Sunday, December 16, 2007

A Late-Night Followup on Bots and Timing Tells

Re-reading the post on the 'sillysal' situation as detailed expansively on 2p2, I started thinking about the nature of timing tells themselves, apart from whether or not straight-up bot usage by that account is what happened in that instance.

Let's head off on another tangent. It's no surprise that timing tells are one of many weapons used by sites to unmask bots, but I'm gonna go way-y-y out on a limb here. Blame it on overwork and a lack of sleep, but in giving some concepts a bit of thought, I've convinced myself of one thing: Properly developed, timing-tell algorithms would, by themselves and independent of any other evidence, be sufficient to uncover almost all instances of bot usage where the bot actually made the betting action.

Why this is, I don't want to say at this point, though it's not something as simple as instantaneous raises or folds or whatever. It's made a bit more challenging by the fact that Internet lag times must be accounted for, but there are still some things that humans do that machines imitating them do not.

As time allows, I may do a bit more digging on the concept. In a previous career I became enough of an expert at reverse-engineering a mechanical process designed to emulate a certain human expectation that I was called as a witness in a federal case, and something I learned in that career also applies to this topic. I'm also sure it's nothing new to the experts already working on these matters --- they absolutely already know what I've just realized. In a way, there's a realization here that bots have less of a chance of taking hold and running for lengthy periods than most people believe... given that sites devote enough time and resources to uncovering them.

Color me dumbassed --- but that's a human quality, too.

No comments: