each player you have. ARPPU is how much money you get on average from your paying players. LTV is how much money a player is likely to pay, on average, before they quit playing.

What This All Means

The success of a freemium game can (in extremely broad terms) be summarized by comparing the Cost Per Acquisition to the Lifetime Value. This is how much money you spend to get someone to start playing your game and how much money they’re likely to give you before they stop playing. If you have a low CPA and a high LTV, you’re doing great! On the other hand, if you reach a point where you’re paying more to get a user than they’re going to pay, it’s probably time to sunset the game.

Where Design Fits In

For the most part, design in a free to play game is about retention, retention, retention. That isn’t to say that designers can ignorer acquisition or monetization; the game would fail if they didn’t. I just mean that retention helps with acquisition and monetization all by itself. A healthy retention rate means players stick with a game for longer, giving them more time to convert to being a paying user and giving them more time to convert into a paying user and boost LTV. A game with great retention means that the game gets a large community with a lot of seasoned players who can drive organic traffic and bring in more users, who will themselves bring in more users, which drives down CPA.

There are techniques you can use to design retention into a game. Giving players rewards for logging in on consecutive days is a popular one, and getting the player to schedule appointments to check in on the game (Come back in 24 hours to harvest your strawberries!) is also common. There’s one technique even more common than those two, though, and it’s a technique that every game, freemium or not, strives for, and achieving it has been the subject of countless books and endless debate: Just make the game fun.


Posted by on August 11, 2014 in Game Thoughts


Tags: , , , ,

Randomness in Puzzle Design

After reading my last post, a friend of mine (who hadn’t really considered that puzzle levels could be designed) asked me if the tile colors of puzzle levels were designed as well. The answer is that it really depends.

Your typical “swap” tile matching game generates tiles completely randomly upon initial setup (except for things like not having matches before the player’s even had a chance to take a turn). This has a few advantages. A player who just experienced a loss knows that it’s random, and maybe their next initial setup will be more favorable. This is especially important for the free to play market, where retaining players is the arguably the single most important element to success. That feeling that “Maybe I’ll have better luck next time” makes them much more likely to click on the Retry button.

That’s not to say that they don’t sometimes have fixed pieces. This happens, and when it does, there’s always a reason for it. For example, let’s have a look at level 22 of Cookie Jam:



The level objective is to clear all of the wafer tiles. They’re functionally equivalent to the jellies from Candy Crush. There are also those donut obstacles. They can be swapped, but they can’t be matched. The only way to get rid of them is to make a match next to them. These donuts cover up 30 of the map’s 50 waffle spaces. 12 of these donut-covered waffle spaces are also chocolate waffle spaces, which act like Candy Crush’s double jellies.  The player needs to match 62 cookies on these spaces, but the donuts are blocking 42 of these necessary matches.

The cookies are generated randomly, but look at the right side of each of these maps. Every time this level is generated, there are two tiles that can be swapped that will match five cookies of the same color. This gives players an optimistic start to the level, since matching five cookies feels good, and also generates a rainbow cake that will eliminate all cookies of any color it is matched with.

The actual benefit of the rainbow cake this early in the game is marginal at best. When it’s created, it’s going to be sitting in the bottom right corner, with a donut on its left. The only color the player can use it to eliminate is whatever color happens to be right on top of it. Is the rainbow cake useless, then?

No! Far from it. In this level, its purpose isn’t to be used. Its purpose is to be created. When the player sees that 5-cookie match, they go for it because they want that rainbow cake, but its real purpose is to clear away three of the donuts and give the player a foothold in the lower right.

When the level begins, the easiest place to make matches is along the top edge of the donut structure. Because of the voids in the level, though, vertical matches are difficult. The left side is practically impossible to match, since it’s not only a map edge, but also constrained by the donuts. Same situation in the right, except for that five-cookie match. Once that happens, the right side is much like the top edge of the donut structure: In practical terms, you can really only match parallel with it, but you do have another row (for the top edge) or column (on the right edge) to swap tiles with.

You might wonder why they don’t just have the donuts on the right removed at the start in the first place. My feeling is that because this is a bit of an intimidating level to start, throwing in a free rainbow cake encourages the user. It also gives them an additional edge they can work with, while preserving the level’s pleasant symmetry. People love symmetry. It also serves as a nice reward for players who hold off on using the rainbow cake, since they can use it near the end of the level to collect difficult wafer spaces.

So, there you have it. An example of a Match-3 level where tile placement is fixed. It’s a different color each time, but mechanically-speaking, it’s always the same, and it serves a very clear purpose.

Leave a comment

Posted by on August 9, 2014 in Game Thoughts


Tags: , , ,

The Infamous Level 65

The last time I talked about the modern Match-3 game, I said there was a lot more than just a shift to move limits. The biggest identifier of the modern Match-3 is that it has level design. I’d like to demonstrate that by breaking down level 65 of Candy Crush Saga. When Candy Crush Saga first launched, it had 65 levels. Level 65 was notoriously difficult, and it’s been altered to make it easier. I’d like to go over what made it so difficult, and how it was changed.


This post is going to be a long one, so bear with me.

Read the rest of this entry »


Posted by on August 9, 2014 in Game Thoughts, Puzzle Design


Tags: , , , ,

Modern Match-3


I was playing Candy Crush on my phone on Saturday, and a friend asked what I was playing. I said it was Candy Crush, and I forget how the rest of the talk went, but they called it a Bejeweled clone, and that stuck with me. It reminded me of when people used to refer to first-person shooters as Doom clones, and how for a while, League of Legends was called a DotA clone.

Match-3 games have come a long way since Bejeweled, but why are they still compared to Bejeweled specifically? The first thing that comes to mind is that for many people, Bejeweled was the first Match-3 that stuck with them. They haven’t really played Match-3 games since, and when they look at a modern one, they think back to Bejeweled. Of course, modern Match-3 games are vastly different beasts, although people don’t quite realize it yet.

Bejeweled 2 (My baseline for this discussion) is based around making matches until you reach a target score, then you’re presented with a new level (keeping your special gems) and a new target score. This target score keeps rising (and so would your points per match). However, if you end up in a situation where no more matches are possible, the game  ends. The longer you play in a level, the more likely, you’ll end up in a losing situation, so over time, you need to make more and more of your matches into higher-scoring ones and combos to get more points out of the time you have in a level.

Most modern Match-3 games still have a score, but you no longer lose if you run out of possible matches. The board is just reshuffled. The loss condition is different: You’re given a certain number of moves to reach your goals (the old target score is now just a formality or omitted entirely). If you run out of moves, you lose. Why is that?


You have limited moves instead of playing until you’re stuck for a few big reasons: The first is that the player has a clearer idea of how close they are to losing. Match-3 play is about identifying possible matches, and a fair amount of a player’s time is spent looking around for the possible matches they can make. They will clearly overlook some, and they know this, so if they make a match, they won’t know for sure that the end is coming. They might know it’s near, or it might catch them completely by surprise. “What? The game is over?” Giving the player a limited number of moves more clearly communicates to them the resources they have remaining and gives them a stronger feeling of control over the outcome.

The moves system also gives the designer more control. Many new Match-3 games involve objectives other than scoring as high as they can. I’ll go into more detail on this in a future post, but the general idea is that Match-3 levels can vary widely in difficulty. A designer can create a level, do some testing, and conclude “This level takes, on average, 50 moves to beat.” They can give the players 50 moves, and then the player has an exciting experience where they’re running out of time, but victory is just within their grasp, and they make it with their final move. Or they might lose with just one move left to make (and consider spending money on buying just one more move so they can beat the level).

This is just one way that modern Match-3 games are characterized. I’ll talk about this more later.

Leave a comment

Posted by on August 7, 2014 in Game Thoughts, Puzzle Design


Tags: , , , ,