How would you look at ratings to determine our tastemakers?

We all have our favorite reviewers, often for specific styles, who we think are very authoritative, even if sometimes our tastebuds disagree with them.

I was just thinking about how we might surface tastemakers like this… Are these people who’ve tried beers early on that have turned out to be fantastic? People who tend away from hype generally but lit up for a particular beer that eventually others found to be world class? People who’ve sampled the most of enough of a particular variety? Is there a geographic component? Would you have to be in Germany to be an authority on German beer?

What would be your criteria?

Hard to tell. Right off the top of my head I thought of it in terms of people who rate beyond standard deviation consistently on a style, over is good for low rated styles like pils and kölsch, but I would think under is good for highly rated styles like Imperial stout (someone who isn’t likely to be blown away by the most recent hyped 4.8 pastry stout). Looking at the alone would leave us open to haters though (sorry @cheap but I don’t want you to be a tastemaker for IPA despite you being far off the standard deviation for most of the beers you rate). In order to mitigate this, maybe include a factor of high standard deviation in the style for the user, or an average rating that fits within the beer’s average rating)

For example, if 3fourths gives out more 4.8s than anyone on kellerbiers that score on average 3.0 with a standard deviation of 0.3 but also has an average score of 3.3 on the style with a standard deviation >0.3 he could be considered someone who knows the distinctions of the style.

That would be a math/stats way I guess (I’m no statistician so go easy guys).

The other simple social-media/user-input data method is to put in place a like system and count likes for a user/style and have a leaderboard.

1 Like

The like system is coming (slowly) but the stats method seems like it should yield better results. I loved your proposed use of std dev to help determine this.

1 Like

Re the like system, how is that going to be intended to be functioning? Do RBs have to be following, like how Untappd uses friends? The only follower and following RB member I had was Pinkie and well, she left.

I’d like to know this too.

If they made it followers only it’s pretty much just friends liking friends’ ratings and doesn’t really help with stats, influencers or “tastemakers” as they’re being called here, it’ll only be relevant inside those particular cliques, useless to the site as a whole. That and it would be basically copying Untappd’s “Toast” thingy and I can’t see what value that would bring.

So… I’m imagining it will be that anyone can like a rating, regardless of whether or not you are following that person. I think that would be more useful overall. But then again, it depends what’s going to be done with the data. Hopefully something useful would come of it, rather than it just being a popularity contest.

1 Like

That’s how it works on BA. No friendship/following needed but I don’t understand how a system like that, though I am in favour of it, affects stats. Can you enlighten me on how that can possibly be achieved?

Open question, not just directed at you.

Yeah I mean joet’s the best one to answer as I don’t know exactly what his intentions are for this.

Anyway I was using stats in a very vague/general sense there, not relating to any of the existing stuff, but more to what joet was saying about a method of determining “tastemakers”. The way I see it is at a very basic level the more likes your ratings receive the better judge of beer you are considered to be. So there might be a leaderboard of users whose ratings are well received, perhaps broken down by style, region, abv, those sorts of things. Kinda like a list of the most respected users.

Of course there are drawbacks to this, a good example would be reddit where a like (upvote) there is often used as an “I agree” button rather than an indicator of a good comment. So someone here might write an incredible review of a beer, but if it’s contrarian to most others, many people might not give it a like because they don’t share the same tastes.

Another problem is people who write funny/joke reviews will probably get loads of likes simply because it made us laugh and it was good entertainment.

This is probably why joet is leaning towards implementing something like Julien explained above, rather than a basic like system.

EDIT: just to clarify both systems could of course co-exist, just that Julien’s one would be better for the tastemaker thing, while likes were more for vanity/respect

Other than a leaderboard, what would you do with these newfound tastemakers? Highlight their ratings somewhere?

1 Like

Thanks. Oh yeah, I’m a firm believer in rating to style so whatever system finds itself in place, watch out, suckers! :slight_smile:

1 Like

So if I consistently like beers that most people don’t, doesn’t that suggest that my tastes aren’t necessarily in line with the masses and therefore most people will not benefit from trying the beers I rate highly? :wink:

You could based it on the groups - Like the top ciders rated by the Outciders group for example?

What about people who consistently rate lower? My 3.7s are probably a 3.9-4.2 for a lot of people just because I am habitual / internally consistent. “Stingy jerc gave it a 4.1, it must be awesome!” :slight_smile:

2 Likes

definitely not, think of 3fourths, TAR, Dedollewaitor, MartinT, and others for Franconian beer.

And I’d like to think that my knowledge of German beer will not all of a sudden decrease when I move away from here.

1 Like

Good point about stingy raters, I was trying to include them without including the haters of a particular style. I need to think some more about that :stuck_out_tongue:

1 Like

Thin ice… But I can meet you half way. When a craft brewer tries his hand at a traditional boring style I know well, and brews something modern, edgy, and crafty, and therefore isn’t brewing to style at all, the thought “this really isn’t a particularly good [boring style] at all, I’d much rather have a [boring example]” goes through my mind, followed by “the brewer hasn’t got a fucking clue what he’s doing”. My brain then just flips into beer judge mode, and I start spotting the flaws and off flavours, and the score naturally tumbles.

But a local unfiltered traditional czech lager, laced with diacetyl, I’ll love, I simply don’t care about the diacetyl when my mind is in that mode/mood, that’s just part of the scene.

2 Likes

What I used to use was the chart of score over time. That way you could see beers that had really high scores early on, then look to see who gave scores commensurate with the score years or hundreds of ratings later. You could use that to build a picture of who outperforms the algorithm I guess, but it would be nice just to have the old tools back.

Nothing is strictly defined with regard to likes of ratings so it could be anything really. What I’m imagining right now is a like system that’s open to any registered and possibly non registered user.

1 Like