Archive for 2008


Thursday, November 13th, 2008

pirate.pngyes, that’s the approximate piracy rate for the pc version of world of goo.  we casually mentioned that number recently and the news seemed to propagate far and wide, so we’d like to follow up with some more details, for those interested.

first, and most importantly, how we came up with this number:  the game allows players to have their high scores reported to our server (it’s an optional checkbox).  we record each score and the IP from which it came.  we divided the total number of sales we had from all sources by the total number of unique IPs in our database, and came up with about 0.1.  that’s how we came up with 90%.

it’s just an estimate though… there are factors that we couldn’t account for that would make the actual piracy rate lower than our estimate:

  • some people install the game on more than one machine
  • most people have dynamic IP addresses that change from time to time

there are also factors that would make the actual piracy rate higher than our estimate:

  • more than one installation behind the same router/firewall (would be common in an office environment)
  • not everyone opts to have their scores submitted

for simplicity’s sake, we just assumed those would balance out.  so take take the 90% as a rough estimate.

this is in line with a previous estimate by russell carroll (director of marketing at reflexive) for the game ricochet infinity.  russell estimated a 92% piracy rate and i found his analysis quite interesting (check it out here if you’re curious).  one thing that really jumped out at me was his estimate that preventing 1000 piracy attempts results in only a single additional sale.  this supports our intuitive assessment that people who pirate our game aren’t people who would have purchased it had they not been able to get it without paying.

in our case, we might have even converted more than 1 in a 1000 pirates into legit purchases.  either way, ricochet shipped with DRM, world of goo shipped without it, and there seems to be no difference in the outcomes.  we can’t draw any conclusions based on two data points, but i’m hoping that others will release information about piracy rates so that everyone could see if DRM is the waste of time and money that we think it is.

nerd.jpgUPDATE (and nerd alert): a lot of smart people have been questioning the accuracy of our 90% estimate, and with good reason, it’s a very rough estimate and the measurements are flawed. so we did some more digging to see if we might have missed the mark by a significant amount.  here’s what we found:

  1. based on the number of unique IPs and unique player IDs, we found that on average, there are 1.3 unique IP addresses per player (there is 1 player id for each profile created on any installation that submits scores to our server)
    76% of players have contacted the server from 1 IP
    13% from 2 IPs
    5% from 3 IPs
    3% from 4 IPs
    1% from 5 IPs
    1% from 6 IPs
    1% from more than 6
    this tells us that the dynamic IP issue is a relatively small factor in this calculation
  2. we also looked at how many players IDs were created (rather than used) from each IP address.  given that the vast majority of player IDs are associated with only a single IP, this is a fairly accurate measure of how many profiles the average user created.  on average, a player has 1.15 profiles per installation.

when we take the total number of player IDs (which is smaller than the number of unique IPs from which leaderboard entries came) and divide it by 1.15 (the average number of profiles per installation) the number of estimated unique installations drops by about 35% as compared to the estimate based on unique IPs.  let us further say that the average user installs the game on 1.25 computers with different IPs (i.e. not behind the same router), which i think is a high estimate.  that lowers the estimated unique installations by another 20%.  after factoring both of these in, the piracy rate would still be 82%, and we should keep in mind that this number doesn’t include those who never opted to submit scores to the leaderboard (it’s an option that’s off by default).  so while it’s possible that the actual piracy rate is lower than 90%, it’s unlikely that it’s significantly lower.  2d boy hopes this satisfies the more rigorous number crunchers out there :)

oh, and yes, these numbers are exclusive of the demo… those scores are submitted to a different server entirely.