First, we need to form a hypothesis and then test that hypothesis against a real world data you acquired from playing the game. For example, you might guess maybe beating a level is determined by the number of moves you make on that level. When you make enough moves, you progress to the next level. All right, so you’ve got your own hypothesis in mind? Awesome. Now pause us and go test it out.

You’re back? Cool. How did it go? Was your hypothesis correct? You probably found that progressing through levels is actually determined by the player’s score. Now we need to dig a little deeper. We’re going to try to understand if there’s a purpose behind this system. The first thing we’re going to do is establish how many points does it take to complete the introductory levels? It’s time to collect some data ie. play some games. Pause this, go beat the first six stages, record your score, and come back. Go.

All right. On to the numbers. To beat level 1, it takes 500 points. To beat level 2, it takes 1,500 points. To beat level 3, it takes 3,000 points. Level 4 takes 5,500 points. Level 5, 8,800 points. Level 6, 13,200 points.

A designer should always be thinking about how the player experiences the game. How does the player experience those numbers? What would we have to do to our current data set to make it more accurately represent the player’s experience? Well the player rarely starts at level 1 thinking, “Oh man, only 13,200 more points till I reach level 7.” No, no, no, no. They spend much more of their time thinking how long until I beat this stage. Were going to have to take the difference between each level value to see how long it takes to get from one level to the next, much more useful than simply looking at raw point values required to get to a given level. Pause, do the math, and come on back.

You got it? Here’s how the numbers should break down. It takes 500 points from level 1 to level 2. It take a thousand points to get from level 2 to level 3; 1,500 points to go from level 3 to level 4; then 2,500 points from 4 to 5; 3,300 points from level 5 to level 6; and to get from level 6 to level 7 takes 4,400 points.

Now that we know how many points it takes to get between the first few levels, we can say it gives us a weird, slow exponential curve. Now what? How do we make this data useful? Well, let’s return to the player. What other system affect how the player experiences this system? Well of course, the scoring system. Now we have something else we need to get data on. How many points does matching 3 gems on any given level earn you? How about matching 4 gems on any given level? How about matching 3 gems and chaining into another set of 3? Pause, grab some data, come on back.

You got it? Awesome. Matching 3 gems at level 1 will give you 10 points. Matching 3 gems at level 2 will you 15. Matching 3 gems at level 3 gives you 20 points, then 25 points, 30 points, and at level 6, it gives you 35 points. We see that all other block combinations simply give a multiplier to this match 3 base score. How does the player experience the conjunction of these two systems?

To put these two systems together, we’re going to have to divide the amount it takes to get to the next level by the amount of points that a base match 3 will get you on any given level. Now you may ask why we’re doing that. What will the number we get out of that mean? If you guessed it will define how hard the level is, you’re dead on. There’s another important variable here. What helps define difficulty? Your losing condition. In Bejeweled 2 classic mode, what’s the losing condition? When you can’t make any more moves.

Also worth noting is that at the beginning of every level, the board is shuffled. I’ll leave it to you guys to piece out why on average, making a valid move in Bejeweled leaves fewer total valid moves on the board. Just assuming that’s case, the number we’ve now come to, the amount of points to progress to the next level divided by the base points on a given level tells us directly numerically how challenging we can expect the given stage to be. What’s amazing about this system is the level of control over the difficulty it gives to the designer. You can see it right there in our initial data set. PopCap took a vaguely exponential set of numbers, used a completely linear number set to smooth out that exponential curve, and then tweaks the numbers wherever play test told them that players were having a hard time or not being challenged enough.

For example, if you’ve been doing all the math with us up to this point, you may have already noticed that the increase in difficulty from level 3 to level 4 is much greater than for level 4 to level 5. This relatively complex system is totally invisible to the player. To them, it simply feels like they’re progressing since the system is built to continuously increase the amount the player is rewarded and make them feel empowered without breaking the difficulty curve. In fact, its this system that let the designers construct Bejeweled carefully hone the challenge curve, but this entire system is dependent on the lose condition and board re-shuffling it every new level. A slightly lazier designer who didn’t shuffle the board between rounds would’ve created a totally different and probably inferior game.

The really incredible thing is that you’ll see players who’ve played a lot of Bejeweled no matter causal, no matter how uninterested in strategic thinking, playing around with this system. Ask any long time Bejeweled player, and they’ll tell you not to use your special gems until you need them since they carry over between levels. If you ask them why, they’ll probably tell you, “Well, because it’s better,” without a word about the math going on under the hood. The math ensures that that’s the way things work, and it also ensures that they get a sense of that for playing.

Also, if you really want to know how deep this rabbit hole goes, answer me this. How much harder is it to make your first match on level 5 as compared to level 1? That’s correct. No harder at all but your road for doing so is vastly increased. This system is even designed to make you feel like a bad ass every time you get to a new level just by the way they combine their point system and their lose condition.

You can see the depth of thinking that goes into making a good game, puzzle game, or any other kind. The next time you hear someone trying to claim that building little puzzle games like this is simple, you will know better. Just for the masochists out there, here’s a little easy homework assignment for you. See if you can figure out what they changed in Bejeweled 3 or the Facebook version of Bejeweled and then see if you can figure out why. If you’re ever feeling stumped, just come back to the golden question: how does the player experience this?

Additionally, early testing gives you data that you’re going to need for later. Each time you test, you’re going to see if the adjustments you made to the game, made the issues you encountered better, or worse. To do this, you need as many data points as possible. The longer your trend line, the more in depth your understanding will be. One final word about early testing. You may be wrong about your game. Embrace that. You may well find that what you thought was the core engagement of your work, wasn’t what people actually found engaging. Don’t reject that, that is the most valuable data you can get. As anyone who has built a game will tell you, the earlier you make changes, large or small, the cheaper it is to make them. Don’t cost everyone else on your team weeks of work, just because you were afraid to playtest early.

Now, let’s go over a few quick and dirty rules of how to test. First, try and talk to your players as little as possible. Anything you say to them, biases them, and prevents you from getting good data. You may feel the need to explain your game before starting, but resist that urge. Let them take a crack at it, before you tell them what they should be doing. This will allow you to observe what they naturally want to do in your experience, and gives you a better sense of where they get stuck, or what needs to be explained. Once they’ve had a go, or 2, then you can inform them.

When observing them, listen carefully for when they vocalize. Anytime your game draws an utterance from them, a sigh, a gasp, a thoughtful hmm, you’ve either done something right, or something very wrong. Also, try and see where they spend most of their time looking in your game. It will give you deep clues into what they feel is important, and what data they felt like they had to spend most of the time processing. Admittedly, this can be a very hard one to do, and you’ll often have to resort to asking the player, unless you can get all high tech and have a camera set up to record the session.

Additionally, create a test survey, rather than trying to get everything from simply talking to your players afterward. This will not only allow you to get data from a lot more people at once, it’s also unbiased data that can be cleanly turned into metrics, if you give your playtesters a 1 to 10 scale question, which I highly recommend doing. It puts less of a burden on your playtester, than making them write out whole paragraphs. It gives you an easy way to know if you’re making changes for the better in future tests. Just watch to see if the numbers trend upward or down. Just be sure to leave the playtesters room for comments at the end too, because that’s also helpful.

When you do talk to your testers, just be aware that as human beings, we all have a tendency to jump to solutions, rather than stating the underlying problem. When a playtester offers you a solution, like, “There aren’t enough power ups,” really dig in, and try to establish what underlying problem they are trying to solve. You’ll often find, it’s something else like, “My gun doesn’t feel powerful enough,” which may require a totally different solution than the one they offered.

Lastly, be aware that we are all hardwired to avoid suffering, so make sure you’re testing in an environment where you tester feels okay walking away from your game. If you lock them in a room, or even have them in an environment where they feel socially compelled to play, they’ll try to make the best of it. They’ll try to get into your game. That will deny you one of the most important data points you can get, where your player gives up on your game, or walks away to do something else. Remember when you launch your game, you’re not only going to be competing with all the other games out there, but with every book, every movie, every TV show, even the the internet itself, so your game better be ready for it.

If you don’t have to worry about NDA’s, it can often be fun to grab several of your other game making friends, and test as a group. That at least gives your players other options for what to do with their time. Before we wrap this up, let’s talk about who to test with. Test with everyone, but know that the test results you get vary in value. For example, you are the worst tester of your game. You’re just too close to it. You’re naturally going to test your own game a thousand times over the course of its creation, but that sort of testing only goes so far. There’s all sorts of things that you’re going to get used to, things that will become second nature to you, that will seem totally alien to anyone picking up your game for the first time, so you need other people.

As a rule of thumb, if you think the difficulty on your game is just about right, it’s probably way too hard. Next, family and friends. Second worst testers. They’re biased, they can’t help it. Next comes the hard core. Hard core players are often the most willing to test but unless they’re the only audience you’re looking for, and they are a niche, you need to find other testers too. As hard core players, we simply have too much ingrained in us from the thousands of hours we spend playing these things. You’re really looking for people without preconceptions or biases. People who like games, but aren’t in any way invested in them. Strangers who owe nothing to you, or better still, know nothing about you.

Who makes the best tester? In my experience, little kids. They may not be your audience, but if they are, you can always count on them to be open, eager, and 100 percent honest. I hope that was useful, good luck honing those games to perfection.