Getting Started Reverse Engineering the Diablo 3 Auction House

larrystechblog.tumblr.com

4 points·arange··4 comments

Once you have all this data, simulate mouse clicks on the next and previous buttons to paginate and get more data. It’s all a tricky process, very hacky, but you’re getting really valuable data.

If you go all the way to simulate mouse clicks, why not OCR the prices off the screen? I guess that's somewhat simpler (I used to do it for a poker bot, there's a few open source screen scraping tools available). Maybe there's more information available in memory than is shown to the user, of course.

I'm not sure how easy it would be to do OCR due to the weird font the game uses. The numbers seem easy enough to read though (Way better than the ones used diablo I/II), and I guess that's what count. If they changed the font to the old one it would probably not work though, since the '5' and the '6' characters where pretty similar.

Well, the advantage that screen scraping has over paper OCR is that the font is completely deterministic and there's (usually) no degrading effects. Even if '5' and '6' are pretty similar, there must be some pixel that can be used to distinguish them.

What could make it more tricky is if alpha transparency is used for both the foreground and background. Still, the complexity isn't anywhere near "real life" OCR.

This guy is a badass.

Reply on news.ycombinator.com