Tetris DS rating collection

[Xkeeper]

Copied from the Nintendo forums, more likely to work here since I'm assuming you aren't all 13-year-olds =P


Simple enough: start a 2P standard game.
Make a note of your ratings before the game starts, such as

Your rating | Opponent
#### ####

After the game is over, put your new ratings down:

#### ####
#### #### - New ratings here

Continue doing this until you quit, where you put in
#### 0 - current rating and a zero

Then simply repeat. I can go ahead and add them to my various charts and graphs, and possibly even make unique graphs for each user if I have enough samples from you

As an example, here's a few matches in the proper format:
7377 6366
7382 6357
7387 6347
7392 6336
7397 6326
7402 6316
7407 6306
7320 6469
7248 6587
7248 0

[caffeine]

from start to finish.


7714 7547

7724 7535

7734 7523

7743 7512

7752 7501

7761 7490

7732 7526

7741 7515

7750 7504

7759 7493

7766 7484

7774 7474

7781 7465

7749 7506

7757 7495

*forgot to write down score-- was a win (maybe 7772 to something)

7780 7466

7787 7457

7794 7447

7800 7439

7807 7430

7813 7422

7818 7414

7824 7405

7830 7397

player leaves



7830 1310 (whoops, i wrote down his wins instead of rating-- sorry)

7807 7663

7785 7687

7765 7709

7747 7727

7743 7731

7746 7728

player leaves


weird games-- like i was watching him towards the end, and his pieces were going through garbage and stuff, especially at the top of his stack. maybe lag or something?


*game started before i could write ratings down

7766 7834

player leaves


7766 7671

7777 7659

7788 7647

7767 7670

7779 7658

player leaves

this is the same guy as before (the weird games).


7779 7256

7783 7248

7788 7241

7793 7233

7798 7226

7803 7218

7808 7210

7756 7289

7761 7282

7766 7275

7771 7267

7776 7260

player leaves


7776 7587

7785 7577

7793 7566

7802 7556

7809 7549

*i leave (once again the weird games guy-- weird stuff was beginning to happen, so i decided this would make a good stopping point).


hope you can use it, despite those couple of goofs i made.

[Nick15]

I purposefully decimated my own Wifi rating. Mostly for kicks, but I suppose getting a collection of losing stats might prove useful as well.


MINE OPPONENTS

???? ???? *(I forgot to write down the initial rating for this match)

7090 6740

7047 6792

7010 6835

6976 6872

6944 6906

6913 6937

6884 6965

6884


6884 6485

6835 6545

6792 6595

6753 6638

6715 6677

6715


6715 6031

6647 6127

6587 6204

6587


6587 6204

6533 6268

6533


6533 5098

6418 5310

6323 5464

6240 5584

6164 5685

6194 5647

6194


6194 5288

6105 5418

6027 5520

5955 5609

5889 5685

5930 5637

5930


5930 5863

5976 5918

5824 5967

5824

[Xkeeper]

Wow, yes, those were helpful.


Graph here, also (http://xkeeper.shacknet.nu_3A3/e/tetris.php?view=nick15)

[colour_thief]

I purposefully decimated my own Wifi rating. Mostly for kicks, but I suppose getting a collection of losing stats might prove useful as well.


OMG I've corrupted you.


(Yes, I was that opponent you played.)

[Billmaan]

Here are some 5000-rating stats. Sorry I didn't collect more...after this I spent quite a while trying to get into a 5000 vs. 5000 game. (I failed, but the rating change for the winner would probably be 72 or 73.)


5000 6108

5157 6007

5283 5918

5386 5837

5337 5876

5294 5910

5249 5943

5209 5971

5169 6000

5127 6027

5092 6051

5055 6074

5018 6097

4982 6120

4982

[Billmaan]

Here's some more. Some of this stuff is really screwy, and if I hadn't been collecting it myself, I'd swear it was bogus.


5000 5316

5085 5238

5163 5163

5232 5092

5157 5169

5085 5238

5018 5299

4957 5354

4957


5000 4386

5049 4348

5098 4311

5139 4279

5139


5000 5854

5127 5762

5232 5680

5232


5000 5614

5110 5525

5203 5444

5203


5000 4652

5061 4598

5061



I was trying to win a few, then lose my way back to 5000 to collect data on wins and losses (and to avoid screwing up people's ratings), but most people leave after the first couple wins. http://www.tetrisconcept.net/forum/images/smilies/icon_sad.gif

[Nick15]

I purposefully decimated my own Wifi rating. Mostly for kicks, but I suppose getting a collection of losing stats might prove useful as well.

OMG I've corrupted you.

(Yes, I was that opponent you played.)


I played you last night? I didn't play you, did I?

[colour_thief]

Not last night... Maybe 2 weeks ago or so?


I had a higher rating than you, and we played a few games. Eventually I just sort of threw every game as fast as I could. Until you quit. It was a social experiment to see if you were playing for Tetris, or playing for rating points. I don't think you took more than about 5 free wins.


Such an opponent was not memorable?

[Nick15]

Not last night... Maybe 2 weeks ago or so?

I had a higher rating than you, and we played a few games. Eventually I just sort of threw every game as fast as I could. Until you quit. It was a social experiment to see if you were playing for Tetris, or playing for rating points. I don't think you took more than about 5 free wins.

Such an opponent was not memorable?

Ahhh OK, now I remember. It actually happened to me twice, though I'm willing to bet both times were you. At first I though you were off your rocker there, or you were pissed off at losing or something... I thought maybe if I held out for a little bit, you'd snap back to normal. After, well, five games, you didn't, so I left.


Anyways, that's what I did to sabotage my rating, and I got the idea from you. So I suppose you did corrupt me. http://www.tetrisconcept.net/forum/images/smilies/icon_biggrin.gif

Another reason why I sabotaged my rating was that, since I've already proven to myself (and a few others) that I'm a fairly OK Tetris player, I thought I'd do a little shit disturbing by dropping my score, then kicking the crap out of higher ranked players. I can only imagine the kind of panic gripping their soul when they lose like 500 points after playing me. http://www.tetrisconcept.net/forum/images/smilies/icon_razz.gif

[Sully]

XKeeper- are you still collecting info? Haven't seen you around lately.

[colour_thief]

I believe he's on vacation for the next while. I think he wouldn't mind more data waiting for him when he returns though. http://www.tetrisconcept.net/forum/images/smilies/icon_smile.gif

[jjdb210]

In order to help XKeeper get this data faster, as well as hopefully make it easier to keep track of... Blockstats has added a tool to the members section to record this sort of data. It will feature Xkeeper's graph shortly, but the mean time, we can use it to start collecting data. I'm going to try to import as much as I can into the table, hopefully at some point, we can figure out exactly what the formula is (even if it's not really a formula and just a pairing of most of the combos between 3000 and 9000, ok so that's harder to do but eh...)


Let me know if you think anything can be done to make it better.

There most likely is a formula combined with logistics and range factors. This is similar to how chess rating systems work (i.e. Elo). I have gone through the data myself using these simple logistics and I have gotten values that are very similar. However, it does not remain consistent over different groups of data which suggest various factors (or methods of adding factors) are used depending on the point different between two players, the rating of the player, etc.


I have done a lot of analysis and have found, so far, the following (some of these may be obvious, and assume the general logistical model is used with a similar system as in chess which is most likely what it is based around):


1) When the players are closer together, this factor appears to be lower then when they are further apart.

2) The higher-rated player generally tends to take on a higher factor then the lower-rated player.

3) The relationship between the difference in the players ratings before playing the game appears to be fairly linear. However, as this approaches 0, it becomes more non-linear suggesting logistics and the possibility of an exponential function within (as was used in the general logistical model to approximate an expected win). There is a slight curvature to this line which may make it very difficult to figure out.

3) The relationship in the summation of the two players differences of rating change for the game appears to be approximately linear too. However, form the data I have, it is difficult to tell, but it appears to become very non-linear as it approaches zero. As with the previous relationship, there does not appear to be a complete linear regression within the curve which will could make it very difficult to find a suitable equation without guessing or making an approximation.

4) There appears to be a clear relatioship between the two previously mention curves (from 3 and 4). As one decreases, the other also appears to decreas. This just means that the closer the two players are together in rating, the smaller the difference will be between gained and lost points (taken as an absolute value to find the magnitude as negative values are not significant in this situation).

5) As the two players increase in rating (or as their total rating (P1 + P2) inscreases), while still close together, it appears that the absolute value of the overall points |won + lost| approaches zero. This would state that at high ratings (both players at about ~7500+, and close together) there are no net loss of points.


This is all that I have found so far that seems to comply with all of my compiled data. If anyone else wants to post their data I'll look at it. I am really bored. LOL. Also, there are a couple data points in some of the data which are very odd and I can't even figure out why they do what they do. I just hope that XKeeper has more luck then me. It seems that my Mechanical Engineering labs had us finding similar patterns to this one, but most of them have approximations which are much easier to see when graphed then this.


Cheers

In my previous post there, you may notice that my numbers are a little out of order. That second 3 should actually be a 4, the 4 a 5, etc... Sorry about that.