Show me your data!

[brötchen]

Hi
So recently my interest in nice data-sets has sparked up again and I would be interested in hearing stories about interesting data-sets and statistics you created or found.
The thing that caused me to start this thread was my frustration with the fact that on the official online instance of tetris there are no good statistics available and so i head to make my own statistics to evaluate my skills (I started today and have 38 data-points by now (which equates to almost 76 minutes of game time). Sadly they don't look very normally distributed jet but at least there is a clear spike at around 14000 points per two minute game).
Another nice data-set I recently helped create was meant to increase the measurement precision of the light intensity of a flashlight. basically we measured the beam intensity at different distances and than fitted an inverse square law curve to the data points. although the inverse square low itself isn't all that interesting it was still very satisfying to see the measurements align almost perfectly with the theoretical curve (also someone else posted the same kind of measurement of the same brand and model of flashlight and the values matched ours within a few percent)

So what have you measured? Am i the only one that gets thrilled when a data-set turns out nicely?

(also excuse my less than perfect englisch, its not my native language)

[Sandor]

Am i the only one that gets thrilled when a data-set turns out nicely?

No you're not the only one. People got quite excited about this one:

[Gigano]

Am i the only one that gets thrilled when a data-set turns out nicely?

No you're not the only one. People got quite excited about this one:

Is that the one where the model had an accurate fit with the data within 2000 sigma?

[qetzal]

I used to play a lot of on-line spades, and I'd hear people complain that the deals weren't fair. So I kept track of about 200 consecutive hands dealt to me, and analyzed them for things like number of spades, suit distribution, etc., to see if they conformed to the expectations of random deals. I was pleased to see things match up quite well.

[idobox]

I'm surprised by the smoothness of the curve. I would have expected to see the red shifted absorption spectrum of the random stuff it went through, like hydrogen.

[mfb]

Maybe that is removed from the raw data.
It is important to point out that the error bars are increased by some really large factor - I think I saw plots with factor 20 or 50 - to be visible in the plot.

[PerchloricAcid]

Perhaps offtopic, perhaps not.
My boyfriend and I somehow came up with the idea of having some fun with Google Trends. Among various queries, homework/school related queries seemed to manifest spectacular [periodic] trends

integral, calculus, matlab, equation... all seem to be unpopular during the summer holidays (and highly unpopular during the Christmas/New Year's holidays), but popular during school periods. Oh, and by the way, a slight decrease in search for these terms can be seen somewhere around, say, March - I'd say this is the spring/Easter holiday effect.

We also found the following interesting:
Hamlet, Othello, and Shakespeare seem to be following periodic trends, BUT, manifest long-term decrease. On the other hand, sparknotes, apart from also showing periodic trends, manifests a long-term increase in popularity. Fill in the blanks

Oh, and, porn becomes kinda popular at the end of the year

[mfb]

It tends to become more spiky if you restrict it to single countries, as you don't have the effect of different holiday times there.

[PerchloricAcid]

Sure, we tried that too.

[Username4242]

Summed bobcat and coyote track intrusions in sand stations placed on game trails on a privately owned ranch. The low sample sizes of bobcat intrusions precluded statistical analysis, but I thought that it was an interesting pattern all the same.

[brötchen]

My tetris dataset has become large enugh to be somewhat interesting (N>100)

Here is the scoredistribution:


Here a scatterplot of efficiency (which is calculated by dividing the score by the number of lines cleared) vs. score


Here is a scatterplot of speed vs. efficiency


And here is a scatterplot of speed vs. score


The data points I captured where: score, lines cleared and speed. if any one has a sugestion for other interesting plots let me know

[mfb]

What is the definition of speed? Points per time, lines per time, ...?
I think most of the correlation effects you can find come from a variable which you do not have: Sanity and height of the stack. A nice, small stack allows you to get a good efficiency, a good high score and maybe even a high speed at the same time.

What happens at 14000 points? Why is there no point above the 21000-bin?

[brötchen]

What is the definition of speed? Points per time, lines per time, ...?
I think most of the correlation effects you can find come from a variable which you do not have: Sanity and height of the stack. A nice, small stack allows you to get a good efficiency, a good high score and maybe even a high speed at the same time.

What happens at 14000 points? Why is there no point above the 21000-bin?

The definition of speed is tetriminos (blocks) per minute.
the reason there is no point beyond the 21000 bin is simply because i didnt think I would go that high. I managed to go almost that high a few times as you can see by the fact that the 21000 bin is not 0 but not above so it dosnt realy change anything.
what do you mean by :"What happens at 14000 points?"? its simply the score that occoured most often.

[mfb]

its simply the score that occoured most often.

I can see that in the graph, thanks . But why does it occur so often? The peak looks significant. Does anything special happen at this amount of points? Higher block falling speed? Any other challenge?
More speculative: The even bins (10000, 12000, ...) seem to have more entries than uneven bins. I wonder whether that just happened by chance, or is there any reason for it?
Your 14000-bin goes from 13000 to 14000?

[brötchen]

So the rules of the game mode I played are that a game always lasts at most 2 minutes (of course you can lose before the 2 minute mark by just stacking blocks to high but that happened very rarely because the default block falling speed is so slow) the block falling speed doesn't seem to change much at all and is always fairly slow so the only limiting factor for speed is really how quickly you can decide where to place a block and than drop it (its a hard-drop variant of tetris meaning that you can place the block instantly).
Yes the 14000 bin ranges from (not including) 13000 to (including) 14000.
The scoring system for this particular game is a bit strange as it scores combos (lines cleared with consecutive block) and T-spin (turning a T-block into a place it wouldn't fit in if you just dropped it straight down) very highly but neither of those moves makes up a significant part of my score because I found that i reach the highest scores when i just go with the classic strategy of leaving a slot open and waiting for a straight-block (I stayed consistent with this strategy for the duration of the experiment).
An other interesting fat about the game mode is that it allows you to "hold" one block. At the beginning of the game the "hold slot" or what ever you want to call it is empty and when you press the hold key the current block is placed in the hold slot and you get the nex block, when you later press the hold key again the current block goes into the hold slot and you get the one you placed there before, also you can't switch between the same two blocks multiple times.
I guess the reason for such a sharp spike in the data is simply that my speed was fairly consistent and that there is a time limit. Maybe things would look a little bit more like a bell curve if i used narrower bins but the amount of data doesn't really allow for much narrower bins.

[mfb]

Ah, the time-limit explains the high correlation between speed and score.
And it helps to understand the score distribution.