May 21
Simulation vs. Game
So I’ve wired up my initial “Forum Buzz” calculation and discovered.. that while my plan for its calculation is probably a good mirror of what happens in the real world, I’m not convinced that it works in the context of a game.
Basically, here are the game rules as they currently stand: Only players who are addicted to your MMORPG will generate any kind of forum buzz, on the theory that nobody else will bother to post on your forum. The addicted player will add either positive or negative buzz, based upon their current satisfaction with the game. (So if you’ve just nerfed their class, they probably will be generating negative buzz). Buzz decays over time (currently at a rate of 25% per day.. I’ll probably increase this to 50% per day, but it needs more testing). I then base the rate of new subscribers on a weighted buzz value which is equal to: (2 * positive_buzz) + (1 * negative_buzz).
So as a rough rule of thumb, if you have two hundred people posting negative comments about your game, you’ll get half as many new members as if you have two hundred people posting postive comments about your game. This works off the assumption that any sort of exposure will bring you more customers than no exposure, regardless of whether the exposure is good or bad (although obviously good will be more effective than bad). This seems to make sense, and actually appears to be more or less what we see happening in real MMORPGs. The problem is that I’m not sure whether this works well in a game format.
See, what it effectively means is “big games get more buzz, and more buzz generates more players for the big games”. Big games get bigger, small games get smaller, almost regardless of whether the buzz is good or bad. It really can’t be argued that that’s what’s happening; certainly most of the new MMORPG players are going into the big MMORPGs like WoW. But I don’t think that’s going to be fun to play; it kind of makes having lots of subscribers become the only requirement for success at the game; if you have lots of subscribers, your new subscriber base will grow massively, pretty much regardless of whether or not your player base actually likes what you’re doing, because either way, they’ll be generating a lot of buzz on the forums which will bring a lot of new players into the game. This is, of course, assuming that you manage to keep your players addicted. But so far, that’s actually not turning out to be at all difficult to do. Maybe I need to improve the ‘addiction’ algorithms, to make that a little trickier to generate.
Need to think about this more.
May 22nd, 2008 at 4:46 am
Something reasonable might be having negative forum buzz influence addicted players to quit. It’s realistically reasonable — if you’re on a forum sharing your dissatisfaction with a bunch of people every day, you’re going to be more likely to quit. This would also mean a small game with lots of happily addicted people would be able to keep those people, while a large games that aren’t satisfying would tend to have low numbers of addicted people (since they quit because of bad buzz), which would stunt their growth.
Or you could have buzz weighted by how many players are addicted. You could weigh down the negative buzz, giving something like (1-negative_buzz/total_buzz)*negative_buzz, while keeping positive the same. Say you have 10 players generating forum buzz, and 9 of those are dissatisfied. Currently your buzz value is 1*2+9 = 11. In this system it would be 1*2+(1-.9)*9 = 2.9. It would reward have low numbers of negative buzz. If you reverse the example to 10 total people and 9 are positive, you’d be getting 9*2+(1-.1)*1 = 18.9. In this system, negative buzz still raises your exposure, but having a lot of it is bad.
Just some ideas.
May 22nd, 2008 at 8:36 am
Will it at all be possible to run the MMO on one server such as Eve?
I always thought that was neat how CCP did that. It’d be fun to emulate here.
Also, how about addicted players quitting due to becoming too addicted and having their lives wrecked in the process?
May 22nd, 2008 at 10:13 am
Right now, no, each region of the MMO runs on its own simulated server. In the original design, you allocated servers independantly from regions, and could resize and move the areas each server handled around on the world map to do load-balancing. But I’ve cut those features from the first release, in the interest of simplicity and being able to finish on time.
Right now, the level of addiction is only used to make it less likely that someone unhappy with the game will unsubscribe. It does not (yet) actually make them spend more time with the game. With 10,000 users being simulated at once, I’m not sure how much value there’d be to simulating something as subtle as the quality of their outside lives. ;)
jyrenth, simplifying the math of your suggestion (the ’1-’ operation was breaking my brain), the buzz generated by negative comments would look like this: “(positive_buzz/total_buzz)*negative_buzz”. So effectively as you get less and less positive buzz, the negative buzz stops being as effective. I’m not sure whether or not it works that way in reality, but it’d certainly work a lot better in a game context. I think I’ll probably go with that. Thanks! :)
Now I just have to do something to try to balance out the buzz a little; it probably goes heavily negative just a bit too fast when a server goes down, taking a hundred or so players with it. :)
May 22nd, 2008 at 12:26 pm
How about this: Negative buzz follows the derivative of a logistic curve in terms of how many new subscribers it brings in whereas positive buzz attracts a logarithmic function of itself in new subscribers. That way, positive buzz will be better passed a certain point. To make negative buzz worthwhile, you’d have to manage it to be in the sweet spot of the logistic curve.
This is assuming forum buzz grows proportionately to the amount of users, eventually people get so angry and dissatisfied with a game that the sheer mass of negative buzz keeps people away instead of piquing their interests. Just an idea, you could use any functions to model the change I guess.