Summary data from 100K MGT planets.

captainjack23

Cosmic Mongoose
Okay. THIS is the file with 100K summary results. For those of you who caught my earlier (brief) post, you'll notice that this DOES have 100K planets (+2, actually).

See, I made what we statistical types call a Type VI error : Failure to send the correct file.

Okay, so I ran off 100K planets last night by MGP rules. The below is the basic summary data. For % just lop off the last two digits.

Caveats Mentor:
1. The list is generated with my best attempt at pure 3.2; ie the document + Gar's explicit clarifications (In worlds and Pop 0 worlds)
2. No, i didn't comb this data or code to journal review standards - I have other things to do next week....
3. With this size sample, duplicate uWps will pop up, tweaking the results somewhat.
4. Why so many ? Well, it was about as easy as running 10,000, looks cooler, and provides a good pool of datapoints for cross tabulation - which when counting combinations, can make your datapool dimish VERY quickly.
5. Who cares ? Well, me. Plus, I can start running what for me is the interesting analysies, what combinations of planets are found : for instance, how many tech 4 Pop A Starport E In worlds are there ?


Code:
The SAS System                              19:17 Thursday, January 24, 2008 262

     SIZ
------------------
       0      3968
       1      8040
       2     12348
       3     16306
       4     15582
       5     15722
       6     12140
       7      7960
       8      4972
       9      1960
      10      1004

The SAS System                              19:17 Thursday, January 24, 2008 263

     ATM
------------------
       0     19802
       1      9370
       2     11256
       3     12216
       4     11642
       5     10394
       6      8688
       7      6664
       8      4358
       9      2722
      10      1756
      11       788
      12       346

The SAS System                              19:17 Thursday, January 24, 2008 264

     HYD
------------------
       0     28000
       1      8140
       2      9316
       3     10252
       4     10494
       5      9514
       6      8018
       7      6138
       8      4350
       9      2766
      10      3014

The SAS System                              19:17 Thursday, January 24, 2008 265

     POP
------------------
       0      3976
       1      8134
       2     11968
       3     15812
       4     15978
       5     16264
       6     12030
       7      7868
       8      5022
       9      1930
      10      1020

The SAS System                              19:17 Thursday, January 24, 2008 266

     GOV
------------------
       0     19572
       1      9296
       2     11212
       3     12210
       4     11920
       5     10506
       6      8616
       7      6504
       8      4402
       9      2776
      10      1606
      11       844
      12       374
      13       164

The SAS System                              19:17 Thursday, January 24, 2008 267

     LAW
------------------
       0     29302
       1      9790
       2     10262
       3     10180
       4      9436
       5      8178
       6      6762
       7      5318
       8      3948
       9      2744
      10      4082

The SAS System                              19:17 Thursday, January 24, 2008 268

     TLv
------------------
       0      5814
       1      2202
       2      4026
       3      7750
       4     10806
       5     13242
       6     14028
       7     12570
       8     10684
       9      7862
      10      5208
      11      3010
      12      1702
      13       650
      14       300
      15       116
      16        28
      17         4

The SAS System                              19:17 Thursday, January 24, 2008 269

PORT
--------------
A         2998
B        11752
C        25408
D        31128
E        22054
X         6662

------------
Ag     12590





------------
Ba      1126



------------
De      7396


------------
Fl      2740



------------
Ga     15436




------------
Hi      2950




------------
Ht      2800




------------
IC      6288




------------
In       618




------------
Lo     35914







------------
Lt     39864



------------
Na     10968




------------
NI     80186


------------
PO     22284


------------
Ri      4904



------------
Wa      2914


*for reference: Type I error False reject of null hypothesis.
Type II error False accept of null hypothesis.
Type III error Failure to avoid a math error
Type IV error: Failue to use the correct stat test.
Type V error: Failue to debug code .
Yes, iii-VI are what passes for stat humor.
 

captainjack23

Cosmic Mongoose
And the answer is, FWIW:
(Pop = A and Tlv =4 and Port ='E' and In =1) = 8

The breakdown of tech levels for In planets is interesting:
(collapsed across all other vars)

Code:
  TLv X In code
    CODE   COUNT
------------------
       0         2
       1        16
       2        22
       3        42
       4        60
       5        76
       6        84
       7        88
       8        80
       9        52
      10        50
      11        22
      12        20
      13         4

Thefrom 5-8, the distribution is pretty flat with a bit of a fall off for 2-4 and 9-10.

Interstingly, there is more chanvce of a Tech 3- world than of a tech 11+ world.
(note that all counts are <1% all planets (100,002); total count of all In planets = 618 ).

So, Lo tech In planets are the norm. The heck ?
 

captainjack23

Cosmic Mongoose
Just for S&G here are the base results for the worst offenders.
Code:
   PLANET       SIZ       ATM       HYD       POP       GOV       LAW 	TLv         PORT  
--------------------------------------------------------------------------------------------------------
   14350         6         4         3        10         6         4 	1 	E

   20941         6         4         5        10         9         5  	1   	E

   28212         7         9         4        10        10         5 	1      E



          .         .         .         .         .
         
* Pop = 10 and Tlv =1 and Port ='E' and In =1
 

EDG

Mongoose
You don't have all those values as percentages of the total, do you? Would make it easier to compare our results...

(EDIT: And OK, you win, I'll do runs of 100k. Turned out it wasn't as fiddly as I thought it would be).
 

captainjack23

Cosmic Mongoose
EDG said:
You don't have all those values as percentages of the total, do you? Would make it easier to compare our results...

(EDIT: And OK, you win, I'll do runs of 100k. Turned out it wasn't as fiddly as I thought it would be).

Well, glad that worked out for you. For me, with monster grade stat equiptment, its mainbly just fitting it in between real work. With MUCH bigger sets (.5G row tables).

As to %, just lop off the last two digits of the count.
 

EDG

Mongoose
Code:
     SIZ
------------------
       0      3968
       1      8040
       2     12348
       3     16306
       4     15582
       5     15722
       6     12140
       7      7960
       8      4972
       9      1960
      10      1004

You may want to check this. Size is an unmodified 2d6-2 roll, so that distribution should be exactly like a 2d6 bellcurve, but shifted downwards by 2 steps. But it isn't. In fact, it peaks at size 3 for some odd reason - it definitely should peak at 5.

(also the total is 100,002, not 100,000).
 

EDG

Mongoose
Here's a comparison:

Code:
   EDG#    EDG%      2d%     CJ%      CJ#
0  6918     2.8      2.8     4.0     3968
1  13394    5.6      5.6     8.0     8040
2  20907    8.4      8.3    12.3    12348 
3  27570   11.0     11.1    16.3    16306  
4  34711   13.9     13.9    15.6    15582 
5  41641   16.7     16.7    15.7    15722 
6  34840   13.9     13.9    12.1    12140 
7  27941   11.2     11.1     8.0     7960 
8  20614    8.2      8.3     5.0     4972
9  13907    5.6      5.6     2.0     1960
A  6957     2.8      2.8     1.0     1004

The first column is the MGT world size (determined using 2d-2).
The second column shows the number of worlds I got with that size on my 250k pure MGT run.
The third column shows my percentages - the numbers from column two divided by 250,000.
The fourth column shows a comparison with the percentages of a standard 2d6-2 roll.
The fifth column shows captainjack's percentages - the numbers in his size table divided by 100,002.
The sixth column shows captainjack's numbers from his 100k run.

As you can see, my results follow the 2d6 probability curve very closely, but captainjack's results deviate significantly from the expected 2d-2 curve. So either he's making a wrong assumption somewhere, or there's a serious bug in his code (I'm not entirely sure what can distort the percentages in that way in his results though).

CJ's Pop distribution has the same problem, and I think the skewy size and pop are both screwing up the other UWP stats derived from them. Compare this with the 250k stats I did on this thread.


EDIT: And before you wonder if it's because I'm using 250k results instead of 100k, it's not - here's the spread for 100k:

Code:
CT100k
2.8
5.5
8.3
11.2
13.9
16.6
13.9
11.2
8.4
5.5
2.8

As you can see, it's also very close to the 2d% curve.

(Also, ignore Aramis' post below, he's referring to a previous version of the tables that I accidentally mis-labelled. My code is fine.)
 

AKAramis

Mongoose
EDG said:
Here's a comparison:

Code:
   EDG#    EDG%      2d%     CJ%
0  3968     2.8      2.8     4.0
1  8040     5.6      5.6     8.0
2  12348    8.4      8.3    12.3
[b]3  16306   11.0  [/b]   11.1    16.3  
[b]4  15582   13.9  [/b]   13.9    15.6
[b]5  15722   16.7 [/b]    16.7    15.7
6  12140   13.9     13.9    12.1
7  7960    11.2     11.1     8.0 
8  4972     8.2      8.3     5.0
9  1960     5.6      5.6     2.0
A  1004     2.8      2.8     1.0

The first column is the MGT world size (determined using 2d-2).
The second column shows the number of worlds I got with that size on my 250k pure MGT run.
The third column shows my percentages - the numbers from column two divided by 250,000.
This tells me your code is buggy as a batcave...
your percentages describe a different curve than your numbers column.

If the numbers are accurate, your random function is badly flawed.
 

EDG

Mongoose
AKAramis said:
This tells me your code is buggy as a batcave...
your percentages describe a different curve than your numbers column.

If the numbers are accurate, your random function is badly flawed.

Oops. I evidently mis-labelled the columns (doh!). The second column isn't my numbers, they're CJ's (as is apparent if you look at his table in the first post - they're the same numbers).

So actually, my code is fine - it's CJ's that has the problems, like I said.

Hang on, I'll do a better chart to avoid confusion (showing my numbers as well) - I've edited the table in my post now. (so ignore the one Aramis quoted as it's mislabelled and therefore confusing - in fact if Aramis can delete his post or at least the quoted table then that'd be better all around).
 

captainjack23

Cosmic Mongoose
EDG said:
AKAramis said:
This tells me your code is buggy as a batcave...
your percentages describe a different curve than your numbers column.

If the numbers are accurate, your random function is badly flawed.

Oops. I evidently mis-labelled the columns (doh!). The second column isn't my numbers, they're CJ's (as is apparent if you look at his table in the first post - they're the same numbers).

So actually, my code is fine - it's CJ's that has the problems, like I said.

Hang on, I'll do a better chart to avoid confusion (showing my numbers as well) - I've edited the table in my post now. (so ignore the one Aramis quoted as it's mislabelled and therefore confusing - in fact if Aramis can delete his post or at least the quoted table then that'd be better all around).

Well, thats why I've been posting it - for review. Not JUST for you and Aramis to get into an argument... :arrow: :mrgreen:

Sure looks odd to me, so thanks. You're about two revs ahead of me on debugging, thus far, so go with what you've got. Plus, I have only had at best 25% attention to the work due to...well, real work.

Without starting a fight ,this is why I prefer working with the earlier prob counts as opposed to sampling pseudo random data.

I'll check it again when I have access to it (Monday), or not - time is very short as it is for the playtest, so go ahead and ignore these results ,I may just shelve them for now.

"Buggy as a batcave "? Now ,is that nice to Batman ?
 
Top