To Catch a Farmer

The page refreshes and I see a list of names: Jimfun, Jimgun, Jimrun, Jimkun. All elves, all level 1, each belonging to different accounts but all charged to the same credit card. Every one of them has traded with the same high-level character. I’m willing to bet dollars to dimes that these guys are gold farmers. But I send their names over to customer service anyway, to have a look. You can never be too sure.

But I know they’re farmers. It’s my job to know. See, while you and your friends are playing in our world, grinding levels, looking for groups and ganking newbies, our servers sit and collect data about what level you are, who you’re grouped with, and who you just wasted in player-vs.-player combat. All of that data, and much more, is piped into a database.

I’m the keeper of that database. It grows by gigabytes per day, containing vast amounts of information about a given MMOG, all encoded numerically to save space. Some days I feel like Cypher in The Matrix, staring at a bunch of symbols on a screen: “All I see now is dwarf, elf, troll.”

An MMOG is an amazingly complex entity. You might believe a gold farmer could easily hide among the millions of other characters on a server, like a needle in a haystack. But farmers behave fundamentally differently than a normal player. The farmer isn’t trying to have fun. In fact, if you look at the act of farming, it’s probably the most boring thing you can imagine. But it’s efficient, and efficiency is what the farmers are optimizing for. That efficient boredom sticks out like a sore thumb. We can see this stuff happen. So it’s like finding a needle in a haystack where the needle is colored bright orange and we happen to know the density of each cubic centimeter of the haystack.

Rooting out farmers is about finding patterns. Fortunately, the human mind is actually incredibly good at pattern recognition. One of my favorite tricks is to take an incredible amount of data, say every trade that’s ever occurred between players in a game, and then just render the whole thing to the screen as nodes and edges in a graph. It’s an incomprehensible mess of spaghetti. And yet anyone, after just a split second of looking at the screen, will begin to spot anomalies, patterns in the noise. Zoom in on these fleeting anomalies, and, quite often, you’ll find there’s illegitimate activity going on there.

But of course, the real job of rooting out farmers is far less psychological and far more rooted in hard data. For example, one morning I was scanning my dashboard of macro visualizations of the world. Immediately, the graphs revealed a potential hot spot. The Cave of P’tath quest was being completed far too often in relation to similarly designed quests. I whipped up an SQL query and found out the quest’s completion rate was being skewed by a single group. Most people that complete the quest move on to bigger and better things, but these guys banged through it 172 times in the last four days. Their names: Abcde, Fghij, Klmnop, and Qrstuv.

I drilled into the characters’ recent histories. They had an awful lot of gold on hand, especially for characters of their level: They were more than five standard deviations away from the mean for gold. But where’d the gold come from? I looked at their gold acquisition records and found most of the gold came from The Cave of P’tath. They were grabbing Loki’s Amulet, an overpriced reward item, and selling it to NPC vendors for a great profit. It turns out there was a typo in a spreadsheet. QA was not pleased.

So, I knew where the gold came from, but where was it going? That’s the tricky part. First, I had to put the characters on a watch list and send it over to customer service, who can monitor the players’ behavior a little more closely. On my end, I keep an eye on their trade records. They have to move the gold at some point, and the only way to do that in most MMOGs is by player-to-player trading. (You lose too much gold buying and selling items using NPC vendors to launder things that way.) Abcde and friends are pretty high-level, and they don’t want to risk having those characters banned, since those characters took hours to level up. In most cases they’ll move their gold to a low-level alternate character (alt) on a completely different account.

I got lucky this time. They made a common mistake: In a 24-hour period, they all traded gold to the same alt, and while the alt belonged to a different account than our four farmers, that account shared the same credit card with Qrstuv. When we see that kind of behavior, it’s banning time, and this case was no different.

Granted, maybe we’re only catching the really dumb farmers. It’s like that saying about crime: The best criminals are the ones you never hear about. There are almost certainly farmers who are so good at what they do I don’t notice them. But if they’re that stealthy, they’re probably not disrupting anyone else’s game, and they’re probably not unbalancing the economy. And if they’re not doing either of those things, I’m personally fine with their continued existence. I really don’t believe there is anything intrinsically wrong with selling virtual items second-hand on eBay or wherever, despite the fact it cuts into a potential source of revenue for the game developer. As long as it’s still a “potential” source, and the developer isn’t making a concerted effort to actually get involved in the market (see Sony’s Station Exchange for an example of a pretty good effort), I don’t think the developer has the moral high ground to strike down any secondary markets.

Then there’s the question of the game designers’ intent. Having worked with designers, I know they hate farmers more than a lot of players do, because the farmers are finding and exploiting design weaknesses in products they’ve worked on for years. Some people would say, “Hey, tough luck. If your design is weak, it’s your own fault.” Others would take the stance that people who agree to play a game are entering in an implicit contract to not only play by the rules, but to play reasonably within the designer’s intent. I tend to side with the “tough luck” crowd, but I would add that if you’re going to exploit a weak design, you’d better do it in a way that doesn’t undermine the enjoyment of other players.

In the end, it comes down to a cost-benefit analysis: Is banning a suspected farmer worth the $15 per month subscription fee you lose that the farmer is paying, and on the chance that it’s not a farmer, is it worth the bad word of mouth to insinuate you don’t trust the people who play your game? It’s a delicate balance that involves the marketing and customer service departments of a company every bit as much as it involves the designers.

The attitude a developer takes in regard to farmers should be consistent and should be integrated tightly with the game’s design. When we set out to design an MMOG, we should be as concerned with our policies and attitude toward farmers and other exploiters as we are with the art style of our game world, or the pacing of combat. If developers set down some guidelines about farming early on as a core part of the experience, the designers can take that into account during the development process. At that point, the developers will hopefully have a consistent and sensible set of policies about farming the community understands, and then maybe we can go about catching the right farmers for the right reasons.

Darius Kazemi runs Orbus Gameworks, a gameplay metrics middleware company. He also has a blog about the game industry, called Tiny Subversions.