Haven't seen a good measurement/analysis thread in a while.

I usually pass over those because theres alot of science words that my PE and Health degree didnt teach me. Plus big words and me drinking really dont mix.

 

I would love to do lots of things like those, the problem is that the time it would take to harvest the data from BA is astronomical. If there were some way to click a button and get a CSV of the raw data for a beer then it would be worth doing, but as it stands it just isn't.

 

Pfft, have you looked at our national discourse lately? Data doesn't mean anything to most people.

 

ive got a good one. forgive me if someone just did a post on brewer lot-to-lot variation.

i like to see what other people rate for beers that i like. two of my favorite beers (that arent typical "BA" beers for lack of a better term) are Sam Noble Pils and Wolaver's Oatmeal Stout.

in looking through reviews, i find a lot of variation in likes/dislikes for these beers... but here's the thing. the likes/dislikes come in strings.

take the Oatmeal Stout: http://beeradvocate.com/beer/profile/119/14172

here are avg scores for this beer over time periods:

10/15/11 - 11/17/11: 4.28
11/24/11 - 12/21/11: 3.83
12/25/11 - 1/18/12: 4.22
1/21/12 - 2/16/12: 3.74
2/22/12 - 3/16/12: 4.25
3/25/12 - 5/1/12: 3.45

in the high periods, there were no scores below a 4. in the low periods there were no scores ABOVE a 4. i have a really nice graph of this which i cant figure out how to get posted up here.

the two things I get from this: a) it takes Wolaver's about four weeks to sell a batch of this beer from fall until spring; b) they seem to be having some trouble with lot-to-lot consistency. some lots are world-class, some are barely above average.

my guess is you could find this trend for several brewers, and it speaks to the difficulty in producing craft beer consistently from batch to batch. we talk about great beers, and great breweries, but it is easier said than done to make that great beer great in every single batch. consistentcy is what makes a great brewery truly great.

i think this is valid data... or did i just make it all up? talk amongst yourselves.

 

What are the variances on those numbers? It could easily be statistical noise if your sigma is like 0.5 (which is high, but depending on review #'s possible).

 

average stdev for each group (except the last one which had a 4, 3.05, and 2.65 rating) was 0.17, which means there's about 3SDs between each grouping. i tested this using software i have and it shows the two groups are statistically different. this cant be noise.

 

You don't also happen to have software that shaves the leftover fractions of a cent off of every transaction and then deposits that money into a secret bank account, do you?

 

i do... but i usually forget about some small detail that screws everything up too. no talent ass clown.

 

I wonder how much of that is reviewers having doubts about their own ability to review for themselves. If people see that the last 10 reviews or more are all in a similar range, then the reviewer might be thinking that range is what the beer "is supposed to be" reviewed at. So their review puts the beer at the same range as the others so as to appear knowledgeable via consensus.

 

maybe - except the number of days between switches from good to bad (and back) are too specific. days between cycles are: 40, 31, 27, 32, 32. the phenemenon youre describing would show switching more randomized.

 

Wouldn't that phenomenon actually prevent any switching at all? If everybody is following the average, or the last few reviews, the averages would never change and the last reviews would always be similar. Unless of course you'd have for a while mostly people not trying to be the same as the previous and then mostly people trying and so on, but that would be too random to make a difference.

 

Is nickd still around?

 

I disagree. It's accurate data that doesn't mean anything to most people.