Duel2.Com  
•   Home  •  Rules  •  Your Account  •  Forums  • Newsletters  •
Navigate
· Home
· Content
· Encyclopedia
· Forums
· Members List
· Newsletters
· Old Newsletters
· Private Messages
· Setup
· Tourneys
· Your Account
User Info
Welcome, Anonymous
Nickname
Password
(Register)
Membership:
Latest: propertyturkey023
New Today: 1
New Yesterday: 0
Overall: 104549

People Online:
Visitors:
Members:
Total: 0
Duel2.Com: Forums

Duel2 :: View topic - Roll Up Pool
 Forum FAQ  •  Search  •  Memberlist  •  Usergroups   •  Register  •  Profile  •  Log in to check your private messages  •  Log in

 
Post new topicReply to topic
View previous topic Log in to check your private messages View next topic
Author Message
pipthetroll
Advanced Master Poster
Advanced Master Poster


Joined: Nov 04, 2002
Posts: 447
Location: In my underwear, in front of my computer

PostPosted: Mon Oct 30, 2006 10:24 am Reply with quoteBack to top

The Consortium wrote:
It is sounding more and more to us that the issue is not about size (or rather fairness or bell-shape curvedness of size), but rather about wanting "smaller" warriors - both for size sake itself, and because that leaves a likelihood that there will be more WT/WL.

If so, why not just say to RSI exactly that?


This is exactly correct. There is nothing wrong with the size distribution

Look at the distribution, it is a bell curve, based on 3d7. There are 343 different ways to roll 3d7, 1 is 3, 1 is 21; thats a 1/343 chance of getting a size 3. The most common size will be 12, followed by 11 and 13(equal chance), followed by 10 and 14(equal chance). Then comes your problems, size 9 and size 15--same chance;size 8 and 16--same chance; size 7 and 17--same chance. % wise, more size 7-9 rollups are considered "runable". Getting a size 15-17 is usually a kick in the nuts, not many of them are as "runable"(less points = lower starting wit and will).

The distribution is perfect for the range, problem is you are looking for it to be perfect for the size penalties/bonus.

A size 3 warrior has an average starting stat of 67/6 = 11 1/6. A size 21 has an average starting stat of 49/6 = 8 1/6. On average, the smaller you are, the better your other starting stats.

15's and 16's seem like they are prolific, because you get as many of them as size 8 and 9's. Size 8's and 9's are almost always ran, size 15 and 16 are almost always da'ed.

3 options to remedy it.

1. Shift the size distribution. You'll still get alot of horrible replacements. They'll just be smaller on average.

2. Set the code to reroll ones that dont meet certain criteria. The criteria could destroy your chances of getting some wierd yet runnable rollup.

3. Go back to the pool. The easiest to implement. Easier to update. Easy to remove a rollup that is always da'd and set the % of the really good ones.

The pool is probably the best option, and if you randomize the order they pop out in, noone will notice.

I remeber when they changed it. It was the guy who bitched and moaned about the manager allahs name who bitched and moaned about their being a pool. He wanted there to be a natural triple 21 possible. One hasnt popped out yet anyways. I think it was miles or pug who wnet off against it, saying you really dont want random, there are too many really bad combinations that will never get run. I had always assumed when RSI announced they changed it, they just announced it to shut people up and never did. Guess I was wrong, hehe.

So to end the rant and go back to what the consortium says, ask rsi what we want--we want most warriors from the da to be runnable, with some crappy, and some godly. Either have the program reroll the real bad ones(or tighten the parameters if this already exists), or go back to the pool.
View user's profileSend private messageSend e-mailICQ Number
Drake
Master Poster
Master Poster


Joined: Oct 27, 2002
Posts: 249
Location: St. Louis, MO

PostPosted: Mon Oct 30, 2006 11:21 am Reply with quoteBack to top

pipthetroll wrote:
This is exactly correct. There is nothing wrong with the size distribution

Look at the distribution, it is a bell curve, based on 3d7. There are 343 different ways to roll 3d7, 1 is 3, 1 is 21; thats a 1/343 chance of getting a size 3. The most common size will be 12, followed by 11 and 13(equal chance), followed by 10 and 14(equal chance). Then comes your problems, size 9 and size 15--same chance;size 8 and 16--same chance; size 7 and 17--same chance. % wise, more size 7-9 rollups are considered "runable". Getting a size 15-17 is usually a kick in the nuts, not many of them are as "runable"(less points = lower starting wit and will).


For the sake of easy comparison, I've put the theoretical distribution of a 3d7 roll next to the combined Guardian/Consortium data (from Guardians Mon Oct 30, 2006 8:10 am post)

*****3d7***** || **Guard/Cons** || **Diff**
03 | 01 | 00.30% || 03 | 03 | 00.80% || +0.50%
04 | 03 | 00.90% || 04 | 08 | 02.00% || +1.10%
05 | 06 | 01.70% || 05 | 14 | 03.60% || +1.90% <--
06 | 10 | 02.90% || 06 | 15 | 03.80% || +0.90%
07 | 15 | 04.40% || 07 | 18 | 04.60% || +0.20%
08 | 21 | 06.10% || 08 | 28 | 07.10% || +1.00%
09 | 28 | 08.20% || 09 | 27 | 06.90% || -1.30%
10 | 33 | 09.60% || 10 | 29 | 07.40% || -2.20% <--
11 | 36 | 10.50% || 11 | 29 | 07.40% || -3.10% <--
12 | 37 | 10.80% || 12 | 40 | 10.20% || -0.60%
13 | 36 | 10.50% || 13 | 41 | 10.50% || +0.00%
14 | 33 | 09.60% || 14 | 41 | 10.50% || +0.90%
15 | 28 | 08.20% || 15 | 33 | 08.40% || +0.20%
16 | 21 | 06.10% || 16 | 20 | 05.10% || -1.00%
17 | 15 | 04.40% || 17 | 20 | 05.10% || +0.70%
18 | 10 | 02.90% || 18 | 10 | 02.60% || -0.30%
19 | 06 | 01.70% || 19 | 08 | 02.00% || +0.30%
20 | 03 | 00.90% || 20 | 06 | 01.50% || +0.60%
21 | 01 | 00.30% || 21 | 02 | 00.50% || +0.20%
* I've marked those areas where the difference between the expected & actual were >= 1.5%

Broken down into the categories Guardian suggested and we get:

*******3d7******* || ***Guard/Cons**** || **Diff**
03-03 | 001 | 00.30% || 03-03 | 003 | 00.8% || +0.50%
04-05 | 009 | 02.60% || 04-05 | 022 | 05.6% || +3.00% <--
06-07 | 025 | 07.30% || 06-07 | 033 | 08.4% || +1.10%
08-12 | 155 | 45.20% || 08-12 | 153 | 39.0% || -6.20% <--
13-14 | 069 | 20.10% || 13-14 | 082 | 20.9% || +0.80%
15-16 | 049 | 14.30% || 15-16 | 053 | 13.5% || -0.80%
17-20 | 034 | 09.90% || 17-20 | 044 | 11.2% || +1.30%
21-21 | 001 | 00.30% || 21-21 | 002 | 00.5% || +0.20%
* I've marked those areas where the difference between the expected & actual were >= 1.5%

There appears to be a significantly lower number of 8-12s being generated than should be (by about 6%). Now, with a sample size of only about 400 replacements, there's bound to be some variance. However, 6% does seem a bit high. Interestingly, 15-16 does appear to be about right. So, perhaps a fix to the Size generation method is not what is needed, and rather a check on the Wit/Will of the generated warriors. I'll re-propose the following check: If generated Wit and Will are both less than 9, re-roll (coinsides with fix 2 of Pip's post). This will ensure that every roll-up has either a 15 Wit or 15 Will, but not neccessarily both.

_________________
Deric (Drake) Page
Currently MIA
Playing since 1988 (North Fork turn 10)
Known primarily for being a wise-ass
****
Success is not the result of spontaneous combustion. You have to set yourself on fire. - Unknown
View user's profileSend private messageSend e-mailYahoo MessengerICQ Number
The Consortium
ArchMaster Poster
ArchMaster Poster


Joined: Nov 23, 2002
Posts: 10133
Location: on the golf course, in the garden, reading, traveling, and now Consulting

PostPosted: Mon Oct 30, 2006 11:23 am Reply with quoteBack to top

Woody wrote:
I'll try to be blunt.

There is no way to tell if the distribution is right without a big enough random sample size.

How big is big enough?

SZ 3-21 is 19 different "cell" values. You have to have at least 5 show up in your lowest frequency cells (at least 5 SZ 3 and at least 5 SZ 21).

The data shown so far indicates the frequency of SZ 3 and 21 is about 0.5% each (5 out of 1000). The data isn't very useful for a random sample of less than 1000.


In English: Until you've gotten about 1000 random rollups, you can't tell whether or not you're getting hosed.


Woody, thank you very much. Are you sure of this?

_________________
The Consortium: Crapmaster 2013, Crapgiver 2014; 1213 ADM graduates (40+ manager IDs) including 176K+ fights and 118K+ wins plus 4 teams with 1500+ wins (Animal Farm DM11 @2085; Bulldogs DM11 @ 1976; Lenpros DM30 @ 1792; Fandils DM46 @1727
View user's profileSend private messageSend e-mail
The Consortium
ArchMaster Poster
ArchMaster Poster


Joined: Nov 23, 2002
Posts: 10133
Location: on the golf course, in the garden, reading, traveling, and now Consulting

PostPosted: Mon Oct 30, 2006 11:30 am Reply with quoteBack to top

Drake wrote:
pipthetroll wrote:
This is exactly correct. There is nothing wrong with the size distribution

Look at the distribution, it is a bell curve, based on 3d7. There are 343 different ways to roll 3d7, 1 is 3, 1 is 21; thats a 1/343 chance of getting a size 3. The most common size will be 12, followed by 11 and 13(equal chance), followed by 10 and 14(equal chance). Then comes your problems, size 9 and size 15--same chance;size 8 and 16--same chance; size 7 and 17--same chance. % wise, more size 7-9 rollups are considered "runable". Getting a size 15-17 is usually a kick in the nuts, not many of them are as "runable"(less points = lower starting wit and will).


For the sake of easy comparison, I've put the theoretical distribution of a 3d7 roll next to the combined Guardian/Consortium data (from Guardians Mon Oct 30, 2006 8:10 am post)

*****3d7***** || **Guard/Cons** || **Diff**
03 | 01 | 00.30% || 03 | 03 | 00.80% || +0.50%
04 | 03 | 00.90% || 04 | 08 | 02.00% || +1.10%
05 | 06 | 01.70% || 05 | 14 | 03.60% || +1.90% <--
06 | 10 | 02.90% || 06 | 15 | 03.80% || +0.90%
07 | 15 | 04.40% || 07 | 18 | 04.60% || +0.20%
08 | 21 | 06.10% || 08 | 28 | 07.10% || +1.00%
09 | 28 | 08.20% || 09 | 27 | 06.90% || -1.30%
10 | 33 | 09.60% || 10 | 29 | 07.40% || -2.20% <--
11 | 36 | 10.50% || 11 | 29 | 07.40% || -3.10% <--
12 | 37 | 10.80% || 12 | 40 | 10.20% || -0.60%
13 | 36 | 10.50% || 13 | 41 | 10.50% || +0.00%
14 | 33 | 09.60% || 14 | 41 | 10.50% || +0.90%
15 | 28 | 08.20% || 15 | 33 | 08.40% || +0.20%
16 | 21 | 06.10% || 16 | 20 | 05.10% || -1.00%
17 | 15 | 04.40% || 17 | 20 | 05.10% || +0.70%
18 | 10 | 02.90% || 18 | 10 | 02.60% || -0.30%
19 | 06 | 01.70% || 19 | 08 | 02.00% || +0.30%
20 | 03 | 00.90% || 20 | 06 | 01.50% || +0.60%
21 | 01 | 00.30% || 21 | 02 | 00.50% || +0.20%
* I've marked those areas where the difference between the expected & actual were >= 1.5%

Broken down into the categories Guardian suggested and we get:

*******3d7******* || ***Guard/Cons**** || **Diff**
03-03 | 001 | 00.30% || 03-03 | 003 | 00.8% || +0.50%
04-05 | 009 | 02.60% || 04-05 | 022 | 05.6% || +3.00% <--
06-07 | 025 | 07.30% || 06-07 | 033 | 08.4% || +1.10%
08-12 | 155 | 45.20% || 08-12 | 153 | 39.0% || -6.20% <--
13-14 | 069 | 20.10% || 13-14 | 082 | 20.9% || +0.80%
15-16 | 049 | 14.30% || 15-16 | 053 | 13.5% || -0.80%
17-20 | 034 | 09.90% || 17-20 | 044 | 11.2% || +1.30%
21-21 | 001 | 00.30% || 21-21 | 002 | 00.5% || +0.20%
* I've marked those areas where the difference between the expected & actual were >= 1.5%

There appears to be a significantly lower number of 8-12s being generated than should be (by about 6%). Now, with a sample size of only about 400 replacements, there's bound to be some variance. However, 6% does seem a bit high. Interestingly, 15-16 does appear to be about right. So, perhaps a fix to the Size generation method is not what is needed, and rather a check on the Wit/Will of the generated warriors. I'll re-propose the following check: If generated Wit and Will are both less than 9, re-roll (coinsides with fix 2 of Pip's post). This will ensure that every roll-up has either a 15 Wit or 15 Will, but not neccessarily both.


DRAKE/PIP - we are getting somewhere, we think.

Why not add the 44 points of data that Mannequin offered too - just to solidify things?

The idea of EITHER WT or WL @9+ makes great sense, and is simple.

_________________
The Consortium: Crapmaster 2013, Crapgiver 2014; 1213 ADM graduates (40+ manager IDs) including 176K+ fights and 118K+ wins plus 4 teams with 1500+ wins (Animal Farm DM11 @2085; Bulldogs DM11 @ 1976; Lenpros DM30 @ 1792; Fandils DM46 @1727
View user's profileSend private messageSend e-mail
pipthetroll
Advanced Master Poster
Advanced Master Poster


Joined: Nov 04, 2002
Posts: 447
Location: In my underwear, in front of my computer

PostPosted: Mon Oct 30, 2006 12:08 pm Reply with quoteBack to top

Drake wrote:

There appears to be a significantly lower number of 8-12s being generated than should be (by about 6%). Now, with a sample size of only about 400 replacements, there's bound to be some variance. However, 6% does seem a bit high. Interestingly, 15-16 does appear to be about right. So, perhaps a fix to the Size generation method is not what is needed, and rather a check on the Wit/Will of the generated warriors. I'll re-propose the following check: If generated Wit and Will are both less than 9, re-roll (coinsides with fix 2 of Pip's post). This will ensure that every roll-up has either a 15 Wit or 15 Will, but not neccessarily both.


Actually theres a much easier fix. Just let players add 7 or even 8 max on arena replacements. You'd basically be able to make 1 possibly 2 more breakpoints on a replacement. Will make the weaker rollups more runnable, and wont make the uber ones that much better. This would also encourage players to go for da replacements rather than teams, which would stop the team sheet madness and hopefully consolidate things a bit.

RSI would sell less team sheets, but they'd get more da's. Less replacements would be sold, but they'd be charging 50% more for them. Right now you get about 1 tourney warrior for every 2 sheets or so, or 10$ each, You should get 1 tourney warrior back from the da for every 7 da's(7fights @ 1.50=10.50). Right now its not even close.
View user's profileSend private messageSend e-mailICQ Number
Drake
Master Poster
Master Poster


Joined: Oct 27, 2002
Posts: 249
Location: St. Louis, MO

PostPosted: Mon Oct 30, 2006 12:35 pm Reply with quoteBack to top

The Consortium wrote:

DRAKE/PIP - we are getting somewhere, we think.

Why not add the 44 points of data that Mannequin offered too - just to solidify things?

The idea of EITHER WT or WL @9+ makes great sense, and is simple.


*****3d7**** || ****Survey*** || **Diff**
03 | 01 | 00.3% || 03 | 03 | 00.7% || +0.40%
04 | 03 | 00.9% || 04 | 09 | 02.1% || +1.19%
05 | 06 | 01.7% || 05 | 17 | 03.9% || +2.15% <--
06 | 10 | 02.9% || 06 | 17 | 03.9% || +0.98%
07 | 15 | 04.4% || 07 | 19 | 04.4% || -0.02%
08 | 21 | 06.1% || 08 | 29 | 06.7% || +0.53%
09 | 28 | 08.2% || 09 | 30 | 06.9% || -1.28%
10 | 33 | 09.6% || 10 | 31 | 07.1% || -2.51% <--
11 | 36 | 10.5% || 11 | 34 | 07.8% || -2.70% <--
12 | 37 | 10.8% || 12 | 45 | 10.3% || -0.47%
13 | 36 | 10.5% || 13 | 44 | 10.1% || -0.40%
14 | 33 | 09.6% || 14 | 46 | 10.6% || +0.93%
15 | 28 | 08.2% || 15 | 38 | 08.7% || +0.55%
16 | 21 | 06.1% || 16 | 23 | 05.3% || -0.85%
17 | 15 | 04.4% || 17 | 21 | 04.8% || +0.44%
18 | 10 | 02.9% || 18 | 13 | 03.0% || +0.07%
19 | 06 | 01.7% || 19 | 09 | 02.1% || +0.31%
20 | 03 | 00.9% || 20 | 06 | 01.4% || +0.50%
21 | 01 | 00.3% || 21 | 02 | 00.5% || +0.17%
* Differences above 1.5% have been marked.

*******3d7****** || *****Survey***** || **Diff**
03-03 | 001 | 00.3% || 03-03 | 003 | 00.7% || +0.40%
04-05 | 009 | 02.6% || 04-05 | 026 | 06.0% || +3.34% <--
06-07 | 025 | 07.3% || 06-07 | 036 | 08.3% || +0.97%
08-12 | 155 | 45.2% || 08-12 | 169 | 38.8% || -6.43% <--
13-14 | 069 | 20.1% || 13-14 | 090 | 20.6% || +0.53%
15-16 | 049 | 14.3% || 15-16 | 061 | 14.0% || -0.29%
17-20 | 034 | 09.9% || 17-20 | 049 | 11.2% || +1.33%
21-21 | 001 | 00.3% || 21-21 | 002 | 00.5% || +0.17%
* Differences above 1.5% have been marked.

The inclusion of Manequinn's data didn't really make any difference.

_________________
Deric (Drake) Page
Currently MIA
Playing since 1988 (North Fork turn 10)
Known primarily for being a wise-ass
****
Success is not the result of spontaneous combustion. You have to set yourself on fire. - Unknown
View user's profileSend private messageSend e-mailYahoo MessengerICQ Number
guardian
Advanced Master Poster
Advanced Master Poster


Joined: Nov 05, 2002
Posts: 334

PostPosted: Mon Oct 30, 2006 3:13 pm Reply with quoteBack to top

no no no of course getting smaller warriors will help but it is not my point .



creat e a ranbdom number generator that rolls 3d7 and adds them generate 10,ooo numbers quantify the percentages ( and then remember that turbo pascal uses some wierd method to generate numbers that can never be as truely random as 3 dice can be ).

i dont know if to this day any computor can generate truly random numbers , somthing about needing a seed like the system clock and then generating a random string of numbers , but im sure somone with more programming knowledge than me can fill us in . one of you out there like drake prolly .

did i mention i have more than 5000 warriors

my contention is that it was not programmed with the size breakpoints in mind only to achive a random diminishing at the ends curve and it does exactly that . but

based on the sz breakpoints

which are

3-4
5-6
8-14
15-16
17-20
21


those are the classes of sizes in relation to skills therefore they should have different percentage distributions

and they do but

1 the difference breaks acroos them correctly (from a pure math standpoint )

2 from a usfeulness standpoint the distribution is wrong , because of the sze breakpoints alone it should be probably be different , but based on where the skills are in sz vs the extra hitpoints or lack of them they do not distrbute correctly , now if those bigger guys got a shitload more hitponts to adeqetly compensate them for thier lack of defense and parry it would again probably be fine .

i wont bring up inititive because it is depends on very specific sets of circumstances weather or not it matters ,and quite frankly for the most part the difference in inititive is moot any way and can be debated elswhere .


just from a simple math standpoint roll 3 d7 and you will get the disribution we are seeing

while the percentage chance of any one die coming up any number from 1-7 is theoretically the same the chance of rolling multiples of the same increase greatly which is why there are so few sz 3 and sz 21 and why the curve goes directly to the center point .

and if the damage(doing and taking ) advantage at larger sizes was equal to the defense advantage at smaller sizes this would be fine , but it is not .

(you would have to add 2 damage classes and 2 hitpoint classes to make it more equitable )

(unearthly damage is as high as it goes (even sz 21 does not get its own damage class and it should ).

seemingly unnafected by pain is as high as it goes and the difference in hitpoints for a sz 3 or a sz 21 seemingly unnafected by pain warrior doesnt do a lot , in fact if your a rookie and seemingunafected by pain an aimer still takes you out in 4 hits


now look at it this way

clearly sizes 3-6 are significantly rarer than most others sizes
and clearly sizes 18-21 are as well and along the same curve (if you roll 3 d7) ( i wish i had done the actual math first and figured out sz 17 was involved as well but i didnt , i just used the data i had to form an opinion .

hmmm ill bet runequest used 7 sided dice perhaps someone can find an old set of the rules , since dm was based on runequest), i used to have a copy no clue where it is now )

now further analasys

so every number that has the same number of combinations availavble to it will have the same chance of appearing (more or less )

so since the average size does appear to be just between 11 and 12 that means there is a good chance that it is just 3d7 and as such that puts sz 15-17along the same portion of the curve as size 7-9, and lastly due to where the size breakpoints are in the game it makes for way to many size 15-17 warriors that suffer too much a defense penalty vs the damage bonus or inititive bonus they recieve .

and my only point is if the sizes were genered based on percentages equitable to the actual skill breakpoints in size that exist in order for there to be less crap rollups there should be less sz 15-17 warriors .

yes the actual disribution is perfect from a mathamatical standpoint , but becasue of where the skill breakpoints are it makes for to many larger warriors vs smaller ones . along the skill breakpoints .

if a sz 15-17 was as good as size 7-9 it would be fine , but they simply are not , if they could become as good as a 7-9 at complete max it would be fine but they cannot (dont bring up barehanded aimers or prize modified warriors )

achiving an equitable disribution like a minnimum wit/wl combo or a percentage of exceptional stats is a completly seperate issue .


and no i cant tell you how to fix it , i am only saying that if the number of these size 15-17 warriors can be reduced and the reduction percentage goes to sizes 8-14 we will see a lot less crap warriors and a lot more average ones and few more exceptional ones

looking at all the data its pretty obvious that in relation to the size breakpoints that there are more larger warriors being generated by the arnea replacement program than there should be .
of course all of this would be differnt if the penalties and advantages for size were correctly offset but they are not and they are not likley to ever get fixed .


going by the sizes a much more eqitable distribution in relation to the penalties or bonuses inherent to size would be something like

3-1%
4-2%
5-4%
6-4%
7-8%
8-8%
9-8%
10-8%
11-8%
12-8%
13-8%
14-8%
15-6%
16-6%
17-3%
18-3%
19-3%
20-3%
21-1%


again there are lotsa of ways to change it , i only wished to demonstrate that becasue larger warriors can never be as good as middle or smaller ones (under normal circumstance , that to many are generated .

i dont need to roll up anything to know that based on where the size breakpoints are the disribution unfairly generates to many large warriors

you cannot change the laws of physics captain

but when you apply them incorectly it sure can be a mess and clearly the original design of the sizes wasnt good .

guardian

_________________
im guardian who the f... are you !.
View user's profileSend private messageSend e-mail
pipthetroll
Advanced Master Poster
Advanced Master Poster


Joined: Nov 04, 2002
Posts: 447
Location: In my underwear, in front of my computer

PostPosted: Mon Oct 30, 2006 3:43 pm Reply with quoteBack to top

guardian wrote:

hmmm ill bet runequest used 7 sided dice perhaps someone can find an old set of the rules , since dm was based on runequest), i used to have a copy no clue where it is now )


6d4-3 gets you the same #'s and a bell curve.
so does 2d10+1

I dont recall ever seeing a d7, hehe
View user's profileSend private messageSend e-mailICQ Number
Woody
Grandmaster Poster
Grandmaster Poster


Joined: Nov 01, 2005
Posts: 989
Location: Lake Powell

PostPosted: Mon Oct 30, 2006 9:33 pm Reply with quoteBack to top

The Consortium wrote:
Woody wrote:
I'll try to be blunt.

There is no way to tell if the distribution is right without a big enough random sample size.

How big is big enough?

SZ 3-21 is 19 different "cell" values. You have to have at least 5 show up in your lowest frequency cells (at least 5 SZ 3 and at least 5 SZ 21).

The data shown so far indicates the frequency of SZ 3 and 21 is about 0.5% each (5 out of 1000). The data isn't very useful for a random sample of less than 1000.


In English: Until you've gotten about 1000 random rollups, you can't tell whether or not you're getting hosed.


Woody, thank you very much. Are you sure of this?


The 1000 was a shot from the hip. The goodness-of-fit on a multinomial dist like this will be stingy in yielding conclusions (either way).

If the 3d7 method is valid, I should be able to reduce it some.

I'll crunch some numbers tonight if I have insomnia.
View user's profileSend private message
Woody
Grandmaster Poster
Grandmaster Poster


Joined: Nov 01, 2005
Posts: 989
Location: Lake Powell

PostPosted: Mon Oct 30, 2006 10:09 pm Reply with quoteBack to top

The Consortium wrote:


Woody, thank you very much. Are you sure of this?


I just looked it up.

For what we're trying to do here (I'm still a little confused as to the goal Confused ) the recommended minimum cell count is 5.

In English: If the 3d7 method is valid, it would likely take 1700-2000 rollups to determine if you got hosed.

Of course, this depends on what your definition of "hosed" is...
View user's profileSend private message
pipthetroll
Advanced Master Poster
Advanced Master Poster


Joined: Nov 04, 2002
Posts: 447
Location: In my underwear, in front of my computer

PostPosted: Mon Oct 30, 2006 11:49 pm Reply with quoteBack to top

Woody wrote:

Of course, this depends on what your definition of "hosed" is...


Hosed is having too low of starting wit or will.

Since the size appears to be determined first(thats why its curve favors 12), the other 6 stats have a bell curve situated at the average remaining.

A size 12 has 58/6 = 9 2/3 average for the other stats. Most will be able to get either wit or wil to 15, some will get both, a few will get niether. I'd guess it to be in the 10-20% range, im not gonna figure it out.

As you get bigger it gets worse, a sz 16 has a 9 average for the other stats, a slightly larger % will not be able to hit 15 wit or wil.

A size 4 has an average of 11 on its other stats, a significantly larger % will be able to hit 15 wit and will.

So heres where the hosing really is:

Da replacements have an average size of 12, with the average of other stats being 9 2/3; who knows about the team sheets, since they have been filtered by a person, with the worst removed. Team sheets are less per, and mathimatically get you a better chance of a good replacement.

Here's the real bitch about replacements:

Starting size averages 12, 1/7th of 84. The rest average 9 2/3, 1/6 of 70-average.size. If you randomly added your 14 you'd hit an average of 12 again tho, so it was most likely designed this way on purpose. Problem is more warriors fall on the hosed side of size(- def and parry, diminishing returns on damage), and get that larger % chance of not hitting 15 wit or will.

You can take a sample of team sheet sizes if you want to see for sure that team sheets favor a lower size, I'd almost guarantee it.

What we're asking for is that da replacements give you a slightly better or even the same chance of getting a decent warrior than a team sheet.
View user's profileSend private messageSend e-mailICQ Number
pipthetroll
Advanced Master Poster
Advanced Master Poster


Joined: Nov 04, 2002
Posts: 447
Location: In my underwear, in front of my computer

PostPosted: Mon Oct 30, 2006 11:51 pm Reply with quoteBack to top

Woody wrote:

Of course, this depends on what your definition of "hosed" is...


Hosed is having too low of starting wit or will.

Since the size appears to be determined first(thats why its curve favors 12), the other 6 stats have a bell curve situated at the average remaining.

A size 12 has 58/6 = 9 2/3 average for the other stats. Most will be able to get either wit or wil to 15, some will get both, a few will get niether. I'd guess it to be in the 10-20% range, im not gonna figure it out.

As you get bigger it gets worse, a sz 16 has a 9 average for the other stats, a slightly larger % will not be able to hit 15 wit or wil.

A size 4 has an average of 11 on its other stats, a significantly larger % will be able to hit 15 wit and will.

So heres where the hosing really is:

Da replacements have an average size of 12, with the average of other stats being 9 2/3; who knows about the team sheets, since they have been filtered by a person, with the worst removed. Team sheets are less per, and mathimatically get you a better chance of a good replacement.

Here's the real bitch about replacements:

Starting size averages 12, 1/7th of 84. The rest average 9 2/3, 1/6 of 70-average.size. If you randomly added your 14 you'd hit an average of 12 again tho, so it was most likely designed this way on purpose. Problem is more warriors fall on the hosed side of size(- def and parry, diminishing returns on damage), and get that larger % chance of not hitting 15 wit or will.

You can take a sample of team sheet sizes if you want to see for sure that team sheets favor a lower size, I'd almost guarantee it.

What we're asking for is that da replacements give you a slightly better or even the same chance of getting a decent warrior than a team sheet.
View user's profileSend private messageSend e-mailICQ Number
Woody
Grandmaster Poster
Grandmaster Poster


Joined: Nov 01, 2005
Posts: 989
Location: Lake Powell

PostPosted: Tue Oct 31, 2006 7:52 am Reply with quoteBack to top

pipthetroll wrote:
Woody wrote:

Of course, this depends on what your definition of "hosed" is...


Hosed is having too low of starting wit or will.



Then why not have RSI fix the problem by putting a "floor" (or higher floor if there already is one) on the learn/train rates associated with wit/will.

This will fix the primary problems associated with low wit/will without inventing too many new ones (law of unintended consequences).

Changing the size distribution or minimums in wit/will changes the game dramatically for everyone.
---What if some of the players want higher size rollups because it's fun for them? I know that these rollups aren't as competitive, but neither are pikers, and plenty of pikers are still made on a voluntary basis.

---Changing the typical rollup values affects everything (skill base, physicals, weapon suitability, coordination, etc...)

I won't protest changing the rollup pool, but it seems like we're using a baseball bat to kill a fly.
View user's profileSend private message
Drake
Master Poster
Master Poster


Joined: Oct 27, 2002
Posts: 249
Location: St. Louis, MO

PostPosted: Tue Oct 31, 2006 8:04 am Reply with quoteBack to top

Woody wrote:
Then why not have RSI fix the problem by putting a "floor" (or higher floor if there already is one) on the learn/train rates associated with wit/will.

This will fix the primary problems associated with low wit/will without inventing too many new ones (law of unintended consequences).

Changing the size distribution or minimums in wit/will changes the game dramatically for everyone.
---What if some of the players want higher size rollups because it's fun for them? I know that these rollups aren't as competitive, but neither are pikers, and plenty of pikers are still made on a voluntary basis.

---Changing the typical rollup values affects everything (skill base, physicals, weapon suitability, coordination, etc...)

I won't protest changing the rollup pool, but it seems like we're using a baseball bat to kill a fly.


I think it's mainly a matter of what RSI is likely to actually do. Changes to the base code of the main program (which is where the learning rate code probably is) are far less likely to happen than changing the roll-up generation program.

_________________
Deric (Drake) Page
Currently MIA
Playing since 1988 (North Fork turn 10)
Known primarily for being a wise-ass
****
Success is not the result of spontaneous combustion. You have to set yourself on fire. - Unknown
View user's profileSend private messageSend e-mailYahoo MessengerICQ Number
guardian
Advanced Master Poster
Advanced Master Poster


Joined: Nov 05, 2002
Posts: 334

PostPosted: Tue Oct 31, 2006 3:47 pm Reply with quoteBack to top

rofl have ris do somthing that requires code change

funny how these things go

the original anfd only point i was making was that based on the size breakpoints that the distribution of sizes was wrong becasue some larger warriors in a different size classifacation were to close to the center of the bell curve .

becasue the masth of the way they are generated is in fact perfect across the sizes as the middle size shows somwhere beween 11 and 12

that is all nothing else

any one can see that size 15-17 have the same disribution as 7-9 . and if you dont feel that that is unfair or incorrect well more power to you good luck in those 3 team arenas fighting commision warriors every turn

me i ordered 100 roullups so i can run 100 rookies in tempe again


have a nice day

guardian

_________________
im guardian who the f... are you !.
View user's profileSend private messageSend e-mail
Display posts from previous:      
Post new topicReply to topic
View previous topic Log in to check your private messages View next topic
You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum



Powered by phpBB 2.0.10 © 2001 phpBB Group

Version 2.0.6 of PHP-Nuke Port by Tom Nitzschner © 2002 www.toms-home.com
Forums ©
:: fisubsilver shadow phpbb2 style by Daz :: PHP-Nuke theme by coldblooded (www.nukemods.com) ::