Rating to Rank Conversion Visualized

I haven’t seen anyone do this one yet so here it is. Chart showing rank spacing over Glicko2 rating.

20 Likes

I can’t imagine anything other than a Gothic font for this now. :+1:t2:

9 Likes

I was going to do this, but you beat me to it! :smile:

(For anyone wondering, the formula is

850 * e^(0.032r)

e=2.718281828459045…

r=rank, 30k=0, 10k=20, 1k=29, 1d=30, 9d=38)

(Also I made a table for this over at Glicko Rating Changes)

4 Likes

It is probably not Gothic but rather Old English (English blackletter typeface).

4 Likes

To my untrained eyes it looks like Gothic and I stand by my previous statement. :blush:

3 Likes

I am betting “Manuscript”

2 Likes

This plot doesn’t tell much to me. The curve isn’t linear nor much curved.
I just see that there isn’t room for 40k or so.

3 Likes

Yeah that part weirds me out. Glicko2 ratings drop as low as 620 (mine when I started, for instance), which according to the conversion formula above translates roughly to 40k.

2 Likes

3 Likes

Actually, they can drop down to a minimum of 100

2 Likes

I meant in practice, not in theory. :slight_smile: I knew nothing about go except the rules when I started and lost every game. I dont think Ive ever seen anyone with a lower rating than 620.

4 Likes

I’ve seen troll accounts that only wanted to dive as low as they could, but the lowest real ranking I’ve seen is 664…

2 Likes

Behold my badge of shame above. :smiley:

7 Likes

Yeah, the idea was to show how further apart stronger ranks are placed compared to the weaker ones. Numeric values aren’t very good at conveying that. Realistically though, it indeed doesn’t show much. But it’s pretty to look at, I think!

There’s plenty of room in the negative rank values there (aka slums). The graph is a bit misleading here.


Anyway. Let me ask you what happened here? Looked like an ordinary beginner losing a dozen or so games against a bot. Then boom, super low rating out of nowhere? How is this even possible?
https://online-go.com/player/662573/

5 Likes

I’m fairly sure the confusion about how “overall” is calculated is long-lasting. That guy has an overall of 175, but the others are 600 to 900. The only semi-explanation I got from a search was from @yebellz back in 2015:

These above two types of discrepancies can be explained if players have had a general trend of improving or declining (or at least just a general winning or losing streak) over a period of time, while also playing a mix of all three game types. The overall rating will have moved much faster since it has roughly three times as many adjustments as the categorical ratings. In these cases, its just that overall rating is moving quicker toward the more accurate value, and all four should eventually stabilize as they converge toward the player’s true rating. However, small discrepancies are always still possible from winning/losing streaks.

3 Likes

175 is a caching error in the rest API. If you take a look at the rating graph, you can see the overall rating is 788. It should fix itself after another game.

3 Likes

Could you explain a bit more? What would caching error under these circumstances mean?

3 Likes

I don’t know where the 175 comes from, but it’s not in the rating history.

The rest API holds a copy of the current ratings to speed up things. I observed the cached value being some games behind, when games were annulled. It always corrects itself when the rating is updated after the next game.

2 Likes

Indeed, it’s not in the rating history. The last game is also not in the rating history: https://online-go.com/game/18770227 Was it annulled?

2 Likes

Nice graph! It actually visualises quite nicely how the learning curve is really steep at first, but gradually each step gets longer than the last one ^___^

Lowest i’ve seen a valid played drop is 555, which i guess was 42k https://online-go.com/player/610789/

3 Likes