I also wrote this for my LiveJournal several years ago, but I thought I'd update it and post it here too. There are many things which separate good games and bad games, but probably the most essential is that great games were playtested, and bad games weren't sufficiently playtested. Here are my recommendations on playtesting.

1. The more playtesters, the better, because you can see what a wider variety of people think about your game. Try not to pick only people who are "gamers", as what's pleasing (or obvious) to them might not be pleasing (or obvious) to others who don't play games as much.

2. It's better to watch someone play your game than to read a report about it. What you can learn from watching people play your game you simply can't learn from what other people tell you via email or instant messenger about your game, and I kind of even feel sorry for games which are released without the developer watching anyone play it in front of them -- usually one can tell when they didn't.

3. Constantly add new eyes. Adding new playtesters with each newer version of the game helps: when someone has played earlier versions of your game they tend to look at it differently from someone who hasn't seen the versions without all the current features, since first impressions develop very differently if the first version they played was 10% done or 90% done.

4. When it comes to taking their suggestions seriously, don't do that. Ignore most of what they recommend -- it works better that way. Use what they say as raw data to make your own decisions about the game, and do think about their suggestions, but keep in mind that game players usually don't know much about game development. They do know what they find fun, what they had trouble with, what was confusing, and what they found annoying. Those are the things you're looking for, not suggestions about new weapons or new gameplay elements, you're still the author of the game, don't make it a design-by-committee game.

5. Spend a lot of time playtesting, preferably constantly; I probably spend at least ten percent of my current game's development time watching people play it or talking about it with people who have played it. Think of it this way: you want people to have fun with your game, so what's the best way to ensure that they will? Let them play it to learn what they find fun and what they don't. It's simple empiricism.

6. Your team, and you yourself, do *not* count as playtesters. You can play your game for fun and look for bugs, but you know far too much about it to be a good judge of many aspects about the game, particularly its difficulty level and how confusing or easy to learn it is. Likewise, what's intuitive to you about the control scheme or even the strategy isn't intuitive to people who didn't make the game. Make sure people can figure it out on their own, just from the game, without your advice.

I found this quote by Shigeru Miyamoto in an interview: "If a fan makes a suggestion, I will often put it in my mind, and I will take in whatever comment I feel is useful. But I make my own predictions of how a user might react to the games I create, and I would say I am sensitive to whether those reactions are in line with what I predicted. People generally have different views and opinions about anything. So I would only listen to whatever information is useful for me. It is interesting to hear what other people say. But instead of reading the blogs, I would rather stand behind a person playing the games and sense how the player is reacting to the game -- whether he is unhappy with the games, or if he is having fun. I can feel all of that directly. It is more useful for me to do that than to read what he thinks of it."

For those of you who want to see a great example of how games should be playtested, read this series of posts by Auriea of Tale-of-Tales, they're wonderful.