misterjerry@seznam.cz's Replies

Forum Replies Created

Viewing 1 post (of 1 total)
  • Author
    Posts
  • in reply to: Suggestions Forum #28576

    Hello,

    I’ve played BB for 400 hour so far. Sometimes, I went to reddit to see recommended builds for some characters but most of the times I didn’t learn anything from all the noisy comments suggesting this and that. Unfortunately, I don’t have time to spend tens of hours to asses if the recommended builds work (funny to say that after playing for that long). Eventually, my learning curve plateued and I abandoned the game.

    Recently, I came across a book Noise from Kahneman who, in chapter 10, cites a study [Clinical vs Statistical Prediction from Paul Meehl](https://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.693.6031&rep=rep1&type=pdf) which gave me an idea.

    The study starts with the similar premise as the reddit recommendations. Essentially, some committee assigned points to job applicants and some other group of people tried to judge if those people succeed or not. After a year, the judgments were evaluated revealing that the they are more less garbage.

    I’d like to use Battle Brothers to reproduce a study in the following way:
    – Generate recruits with various stats and backgrounds (you can do that easily)
    – Let people somehow decide if this recruit succeed or not (this will require some kind of front-end enabling interaction)
    – Perform the experiment (That’s open to debate. So far, I can imagine having 12 brothers with pre-defined skills from which we will build enemy team by cloning all 12 of them while “our” will constitute of just 11 of them + the chosen recruit. We can see that the teams are almost identical so we can expect around 50/50 chance of winning but the scales will be slightly tipped by the stats of the evaluated recruit. If we run such situation enough times we will know if this recruit improves the chances or not. I expect to simplify various aspect of the experiment to reduce complexity.)
    – Measure the results to see if the study reproduces in this environment.

    I would like to participate only as a hobbyist. Maybe we can write some paper on this topic… But for you, if productionalized properly, you can use it to community-building feature to kickstart different kind of discussion and/or create an in-game feature that will allow players to evaluate the recruit (this is far beyond the scope of this experiment I believe).

    I understand if this is not interesting. However, if your are at least a bit interested, I’d like to get in touch and discuss the possibilities how could I help you to pull this through – I understand data and I am skilled SW engineer so maybe I could do the hard work.

    Should you be interested, please let me know.
    Regards,
    JK

Viewing 1 post (of 1 total)