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: stream0pastor
New Today: 5
New Yesterday: 27
Overall: 19589

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
Mannequin
Grandmaster Poster
Grandmaster Poster


Joined: Sep 15, 2002
Posts: 890
Location: East Wenatchee, WA

PostPosted: Sun Oct 29, 2006 11:04 am Reply with quoteBack to top

guardian wrote:
guys

looking at your pool of warriors isn't relevant look only at your dark arena replacements my data comes only from da replacements ,warriors off team sheets are not relevant .


the rollups are still a static pool .


guardian


I run three teams in Mordant (DM-1) and use the Dark Arena on a frequent basis. I went through all of my roll-ups there over the course of the year (44 total) and this was the breakdown:

SZ

3 -- 0
4 -- 1 (2.27%)
5 -- 3 (6.81%)
6 -- 2 (4.54%)
7 -- 1 (2.27%)
8 -- 1 (2.27%)
9 -- 3 (6.81%)
10 -- 2 (4.54%)
11 -- 5 (11.36%)
12 -- 5 (11.36%)
13 -- 3 (6.81%)
14 -- 5 (11.36%)
15 -- 5 (11.36%)
16 -- 3 (6.81%)
17 -- 1 (2.27%)
18 -- 3 (6.81%)
19 -- 1 (2.27%)
20 -- 0
21 -- 0

Total SZ was 530/ 44 (number of roll-ups) = 12.05 average size.

I don't think this proves anything other than the fact that the roll-ups are generated on a curve in regards to their SZ (with the averge SZ being 12). This is the same with new team roll-ups so the only difference between the two seems to be the static pool used for team roll-ups.

Mannequin
View user's profileSend private message
Woody
Grandmaster Poster
Grandmaster Poster


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

PostPosted: Sun Oct 29, 2006 11:37 am Reply with quoteBack to top

LHI wrote:
I looked at my data and for 2097 discreet warriors I had these numbers:

SZ 07 = 4.7%
SZ 08 = 6.2%
SZ 09 = 7.3%
SZ 10 = 9.0%
SZ 11 = 9.7%
SZ 12 = 11.6%
SZ 13 = 11.2%
SZ 14 = 7.9%
SZ 15 = 7.4%
SZ 16 = 6.7 %
SZ 17 = 4.3%

I've only had 11 SZ 3 and 9 SZ 21.


This looks reasonable to me.
View user's profileSend private message
Woody
Grandmaster Poster
Grandmaster Poster


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

PostPosted: Sun Oct 29, 2006 11:39 am Reply with quoteBack to top

Drake wrote:
Woody wrote:
Are all the RU's in your database unique, or are some redundant?

Can you say exactly how many warriors this is?

I'm curious because this distribution can be statistically checked to see how out-of-whack it really is.


I don't have any duplicates. The total size of my warrior pool is 222. The margin of error for that pool is about 8.7% and has an average Standard Error of around 1.4% (assuming I've done my math correct) .

The Terrablood data contained 7701 samples. Some of which are the same warrior training multiple times. However, this only matters if specific Size warriors are training more often than other Sizes and I see no evidence to suggest that. The MoE of that pool is 1.5% and the average SE is about 0.2%.

Also, if you compare the Size distribution of my pool next to the Terrablood data, you'll see that the Size frequencies are pretty close (with the exceptions of Size 15 & 16). I would imagine we'd get similar numbers out of Guardian's pool if we had that data. Wayne's pool is harder to guess since he reports different results. I'd suspect that the only significant differences between Wayne's pool & the others would be around sizes 15 & 16.

If someone has another pool of data for us to analyse, then feel free to point me to it. I suppose that if someone really wanted to do all the work (not me) you could go through and build a database out of all the RU posted to the Roll-Up topic on this site. Of course, that will also have a bias; toward hard-to-design roll-ups. If you're looking for unbiased RU data, you're not going to find it anywhere outside of having RSI send you several hundred DA replacements and posting them all here. So, who has about 400-500 warriors they want to DA all at once to give us some clean data? Razz

However, another thing to consider is that Guardian has had this Size 15 & 16 bias confirmed by a formar RSI programmer, and the frequency of players complaining about the number of high size, low Wit/Will replacements they get has always been high.

There is a lot of evidence to support the theory that the method by which roll-ups are generated is broken. Is there anyone here who thinks it's not?

If we all agree that it is broken, it really doesn't matter by how much. What matters is what we can do to fix it. Several fixes have already been suggested. So, the actual useful thing to discuss at this point would be "which fix do we want employed and how do we get RSI to employ it". Quibbling over just how much the RU pool is out of whack will get us no where.


I was curious about the sample size because there's a test to see if that 16 SZ percentage is an anomaly. (I'll have to look it up but I think is was a chi-squared test).
View user's profileSend private message
keeope
Unchartered Poster


Joined: Sep 18, 2006
Posts: 3
Location: Arkansas

PostPosted: Sun Oct 29, 2006 11:40 am Reply with quoteBack to top

This is an old forum, but still relevant.

I've returned to Duelmasters after a 15+ year break to find many changes in the game. So much so that the warriors and strategies I used to use aren't having much luck with winning. So after two turns, I will have sent 4 warriors to the DA.

The returning roll-ups are crap, even by my 15 year absense's standards. Moderate size, and the best wit/will combo I've seen was a 6/7 respectively.

If this keeps up, my return will be short lived.
View user's profileSend private messageICQ Number
The Consortium
ArchMaster Poster
ArchMaster Poster


Joined: Nov 23, 2002
Posts: 8664
Location: on the golf course, in the garden, reading, traveling, and now Consulting

PostPosted: Sun Oct 29, 2006 2:40 pm Reply with quoteBack to top

Most recent 250 replacements to make calcs easy) - (may have missed a few here and there ...)
SZ-#-%
3-2 (0.Cool
4-6 (2.4)
5-9 (3.6)
6-6 (2.4)
7-12 (4.Cool
8-22 (8.Cool
9-16 (6.4)
10-21 (8.4)
11-16 (6.4)
12-25 (10.0)
13-31 (12.4)
14-27 (10.Cool
15-16 (6.4)
16-9 (3.6)
17-15 (6.0)
18-6 (2.4)
19-5 (2.0)
20-5 (2.0)
21-1 (0.4)

250 warriors
2945 total size
11.8 avg size

_________________
Ye Old Consortium Scribe & Crapmaster 2013, Crapgiver 2014
982 ADM graduates (with 40+ manager IDs) including 101K+ fights and 65K+ wins plus 4 teams with 1000+ wins (Animal Farm DM11 @1479*; Bulldogs DM11 @ 1420; Lenpros DM30 @ 1396; Fandils @ 1305
View user's profileSend private messageSend e-mail
Woody
Grandmaster Poster
Grandmaster Poster


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

PostPosted: Sun Oct 29, 2006 3:03 pm Reply with quoteBack to top

Cleaned up and re-posted.


Last edited by Woody on Sun Oct 29, 2006 9:10 pm; edited 1 time in total
View user's profileSend private message
Woody
Grandmaster Poster
Grandmaster Poster


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

PostPosted: Sun Oct 29, 2006 9:09 pm Reply with quoteBack to top

Drake wrote:



Size 16 does seem to be significantly out of whack.


I finally got around to crunching the numbers.

A chi-squared goodness-of-fit test shows your data from the 7-17 range to be consistent with LHI's given data to a=.05 (barely...the critical value for a=.05 and df=10 is 8.31, the test statistic was 8.30).

In English: the size 16 looks out of whack, but statistically there's only a 1 in 20 chance that your rollup pool was somehow different from LHI's.


-edit-
Upon closer inspection, I believe i've made an error...
I can't research it immediately, but I'll try to look into it when I have some time tonight.

There's lies, damn lies, and then there's statistics...

-edit 2-
The first error was a typo. 8.31 and 8.30 should be 18.31 and 18.30, respectively. This doesn't change the test results.

The second error was the presentation of the conclusion.
A better way of saying it is that we can't be 95% sure that something's out of whack with your distribution compared to LHI's.

Basically, I used a hammer when a wrench was needed.

--------
THE FOLLOWING STATEMENTS ARE TO THE GENERAL D2 COMMUNITY
THEY ARE NOT DIRECTED AT ANY SPECIFIC INDIVIDUAL

Here's a better argument:
Out of 2097 warriors, LHI got 140 that were SZ 16.
Out of 222 warriors, Drake got 23 that were SZ 16.

If Drake gets 1875 more warriors, and 6.24% of them are SZ 16, nobody would think that was "out-of-whack"...and that would result in Drake having 140 SZ 16 warriors, just like LHI.

It is going to take a great deal of DAing before the law of large numbers catches up with most of us.

I don't see very many complaints when people get lucky on the front end, rather than unlucky.

I'll probably offend many on this forum by saying this, but this is the nature of gambling.

Unless you're willing to DA 2000 warriors, you might not see your personal distribution get nice and bell-curved.

If you want your chance of bad luck reduced, everyone else's chance of bad luck gets reduced. You may have better warriors, but so will everybody else. It accomplishes little.

But, that being said, it also does little harm.
View user's profileSend private message
guardian
Advanced Master Poster
Advanced Master Poster


Joined: Nov 05, 2002
Posts: 334

PostPosted: Mon Oct 30, 2006 4:31 am Reply with quoteBack to top

Mannequin wrote:
guardian wrote:
guys

looking at your pool of warriors isn't relevant look only at your dark arena replacements my data comes only from da replacements ,warriors off team sheets are not relevant .


the rollups are still a static pool .


guardian


I run three teams in Mordant (DM-1) and use the Dark Arena on a frequent basis. I went through all of my roll-ups there over the course of the year (44 total) and this was the breakdown:

SZ

3 -- 0
4 -- 1 (2.27%)
5 -- 3 (6.81%)
6 -- 2 (4.54%)
7 -- 1 (2.27%)
8 -- 1 (2.27%)
9 -- 3 (6.81%)
10 -- 2 (4.54%)
11 -- 5 (11.36%)
12 -- 5 (11.36%)
13 -- 3 (6.81%)
14 -- 5 (11.36%)
15 -- 5 (11.36%)
16 -- 3 (6.81%)
17 -- 1 (2.27%)
18 -- 3 (6.81%)
19 -- 1 (2.27%)
20 -- 0
21 -- 0

Total SZ was 530/ 44 (number of roll-ups) = 12.05 average size.

I don't think this proves anything other than the fact that the roll-ups are generated on a curve in regards to their SZ (with the average SZ being 12). This is the same with new team roll-ups so the only difference between the two seems to be the static pool used for team roll-ups.

Mannequin


your data shows exactly what i am saying that sz 15 and 16 are as likely as sz 8-14 which is the only claim i have made and clearly based on the sz breakpoints sz 15-16 should be less likely than sz 8-14

guardian

_________________
im guardian who the f... are you !.

Last edited by guardian on Mon Oct 30, 2006 4:50 am; edited 1 time in total
View user's profileSend private messageSend e-mail
guardian
Advanced Master Poster
Advanced Master Poster


Joined: Nov 05, 2002
Posts: 334

PostPosted: Mon Oct 30, 2006 4:48 am Reply with quoteBack to top

The Consortium wrote:
Most recent 250 replacements to make calcs easy) - (may have missed a few here and there ...)
SZ-#-%
3-2 (0.Cool
4-6 (2.4)
5-9 (3.6)
6-6 (2.4)
7-12 (4.Cool
8-22 (8.Cool
9-16 (6.4)
10-21 (8.4)
11-16 (6.4)
12-25 (10.0)
13-31 (12.4)
14-27 (10.Cool
15-16 (6.4)
16-9 (3.6)
17-15 (6.0)
18-6 (2.4)
19-5 (2.0)
20-5 (2.0)
21-1 (0.4)

250 warriors
2945 total size
11.8 avg size



average size has nothing to do with it your data also tends to show there is more larfer warriors than there should be based on the sz breakpoints

i do find it odd that you have gotten so many sz 17 though ,


my contention has always been and wil not change that too many larger warriors (sz 15-16 , you show 17 ) are being generasted , this was most likley done deliberatley ecasue at the time the uber sizes were considered to be 3-7 add yours and mannequins together and we see it get worse not better , which is a pretty good indicator


the sample i used

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

now add all 3 and tell me it doesnt show it

the only contention i am making is that if the larger warriors i said 15-16 were less likley and it put those rolluops no longer at 15 -16 in the 7-14 distribution we would get more runnable rollups .

the method whatever it is 3d(7) if its as simple as that would automatically send to distribute average sz of 12 but it also would make the extremes about what were seeing i had no vidence to support sizes other than 15-16 add all three of these and it show 15-17 as an issue


keep adding rollups i dont think more will show it changing much

guardian

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


Joined: Nov 05, 2002
Posts: 334

PostPosted: Mon Oct 30, 2006 5:10 am Reply with quoteBack to top

added together


3-3
4-8
5-14
6-15
7-18
8-28
9-27
10-29
11-29
12-40
13-41
14-41
15-33
16-20
17-20
18-10
19-8
20-6
21-2

this data pretty clearly shows that the major portion of the curve runs from sz 7-17 odd isn't it that exactly 4 numbers above and below that curve exist (yes I'm still sarcastic ) .

so the average of between sz 11-12 is correct under the current formula my point is that if we saw less of these (now 15-17)larger warriors that average size would drop to between 8-10 and we would get more runnable warriors from that alone so we would have much less dark arena frustration .


thank you for supplying the data

the sz 17 in Wayne's does still surprise me some i really thought it would only be sz 15 and 16

guardian

_________________
im guardian who the f... are you !.

Last edited by guardian on Mon Oct 30, 2006 1:36 pm; edited 1 time in total
View user's profileSend private messageSend e-mail
blackstorm
ArchMaster Poster
ArchMaster Poster


Joined: May 07, 2006
Posts: 1389
Location: Close enough to GREENLAND to watch it melt.

PostPosted: Mon Oct 30, 2006 5:12 am Reply with quoteBack to top

I have personaly seen ALOT of 18 size warriors at the bottom ranks. of my last 5 were 18.

_________________
BLACKSTORM
Champions of Steel--8,106,105
Sisters of Smash--28
Champions of Rune, Eternal Souls--31
Black Storm--47
Chaos Knights--52
Eye of the Storm--54,104
Eve of Destruction--83
Dark Enclave--81 and more....
View user's profileSend private messageSend e-mailYahoo MessengerMSN Messenger
The Consortium
ArchMaster Poster
ArchMaster Poster


Joined: Nov 23, 2002
Posts: 8664
Location: on the golf course, in the garden, reading, traveling, and now Consulting

PostPosted: Mon Oct 30, 2006 7:20 am Reply with quoteBack to top

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?

_________________
Ye Old Consortium Scribe & Crapmaster 2013, Crapgiver 2014
982 ADM graduates (with 40+ manager IDs) including 101K+ fights and 65K+ wins plus 4 teams with 1000+ wins (Animal Farm DM11 @1479*; Bulldogs DM11 @ 1420; Lenpros DM30 @ 1396; Fandils @ 1305
View user's profileSend private messageSend e-mail
Woody
Grandmaster Poster
Grandmaster Poster


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

PostPosted: Mon Oct 30, 2006 7:36 am Reply with quoteBack to top

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.
View user's profileSend private message
Stik
ArchMaster Poster
ArchMaster Poster


Joined: May 06, 2003
Posts: 2118
Location: Frozen Tundra of North Dakota

PostPosted: Mon Oct 30, 2006 7:48 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?


Good point. The reduction in large sized warriors is wanted/needed so that those extra points will hopefully be randomly dumped into some other category we can use to make at least a halfway runnable warrior.

Adding the modifier to make the WT/WL combo at least equal to, what was it, 15, or reroll one stat until it is, would perhaps take care of the "problem" on its own.

Most of us don't mind a big guy as long as it has some redeeming, runnable quality, short or long term. Look at everyone (or at least me) still drooling over the freaky size 20/21 guys posted recently!
I'm still in support of less big guys, but I think Consortium's on the right track with what he's getting at, too.

Sure, everyone would start getting more decent replacements, Drake, and as long as it affects everyone's chances, the playing field is still level. But now everyone's a little, maybe a LOT, more enthused about playing more warriors, especially in basic! If even just a handful of managers like Guardian decide to start up a couple full time arena teams again, it'll pay off for RSI. More warriors running in basic means more getting to ADM and potentially continuing on, boosting their revenue stream!

_________________
Stik
Grim Fandango (21)
and various DM82/84/100 projects
View user's profileSend private message
Stik
ArchMaster Poster
ArchMaster Poster


Joined: May 06, 2003
Posts: 2118
Location: Frozen Tundra of North Dakota

PostPosted: Mon Oct 30, 2006 7:49 am Reply with quoteBack to top

Whoops, double post! Embarassed Mad

_________________
Stik
Grim Fandango (21)
and various DM82/84/100 projects

Last edited by Stik on Mon Oct 30, 2006 10:27 am; edited 1 time in total
View user's profileSend private message
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) ::