[{"data":1,"prerenderedAt":137},["ShallowReactive",2],{"project-8774473":3},{"id":4,"name":5,"line":6,"line_source":7,"title":8,"url":9,"domain":10,"maker":11,"posted_at":12,"year":13,"points":14,"comments":15,"rank_in_year":16,"shown_in_year":17,"top":18,"tier":19,"category":20,"subcategory":21,"format":22,"keywords":23,"reception":27,"sentiment":28,"stance_mix":29,"scored_comments":30,"aspects":31,"strip":32,"post_text":33,"what":34,"said":52,"maker_in_thread":80,"shape":84,"grid":88,"attempts":97,"same_week":98,"maker_shows":129,"maker_hidden":130,"reader_version":131,"card":132},"8774473","epsilon greedy","The problem with the epsilon greedy method","title","Show HN: The problem with the epsilon greedy method","https:\u002F\u002Fgithub.com\u002Fcrobertsbmw\u002FEpsilonGreedy","github.com","crobertsbmw",1419022742,2014,22,3,1194,2272,53,"shown","learning_science","bandit algorithm simulation","other",[5,24,25,26],"a\u002Fb testing","multi-armed bandit","regret","quiet",null,{"positive":30,"neutral":30,"mixed":30,"critical":30},0,{},[30,30,30],"I decided I wanted to roll my own AB testing app for Django (https:\u002F\u002Fgithub.com\u002Fcrobertsbmw\u002FRobertsAB) when I was finished, I came across this:\n\nhttp:\u002F\u002Fstevehanov.ca\u002Fblog\u002Findex.php?id=132\n\nWhich is a very convincing article on why AB testing sucks and with a few extra lines, you can improve your algorithm to select the best test so you never go back and update your code (yeah right.)\n\nI then thought, how many tests does this thing have to run to truly figure out which is best?\n\nI made 4 tests with probability of success equalling 1\u002F2, 1\u002F4, 1\u002F5, 1\u002F6  and found that for this algorithm to settle ",{"problem":35,"audience":28,"pricing":28,"motivation":41,"time_to_build":28,"stack":45,"openness":50,"stage":51,"team_size":51},{"value":36,"source":37,"evidence":38},"Unclear how many trials epsilon greedy needs to find the best option versus A\u002FB testing","maker",[39],{"id":40},"post",{"value":42,"source":37,"evidence":43},"Wanted to build an A\u002FB testing app for Django and wanted to test a bandit algorithm first",[44],{"id":40},[46],{"value":47,"source":37,"evidence":48},"Python 3",[49],{"id":40},"open_source","unknown",{"praise":53,"objections":61,"requests":69,"questions":70,"alternatives":71},[54],{"value":55,"source":56,"evidence":57},"Useful comparison and referral to bandit reading","commenter",[58],{"id":59,"by":11,"text":60},"8775583","awesome read. This would have satisfied my curiosity and probably saved me an entire day of messing around. Thanks.",[62],{"value":63,"source":56,"evidence":64},"Epsilon greedy and A\u002FB testing solve different problems",[65],{"id":66,"by":67,"text":68},"8774639","marketforlemmas","Interesting comparison but, by my understanding, epsilon greedy and A\u002FB testing do not solve the same problem. Epsilon greedy is a method for minimizing regret, that is the…",[],[],[72,76],{"value":73,"source":37,"evidence":74},"Standard A\u002FB testing",[75],{"id":40},{"value":77,"source":37,"evidence":78},"Epsilon greedy bandit method",[79],{"id":40},{"comments":81,"answered":30,"key":82},1,[83],{"id":59,"text":60},{"commenters":15,"max_depth":85,"maker_replies":81,"first_comment_mins":86,"lifespan_hours":87},2,23,5,[89,91,95],[66,30,30,90,67,30],"other\u002Fneutral\u002Fidea",[92,81,30,93,94,30],"8774600","other\u002Fneutral\u002Fdocs","bcbrown",[59,81,30,96,11,81],"maker\u002Fpositive\u002Fdocs",[],[99,104,109,114,119,124],{"id":100,"name":101,"points":102,"category":103},"8766865","Wysihtml",215,"dev_tools",{"id":105,"name":106,"points":107,"category":108},"8787733","Startup Fighter",117,"games",{"id":110,"name":111,"points":112,"category":113},"8787783","Espion",109,"web_saas",{"id":115,"name":116,"points":117,"category":118},"8774257","ping.gg",95,"infrastructure",{"id":120,"name":121,"points":122,"category":123},"8782320","reread",81,"consumer_social",{"id":125,"name":126,"points":127,"category":128},"8790970","Chicory",76,"languages",4,false,"v2",{"id":133,"name":5,"line":6,"ai":130,"cat":20,"year":13,"points":14,"comments":15,"rec":27,"s":28,"top":18,"kw":134,"maker":11,"domain":10,"tier":19,"attempts":30,"strip":135,"url":9,"mix":136,"praise":55,"objection":63,"sub":21},8774473,[5,24,25,26],"000",[30,30,30,30],1794036526295]