🏅 Top traders

He has been banned so can probably be resolved now lol

@Lorelai You are still able to rate his resolution though.

Trust me, I don't like waiting, but I promised to give 3 days to rate him.

Where’s my mana 😭

turns out the answer is "user was banned for trying to screw with the market"

What if it is below 1? 😅

@4fa

Good point

I think it wouldn't be too controversial to resolve to the lowest rung. I am biased, though.

@lumi How is this possible? You can only rate 1-5 stars no? Not zero.

@Quroe I will resolve this in good faith to 1-2 Stars, but I need to wait 3 days. This is the spirit of the market.

The reason is the following: I was not aware that you could be rated below 1 star. I don't think anyone was. I would love for a @mods to explain how the star rating system works. I didn't think this was possible.

@HillaryClinton "Manifold's weighted rating" must have some wonky negative weights. 😁

@HillaryClinton Sounds good to me. Just wanted to make sure. 👍

I imagine it's mapping the user to a scale of where they sit among all users. Like, a scale of 0 to 100, where 0 on the scale is mapped to 0 stars and 100 on the scale is mapped to 5 stars.

I'm talking out of my butt and this is all conjecture.

@4fa I lost money on this market, so I don't think I'm biased. But if there is considerable controversy I will open it for a vote. That being said, I am heavily leaning towards that being the fairest option.

Yes, wonky weights.

@4fa That rating is some "lower bound of 95th percentile confidence interval" (https://github.com/manifoldmarkets/manifold/blob/0c3e2a968560a949411dd61f460f6f3d79febe62/backend/supabase/functions.sql#L349-L390).

Likely predicting "is it probable that a next market by this user will also be good, excluding outlier reviews?".

@AnT Ah, nice! I first thought they did not think of the bounds, but they thought of the bounds – just incorrectly. They "squash with sigmoid, multiply by 5", but maybe they should use the sigmoid, multiply by 4 and add 1.

@AnT I took a closer look.

It doesn’t appear to explicitly exclude outliers, but it does simplify the data by quantizing everything into two buckets ("4-5 stars" and "1-3 stars"):

```
positive_counts AS (

SELECT 5 + COUNT(*) AS count FROM avg_ratings WHERE avg_rating >= 4.0

),

negative_counts AS (

SELECT COUNT(*) AS count FROM avg_ratings WHERE avg_rating < 4.0

),
```

@Bayesian Do you have any stats on average question rating? Like if a user is 3.00 stars, what percentile does that put them in for user ratings?

@HillaryClinton No mb, but seems mostly predicted by ur number of overall reviews.c most people with that bad rating prolly have only like one bad market anf nothing else. Not sure tho

i would suspect he gets a decent rating here, since ratings get bad during contentioud resolution, wjich this isnt? Excrpt timingwise i spose

@Bayesian He only has 4 other ratings, so I assume whatever this will dominate his rating. (not sure how the weighted system works though).

My theory is a lot of people are not happy, as every news organization and website has called the winner.

That being said, even if I'm right, it could be like Uber where anything below a 4.00 is really bad. People tend to be generous with their reviews. I wish I knew the stat, but if he ends up with anything below 4.00, that's definitely bottom 20% or so.

@HillaryClinton Yeah like 90%+ of reviews are either 5 or 1.

if I might do some publicity for my Peter Njeim market: https://manifold.markets/nikthink/peter-njeims-canadian-election-mark

you can comment your Peter Njeim markets on my market too.

@nikthink I like it! Good market.

If his overall rating is exactly 4 stars, which option does this resolve to?

@TimothyJohnson5c16 I will use the "Unweighted Average" as the tie breaker if it is exactly 4.00 Stars. The one circled in red is the tie breaker.