With your philosophy of the game loop, coming up with a self-contained important core mechanic, do you ever significantly modify it with content updates later?

HS: No. This is one of my key takeaways. What I discovered is that there's numerous examples. There's Roller Coaster Kingdom by Zynga; they changed the game loop after launch -- failed. They killed the game. Nightclub City by Booyah changed the game loop to an energy mechanic -- failed. They changed back.

Even Happy Hospital by Wooga, we changed the game loop of a game we already launched. It wasn't a success, it wasn't a failure -- it didn't move the needle, but it wasn't worth the effort.

We put two months of development in changing the game; it didn't move the needle. So even that was kind of a failure. So you should detect problems in the game loop before launch, and there's numerous ways to do that.

One is do a board game. One is obviously user testing. One is to do an early concept shot. Show the concept shot to people who are not working in the company -- do they understand the game loop just from a static picture? Then you're good, or have one level of assurance.

Do a very early first playable inside the company. So develop for two months, just the core game loop, no fancy graphics, even boxes do. Is it fun? Do people start competing in money or something? That, you should get from a core game loop.

Then you have an early first playable. We check second playable, alpha, beta, do user test again, and then kind of like a final decisive moment -- do a closed beta. So we have done a closed beta with Magic Land in an English speaking country, taking 3,000, 4,000 users, see do they like the game, do they come back tomorrow, do they finish the game loop? And by doing a closed beta in this manner, you won't burn the best customers, because the best customers always find the game first.

So when you invite your friends you think about, "Hmm, who's interested in CityVille? Can I invite the best city building guys?" If you have a crappy game, you'll never get them back in. So closed beta is the last milestone when you should detect problems in your game loop.

After launching if you just do modifications -- small things -- you can change the time of a timer, or the UI, or amount of food you need to supply for a business in CityVille, anything like this. But you can do modifications -- timers or something.

A good example in Monster World, in the game early on, you plant lemonade trees. First, they were five minute contracts, so they would be ready in five minutes. We changed it to three minutes, huge change on one-day retention, because new users got the good feeling of harvesting them two minutes earlier, before they abandoned the game. So these kinds of things you can do. It doesn't change the core game loop -- you just change the statistical value, like a static number. But you shouldn't change the game loop.

How simple should it be? If you're talking about something that you can show people as a static image so they understand it, to use one of your criteria.

HS: For example, Magic Land, the core game loop of the game -- it's a city building game, It really is. It's one of the core themes that generates money. Anything that generates money usually is the core game loop. In-game money. I call it "soft currency". Hard currency is real money.

So, for example, in Magic Land, the first playable, you had four houses, you had a castle in the middle that you cannot do anything. You could place the houses around the castle wherever you wanted and connect them with a road, and click on the houses and select the contract that you want to produce money four minutes, or three minutes. This was the game.

What happened is that people clicked on the moneybags when they were ready and they went "cha-ching!", and they laughed. And people started clicking a little bit -- "Oh, I got more money than you!" And this was already a game. And it can be really simple like that.

One of the pitfalls of game design, or game development, has always been, somebody says that "This game is not yet fun, let's improve the graphics." That's not the way to do it at all. Then you have a problem in the core game loop. So that's why I always limit amount of content for first playable. You shouldn't have 20 houses there, then you have focused on the wrong things; you should have four. And if it's fun with four, it will be super fun with 20. But if it's not fun with four it won't be fun with 20.