Test my survey platform; super bowl prediction; 5 min.

see some guys already in, thanks

done

lets go guys. thanks so far to anyone. lets keep it going. anything over 20 or 30 would be awesome

killed it. One of them was a tie between 3 teams of i picked Denver because Payton is the shit.

haha NO love for the bears so far. i am so excited. first "live" survey of my tool and no error messages and shit

OoOo a quick lesson in german too

Well I clearly know absolutely nothing about the NFL. I'm pretty sure I might have chosen one team as most likely and then later on least likely hahaha

done

did it because its about the nfl and not boring

Fuck the Bears. Go Vikings.

no problem. as long as you said that you dont know too much, everything is fine. will calculate results for groups of people with different levels of knowledge.

maybe helps a little more actually, because most people likely will enter it only if they know their stuff.


+k and +k to all of you

thx. thats actually the point. most people tend to like these types of tasks more than simply saying whos favourite.

for example, i could let you rank the alternatives from 1-10, but thats normally very tough. but saying which is the best and the worst option is normally easy. and if you repeat it 4-6 times, the information level is very high.


bad part is the, lets just say, complex calculation necessary.

Done!

done

i hope you guys tell me if there are bigger issues/errors and so on?

i know i repeat myself but i worked way too long on this project and to sit here on the backend and have the reports ready and stuff is just awesome.


i might bother you with a few other short studies in the future. but then i will hold a draw for a small prize.

Doneski

Done! How exactly do you calculate the results from that?

Done

its a little complicated, but in easy words:

1) when you have 4 options and A is the best and D is the worst, you know the following relations:

A > B
A > C
A > D

B > D
C > D

so basically you just give A the value 3, B & C the value 2 and D value 1.

2) you try to receive a few hundreds of these comparisons and write them into a database

3) then you maximize a function via maximum likelihood prediction. basically you ask, which "utility" value for each of the options gives the best fit given the current inputs?



idk if that made sense, will definitely post results tomorrow or after i reach 30 respondents though

^^ obviously you ESTIMATE the parameters via maximum likelihood. and in order to do that you MAXIMIZE the function.

i worded that pretty bad up there.


thanks so far guys, were at 20!!!!

I know nothing but I did it. No tom wallisch?

Fuck the Bears, fuck the Viqueens, go Packers.


Found myself picking the Texans a LOT.

done

Nah we all know its going to come Down to either falcons or texans vs patriots or giants

How can the Texan play the Patriots if they are in the same conference? Same goes for Falcon/Giants.

came to question 3..... sorry bro i have no idea who can win this... "damn europeans"

so far very interesting, intermediate results


people with sub average knowledge from best to worst:
pats, texans, packers, broncos, giants, 49ers, ravens, bears, falcons


the above average ones from best to worst:
texans, broncos, pats, giants, ravens, 49ers, falcons, packers, bears

no problem. but you still could do it randomly. that probably fits your answer before that you dont have much clue about the NFL

done

ok donzo, thought the green bay......somthing... would lose cause i have no clue what city they are haha

28 completetions guys, keep em coming how long you want. i am a happy camper anyways from now on

I was kidding but those are the top 4 teams in the nfl

your da man

keep em going. really appreciate it. if not, no problem. by far the best test data i received.

for over a year now i always entered the data myself, asuming different "personalities" in order to acquire specific results for different groups.

this way is 10 times faster and the data is easily 20 times better.

thanks a lot


as i said, if you want, keep going, if not, no problem. i know how much i hate this constant "yo 5 minutes survey +kkk" shit, but i really really appreciate any help

yeah makes sense! I'd definitely be interested in seeing the results

hmmm this is interesting. Done. Threads for later results

so. i know its been pretty late but i almost forgot about that. was really stoked back then to see my project up and running and definitely saw how certain things needed to be changed and so on.

the study was about comparing 4-5 randomly selected nfl teams / super bowl favourites (out of 9) and name the most likely super bowl winner and the least likely one out of the given teams.

-----------------------------------------------------

the results are in the pdf here
( http://eagl-e.com/media/doc.pdf ). the results are always split into the overall group "newschoolers", the group who thinks they know more and those who think they know less about the NFL. the last page is the most important.

47 people participated (group newschoolers), 27 classified themselves as being above average informed and 19 under average. 1 person quitted before the question regarding the knowledge.

basically, most participants were around 18 years old. no significant difference between the groups.

then the question with the knowledge is treated.


the last page is most important. this is the main part with the comparing questions. the first table shows simply how often a person mentioned either team as being best or worst, splitted into groups.

so, because a good team can be randomly shown with very bad teams more often (unless you run it 10000+ times) as another good team, the results are skewed. so a function is minimized, asking the question, which utility parameters explain the observed decisions with the highest degree of probability? these parameters are found in the second table of the last page. taking exp(utility parameter) would allow you to simulate "markets" (in other settings, its not really important in this case, ask if you want to know more) and to calculate which products consisting of individual objects (teams) would be the best (again, not easy to think of an easy application of that, maybe of a combo-bet or something on which 2-3 teams win the superbowl).



overall, in beginning of december, ns overall was highly favouring the patriots and texans. ravens and niners were in the lower middle of the pack, while falcons and the bears were considered to be the worst.

the results between the groups were fairly consistent with what one would expect.

but overall, NS FUCKING FAILED TO PREDICT THE SUPERBOWL TEAMS! the experts had the 49ers as the fourth best team, while the ravens were considered to be the second worst. texans were in this group the best, closely followed by the pats.


thanks a lot to anybody who participated and i hope it was slightly interesting.