rangerBlog - June 13th, 2007

03:07 pm - When Autobots Cheat

Last week I talked about the macroeconomics of the Battle for the Allspark online game; how 2/3 of users chose to play as Autobots, giving the Decepticons an economic advantage of scarcity that translated into an overwhelming tactical advantage. On June 2nd, 2007, the Decepticons had conquered every single battle zone.

But then- something new happened. Cheating.

The economics of cheating

'Cheating,' in the Battle for the Allspark online game primarily takes one form- forced-forfeits.

In order to discourage players from simply forfeiting a losing match the game designers created two dis-incentives.
1) The player forfeited to would recieve 25points (only about 10% of what they'd get if they won, but still something)
2) The player forfeited against's team would receive a 'win' in this zone.
Thus, any player skipping out on a match would be helping the other team to win the game. In a game where 1/4 of the zones are routinely held by margins of less than 50 wins, a substantial disincentive, right?

A substantial incentive

A little less than a week after ranking and zone play were enacted, the predominately Decepticon-held zones slowly began to drift back to blue again, equalizing at about a 50/50 split.
Was the the long-awaited equalization that would happen as the Autobot players reached the highest experience level? Perhaps, but...
The 10 zones the Autobots have re-claimed were the 10 zones the Decepticons had held by the narrowest margin. That might be expected- but there was no corresponding rally in the other ten zones. These Autobot wins had been targeted.

Someone with a sharp eye (not me I'm afraid) spotted the telltale sign. In these newly-reclaimed Autobot territories- the Decepticons still had more points- by a wide margin. This suggests that the flood of 'wins' bringing these zones back into the blue were actually forfeits.

Good Visualizatons / Bad Visualizatins

Now, this intrigued me. I had previously created a mashup for tracking the zone wins in a manner somewhat easier to read than the zone map (where you had to do math to figure out the margins of leadership.) Now I amended this mashup to highlight zones were the team winning actually had fewer points by putting up a yellow bar. (These stats are live, but presumably you'll still see the bars. Bear in mind Autobots are blue and Decepticons red in this game- no I don't know why.)

Well, that was helpful, I could now highlight fraudulent zone wins-- but not fraud; the yellow 'alarm bar' was not trigger by cheating- but only when cheating was pervasive enough to make the zone switch hands. This visualization told us something about the scope of the problem... but little else.

Additionally, it's terribly unfriendly. A sea of numbers and colors and precious little explanation. Users who understood what was happening got what it represented- but it was not a tool you could use to explain the problem to new people.

Cheating 1.0

Let me take a minute to explain the nature of a forefeit-cheat. A player (or a group of players) would enter a game zone... and sit there. When they entered into battle, they would do nothing- forcing their opponents to wait the maximum 10 seconds between moves to slowly kill them. Most players forfeited a win to them rather than continue to play.

Alternately, a forfeit-cheat could enter a zone- and them simply close the zone window, leaving a 'ghost' of their player in the zone that could be challenged, but would lock their opponent into a 'waiting for opponent screen' for up to 10 minutes before the session times out in their favor. Again, most other players were forfeit in frustration.

Both of these methods involve creating a target that tempts the other player to forfeit against them-- a lot like an Out-of-Order Coke Machine collects coins from unsuspecting passerby. As you can imagine deploying them on this scale caused lots of complaints that the game wasn't very fun to play-- people were sick of fighting players they had to forfeit against.

Cheating 2.0

Slowly over several days the Decepticons reclaimed all the Autobot zones despite their inflated leads.
And again something new happened and the Autobots regained half the boards. But unlike last time, when it was a slow drift taking half the night, a this time the tide turned in less than 45 minutes, almost like the flip of a switch.
What was happening? Last time this happened there was a wave of forfeits so bad the game was almost unplayable! But everyone agreed the forfeit problem had become greatly reduced, it certainly wasn't ruining the gameplay experience anymore!

Weaponized forfeit

The mechanism is quite simple. Why wait passively for someone to forfeit against you, getting a win every few minutes, when you could simply create another account on the opposing team and forfeit against yourself? Much faster, and much less conspicuous because you were no longer victimizing honest players.

My existing zone stats mashup wasn't up to the task, it was all big picture. I needed to zoom in. In this case, this is Zone 2- the Decepticon heartland. At the time of this writing, the Decepticons hold Zone 2 by a narrow margin, but the internal game statistics tell us that if forfeits were not awarding wins- the Decepticons margin would be more than 600 victories. This mashup eliminates all forfeits on both sides (Because the Decepticons are employing the cheat as well- just not as prolifically.) Clicking through the zones, it paints a grim picture for the Autobots, who are actually losing in almost every zone. (On most days it's simply every zone, but it's a new week so...)

Neat as this visualization is, it's still not very good for explaining things to people. They have to understand not only that forfeits count as wins, but that the Autobots are cheating even thought he Decepticons are forfeitting more, and those Decepticons are proxy agents for... and that's about where most users eyes cross. "What do you mean forfeits are worse? That problem has gotten much better since last week!"

The cheating problem is simply too abstract for most people to understand without a better example. The first mash-up proved it ws happening statically, the second proved it was happening empirically, but for people to understand- I needed to prove it was happening tactilely with an example people could touch.

To catch find a thief

One of the issues I have with the Battle for the Allspark game is its statistics. The reason I created the first mash-up was because the game told you very little about the actual scores- obscuring or failing to provide a lot of meaningful data. That data needs to be free!

So I created mash-up #3- look up another user's score.
This is a wonderful tool from a psychological level. You can look yourself up, or look your friend up and confirm that the tool does exactly what it claims to- accurately tell you someone's wins, losses, forfeits and score in the game- both this week and cumulatively. Thus, its believability factor is high.

It was my theory that Autobot cheaters were using individual accounts for their cheating. So somewhere out there there should be user accounts with insanely high forfeit totals.

I don't have the in-game statistics for every player in the game, the mash-up generates a live database query for a single user. I did a test to estimate how long it would take to individually download the stats of every registered account, and I think my estimate was 36 hours. The number of users has more than doubled since then. So a mass-lookup wasn't an option.

I mounted a fishing expedition; wandering through zones and looking for multiple forfeits. (And playing for fun while doing so!) Whenever another user forfeited more than twice, I looked them up. Unfortunately, while I found some suspicious-looking users, (20 forfeits and no wins...) they could also just be confused newbies, or people with connection problems. I wanted an incontrovertible example. Dejected, I gave up.

What color are their hands now?

When you leave a battle zone, you end up in the world map- and a general chat for all users not currently playing in a zone. I noticed a Decepticon user with the insanely overblown username ULTRA DARK LORD OF DECEPTIOCNS [sic] chatting chummily with a clutch of Autobots. I looked him up.

Bingo. 11 wins, 795 forfeits. This became a thread on the discussion forum- with other users posting their own finds as the worst offender. (and some hilarious logs showing strings of forfeits in zones, and players reactions when confronted about it.) But ultimately what makes this such a successful visualization is that people got it immediately. Those who found the statistics of discussions of dummy accounts too confusing understood at once- this guy was playing for the other side.

Final thoughts

I have to admit, as a Decepticon partisan, I find this whole thing incredibly entertaining.
Last week's entry covered the economics of the game that made the Decepticons better warriors- because everyone wanted to play as the 'good guys.'
This week shows that the good guys cheat. Rampantly, innovating as they go, and to the point it's completely distorting the game. (39% of all matches now end in forfeit.) We'll make Decepticons of them yet!

On another level, it's a sobering lesson about what it takes to make people understand cheating. Statistics- no matter how inarguably correct- don't do it.
In the 2004 United States presidential election every single state with electronic voting machines that don't produce a paper receipt gave final vote tallies that differed significantly from the exit polls- every one of them was won by the Republicans, often differing more than a standard deviation from the exit polls. The odds of Bush carrying those battleground states by that margin was 664,000:1.
Since the voting machines keep no paper record there can never be a recount to prove a machine is lying. All you have are those statistical irregularities that look so so bad.
(I'm not a sheep buying into insane conspiracy theories- I ran the numebers MYSELF on the day of the election, taking the exit poll data off of CNN.com- and I saw later that day when the exit poll data was 'amended' to be more in line with the actual results without informing the public.)

What connection does all that have to the game? The Battle for the Allspark game made the bizarre choice to represent Autobots with the color blue and Decepticons with the color red. That makes the Red States the bad guys, in case you didn't get the connection.
Of course the metaphor breaks down when the good guys are the ones cheating. And strangely- cheating for no individual benefit! The game architecture is actually quite secure, every 'cheat' I detail here is an exploitation of the existing game rules.
In the end, the Autobots are cheating out of team spirit. That's almost... cute! It's not like it was no-good individual players cheating to win the weekly prizes! That would just be creepy. Fortunately, that's not possible.

...right?

(see you next post)