Is there any evidence that it actually uses Markov chains? I don't see any mention. The title here might as well be made up to troll HN?
I did write it, it does use Markov chains, and the whole thing should be self evidently a troll.
For example, it is not necessarily trying to beat you. This was inspired by an experience with https://luduxia.com/whichwayround/ where I found most traffic came from people trying to break it, until one day it hit a mailing list and exploded.
Why not make it a show HN?
I have done other related ones before, and honestly find "Show HN" (understandably) involves a level of expected seriousness that other posts might not. Consequently I prefer to leave that only for things which are more serious in intent.
This one I updated the game a bit, submitted a link, and it went nowhere when I posted it some days ago. I woke to a Linode alert of sustained outgoing traffic and it turns out the post had been resurrected.
I was also wondering that - is the poster the author?
In any case Markov chains are a simple and elegant mechanism for making predictions, easy enough to understand so that even school kids at grade eleven (which is where probability theory got introduced when I underwent it) can follow the popular "weather forecasting" tutorial using MCs.
Onwards, Hidden Markov Models (HMMs) add to MCs a layer of hidden states and associated emission probabilities. To learn these, check out Lawrence Rabiner's beautiful tutorial: https://ieeexplore.ieee.org/document/18626
Apart from their simplicity and mathematical elegance it is remarkable how little data and electricity these models require to do a good job.
Good question. I'm at a 75% win/loss ratio by picking rock every single time. Doesn't appear to be learning in this session, anyway.