Fun With Graphs

Finally I found someone who shows it clearly 

-

Well, I have 2300 pictures now heh.

Most of them are pngs that get converted to gifs, but yeah, saved ~30 people's worth. Got some good examples of different kinds of players.

One type I ran into is kind of interesting.

-

-

Pretty much all of this person's games are 10|0. Their times are shifted a few seconds forward from what's common... but IMO this just means they take their time. If there were NO moves below, say, 5 seconds, then that would be strange, or if the shape were different, but it's a gradual decline on both sides. I think they're just habitually cautious or something.

-

And another of the same type.

-

-

Found 6 graphs like this among the top 100 untitled rapid players.

Untitled meaning overall they're in the top 300, but they're top 100 if you only count untitled players.

I limited it to games no older than 2 years old... if you're wondering who among those 100 played the most overall, it was this dude with over 8300 rapid games (and when adding in blitz and bullet, over 13,000 total):

@chessrook5t1

Like everyone with 1000s of games in a graph, his was nice and smooth

-

-

The most disgusting graph award belongs to this guy (below).

Sure only a few games makes graphs look bad, but there were plenty with only a few games, and none as ugly as this one lol.

-

me with the new updates please(you have the data already so it should be quick)

imo the change from set number of games to moves helps show relative shifts more

-

Yeah, keeps the total area the same.

what was that peak in 2 secs? it went off-screen i think...

Yeah, I set the top to be the average (notice 0.15 is the top of your overall rapid graph).

That way when someone goes much higher or lower I can see that fact easily.

On average, 2 seconds represents ~14% of your rapid moves. For a set of 50-ish games, it went a little higher or lower. I don't want to regraph it to find out whether it's 15.2% or 15.7%

Here's an example of someone going much lower and much higher. Also notice the rating goes from something like 200 to 2500.

-

That ugly graph over there...I see a giant peak around 0. Wonder if those are attributed to premoves.

Yes

The furthest bar to the left is premoves.

May I suggest trying to fit a Poisson distribution to the data? I know movetime is not a discrete variable, but I guess it shouldn't matter, since you're taking intervals anyway.

I tried a few, like a shifted Laplace and a Rayleigh. Nothing looked as good as a shifted exponential where I'm just ignoring that the left side exists  

Since I'm rounding, a discrete variable would make sense.

I looked a bit into chi-squared distributions and, you know, trying to do some best fit curves, and that'd be the next step (along with cleaning up some of the code and getting more data. I have ~120 users). But I'm done messing with it. Mostly this was a fun project to try and prove my idea that players exist who simultaneously have a rating jump and moves that suddenly group around a time (like 10 seconds). I found 4-6 "obvious" cases of cheating, and in total ~30 players who I thought were worth reporting by looking at the top 100 non-titled rapid players. I made a topic about it in the cheating forum.

goddamn llama I’ve seen the “-“ symbol more times in the past 5 minutes I spent reading through this thread than I have in my entire life

Some years ago I got tired of how they handled whitespace, and things like enter vs shift-enter. Simpler to just add a character.

Also, after inserting a game or image, sometimes there's no spacing below it so if you keep typing after that the formatting is all messed up like your text will be centered... so before you even begin, it's useful to block things out with something like this:

-

-

-