{"assets":[],"author_link":"author\/pansapiens\/","author_name":"pansapiens","cat":"LD #29","categories":["LD #29"],"comments":[],"epoch":1410796740,"event":"LD29","likes":4,"metadata":{"p_key":"81980","p_author":"pansapiens","p_authorkey":"0","p_urlkey":"292712","p_title":"LD29 predictions \u2013 how well did I do ?","p_cat":"LD #29","p_event":"LD30","p_time":"1410796740","p_likes":"4","p_comments":"0","p_status":"WAYBACK","us_key":null,"us_name":null,"us_username":null,"event_start":"1408665600","event_key":"24","event_name":"LD30"},"source_url":"2014\/09\/15\/ld29-predictions-how-well-did-i-do\/","text":"

Some of you may remember during LD29 I had a crack at predicting<\/a> some of the highest rated entries using machine learning. Specifically, I dumped out a list of entries predicted to be in the top 10th percentile based on Overall score, using only words in the comments (adjectives and adverbs) that were predicted using K-nearest neighbour classification based on a model trained with LD15-LD28 comments and scores. Phew. Deep breath.<\/em>. I never did the followup post to look at just how many I got correct.<\/p>\n

It turned out in my rush to do something before judging ended last time, I\u2019d made some errors when training the model (mostly to do with dealing with Overall(Jam) vs. Overall scores and unrated entries). Embarrassing<\/em>, but hey, it\u2019s a work in progress, Caveat emptor<\/em> :). As a result, this list is different to the original one I posted (with some overlap) .. however since ratings were available, I could see which of the predictions were actually correct:<\/p>\n

<\/span><\/p>\n

Correct predictions, LD29 compo and jam combined (128):<\/strong>\nThese were correctly predicted to be in the top 10th percentile.<\/p>\n

Hypnohustla<\/a>, DannyG59<\/a>, ApoorvaJ<\/a>, Alex Rose<\/a>, Zarkith<\/a>, Snowflower<\/a>, Tomski<\/a>, rubna<\/a>, Overkill<\/a>, PlasterPhantom<\/a>, snapman_GT<\/a>, Lama<\/a>, ix<\/a>, Robinerd<\/a>, Eirikir<\/a>, Gabriel<\/a>, GamerTaters<\/a>, Mental Atrophy<\/a>, Solifuge<\/a>, GertJohnny<\/a>, xdegtyarev<\/a>, ilovepixel<\/a>, TeamInCharge<\/a>, Pierrec<\/a>, bitserum<\/a>, petey123567<\/a>, Nition<\/a>, Cambrian_Man<\/a>, flrn<\/a>, quill18<\/a>, GarethIW<\/a>, 01101101<\/a>, IAmSpencer<\/a>, Lyje<\/a>, _Rilem<\/a>, voidqk<\/a>, HacksawUnit<\/a>, SaintHeiser<\/a>, Xaychru04<\/a>, Mikhail Lyubimov<\/a>, Maschinen-Mensch<\/a>, 7elephants<\/a>, Andrew Shouldice<\/a>, Elis\u00e9e<\/a>, savethejets1<\/a>, baykush_figtree<\/a>, orangepascal<\/a>, geekdrums<\/a>, Helopol<\/a>, Danman9914<\/a>, Frump<\/a>, TijmenTio<\/a>, Ellian<\/a>, Hemuuuli<\/a>, SecondDimension<\/a>, Mase<\/a>, Yword<\/a>, bvanschooten<\/a>, Double Zero One Zero<\/a>, oatsbarley<\/a>, TheGreenTie<\/a>, DragonXVI<\/a>, zillix<\/a>, Benjamin<\/a>, TobiasW<\/a>, Evilion<\/a>, ZYXer<\/a>, Mr. Jif<\/a>, quickfingers<\/a>, KickBack<\/a>, j_peeba<\/a>, Ditto<\/a>, Dreauw<\/a>, dukope<\/a>, Jezzamon<\/a>, Wolve<\/a>, TeamPurpleDolphins<\/a>, Raiyumi<\/a>, Knighty<\/a>, nka<\/a>, StudioWolfox<\/a>, deepnight<\/a>, Draknek<\/a>, Miltage<\/a>, arkeus<\/a>, CryoGenesis<\/a>, UltimateWalrus<\/a>, Turnipnose<\/a>, DrPrettyPatty<\/a>, 01010111<\/a>, Ludonaut<\/a>, Oddly Shaped Pixels<\/a>, Bekokstover<\/a>, oostap<\/a>, Pixlexia<\/a>, lajos<\/a>, jsmars<\/a>, Toki Anholmes<\/a>, Dr_V<\/a>, Zednaut<\/a>, DDRKirby(ISQ)<\/a>, Romko Pidstryhach<\/a>, Make A Game<\/a>, torcado194<\/a>, Managore<\/a>, SteveSalmond<\/a>, NickZangus<\/a>, EpaceGames<\/a>, rpgwhitelock<\/a>, Jools64<\/a>, TeamFlare<\/a>, Cake&Code<\/a>, jahlgren<\/a>, mark.goetz<\/a>, radmars<\/a>, InfectionTeam<\/a>, greysphere<\/a>, BumbleBirds<\/a>, Diptoman<\/a>, Flai<\/a>, Almost<\/a>, Catman<\/a>, _adamturnbull<\/a>, Valandre<\/a>, Split82<\/a>, lucienpro<\/a>, SimonLarsen<\/a>, Aomeas<\/a><\/p>\n

Incorrect predictions, LD29 compo and jam combined (162):<\/strong>\nThese were predicted to be in the top 10th percentile, but weren\u2019t.<\/p>\n

gigitrix<\/a>, X54321<\/a>, SonnyBone<\/a>, elefantopia<\/a>, SunShiranui<\/a>, brofist56489<\/a>, Beavl<\/a>, d__adee<\/a>, hampa<\/a>, eteeski<\/a>, Otrora Interactivo<\/a>, kiririn51<\/a>, Agecaf<\/a>, Indie Squid<\/a>, halfmage<\/a>, Snoother<\/a>, Photon<\/a>, Coin Flip Games<\/a>, TheElephantSeal<\/a>, Glint Games<\/a>, Gap<\/a>, haloflooder<\/a>, Tuism<\/a>, ToOB<\/a>, joe40001<\/a>, TowelKing<\/a>, Christina Antoinette Neofotistou<\/a>, cwkx<\/a>, Ithildin<\/a>, javierecf<\/a>, Zelun<\/a>, mr_tag<\/a>, conormn<\/a>, Robotic<\/a>, tylerb1<\/a>, jackolondon<\/a>, enderosc<\/a>, NeiloGD<\/a>, Liens<\/a>, zn01wr<\/a>, Dualhammers<\/a>, Jakub Koziol<\/a>, Sparrow<\/a>, Lerc<\/a>, tacs<\/a>, teambrookvale<\/a>, MrShoestore<\/a>, rico.albe<\/a>, Nik Sudan<\/a>, mir007<\/a>, John Drury<\/a>, eufrik<\/a>, Chinchilla<\/a>, Secret_Tunnel<\/a>, Colorvade<\/a>, thinkster<\/a>, grozamorei<\/a>, Tweak<\/a>, admung<\/a>, allurious<\/a>, Kino333<\/a>, praporomsk<\/a>, Akz-<\/a>, Woftles<\/a>, NoahC_<\/a>, AhNinniah<\/a>, Bruno Massa<\/a>, Cawrtz<\/a>, BillToWin<\/a>, Striwx<\/a>, ErikU<\/a>, SalamiChild<\/a>, CitrusPossum<\/a>, lightsoda<\/a>, FGM<\/a>, NiceAlexanderAS<\/a>, nddrylliog<\/a>, Mapboy<\/a>, Figglewatts<\/a>, goerp<\/a>, Thomas Ingham<\/a>, abrie<\/a>, malmazuke<\/a>, Zekronz<\/a>, harleylaurie<\/a>, wimnea<\/a>, Mlle Eole<\/a>, Recursiveanomaly<\/a>, berareu<\/a>, AgentParsec<\/a>, Noah Ratcliff<\/a>, Lrnk<\/a>, Erfeo<\/a>, Rahazan<\/a>, Magdev<\/a>, madjackmcmad<\/a>, Joe Williamson<\/a>, DarkMeatGames<\/a>, Stuffie<\/a>, angelk<\/a>, twelveplusplus<\/a>, Colthor<\/a>, camlang<\/a>, magicspark<\/a>, DNA Yarn<\/a>, LazyParia<\/a>, Zampi<\/a>, suurin<\/a>, Linus Lindberg<\/a>, Joshua_Burr<\/a>, eastes<\/a>, BlazeCell<\/a>, Zutty<\/a>, myachin<\/a>, Sodaware<\/a>, WhiteWolf93<\/a>, TheMarvellousTeam<\/a>, PriorBlue<\/a>, Christophe<\/a>, Hyoga-3D<\/a>, StoneMasters<\/a>, mcapraro<\/a>, iossif<\/a>, Exeneva<\/a>, Kyatric<\/a>, Dark Arts and Sciences<\/a>, Typedeaf<\/a>, Balloonsfor600<\/a>, tenpn<\/a>, ShivanHunter<\/a>, Nuclear_Hammer<\/a>, TheH<\/a>, td1801<\/a>, Dailydo<\/a>, WetDesertRock<\/a>, DarkUser<\/a>, Benn<\/a>, mortus<\/a>, sockfolder<\/a>, lief<\/a>, jtpup0<\/a>, petrih<\/a>, MintArcade<\/a>, LTyrosine<\/a>, sirdorius<\/a>, Nizar89<\/a>, FistBumpGames<\/a>, archaeometrician<\/a>, jurnacsr<\/a>, Pent<\/a>, Cherno<\/a>, brother.byto<\/a>, Thijsku<\/a>, Muu?<\/a>, Kevlanche<\/a>, louis.denizet<\/a>, SuperIzzo<\/a>, geounknown<\/a>, BuKaneerz<\/a>, daandruff<\/a>, BrothersT<\/a>, Detocroix<\/a><\/p>\n

Summary:<\/strong> It looks like I can use the comments on entries to guess if any one entry will be in the top 10th percentile with about 50\/50 chance of getting it right.<\/p>\n

Next post \u2026 LD30 predictions using the same method !<\/p>\n

Tags: predictions<\/a><\/p>","time":"September 15th, 2014 3:59 pm","title":"LD29 predictions \u2013 how well did I do ?","title_was_empty":false}