better leader - combat results (Full Version)

All Forums >> [New Releases from Matrix Games] >> Empires in Arms the Napoleonic Wars of 1805 - 1815



Message

borner -> better leader - combat results (4/3/2009 4:41:00 AM)

I was playing a game against the AI, and have seen two battles now where France under Nappy attacks a Prussian and then Austrian stack with no leader. (Yes, this was intentional.) The Austrians rolled 3 zeros. 0(-1). Prussia never rolled above a 1. I also have seen a live game early on where Nappy hammered the Prussians with equal chits ( counter attack vs Esc Counterattack), in that one Prussia had a bad 1805 leader, and did not roll higher than a 1. France lost 5 factors, Prussia over 40. Again the dice were terrible. I am no stats expert, but what are the odds or rolling a 1 three times in a row?

the game is supposed to have a max +1/-1 mod to the total die rolls. However, this really seems like a trend. Has anyone else seen very out of balance die rolls where one leader was so much better than another?




mr.godo -> RE: better leader - combat results (4/3/2009 7:30:08 AM)

one in 216. but not impossible.

try posting all the results from a game.



mr.godo -> RE: better leader - combat results (4/3/2009 7:43:38 AM)

open your logfile. parse for "roll". find all the numbers rolled for.

i just played a solo game with 2406 rolls from 1 to 20. This is interesting.

Count of Roll
Roll   1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 Grand Total
Total 541 217 219 213 211 200 131 139 136 166 23 26 13 27 18 16 27 30 29 24 2406

This is the *basic* unmodified roll. Why does "1" show up so often???



NeverMan -> RE: better leader - combat results (4/3/2009 12:51:26 PM)

I'd be pretty surprised if Marshall used anything other than a simple "rand()%" function for this but then again, maybe I wouldn't be.



borner -> RE: better leader - combat results (4/3/2009 1:08:37 PM)

one in 216, then how about 9 rolls all below 2 before modifications, to account for what I have seen up to this point in other games? this suuuure sounds like a bug; as in the system giving too much credit / traking too much away for leadership differences. by the way - I re-ran that turn and the same thing happened again.



obsidiandrag -> RE: better leader - combat results (4/3/2009 5:25:02 PM)

I have actually brought better leaders and 2 or 3 times the number of troops the AI will have and still LOST battles due to the fondness of the 1 on the dice...  So it is not tied to leaders or troops, my guess is it just likes to roll ones!!




mr.godo -> RE: better leader - combat results (4/3/2009 6:22:22 PM)

Reviewing combat rolls would be problematic since you have to check and record each one. But if you parse through the game log, you can search each line for a forage roll. I have looked at three different games as well as two faked games and it looks like you will get a 1 twice as likely as any other number. I fail to understand why this would not be an even distribution unless the value of the roll is not 1d6 for foraging with modifications afterwards.

if there is a rolling problem with foraging, i would guess that battle rolls would be compromised as well. anyone not having a problem with die rolls?



Jimmer -> RE: better leader - combat results (4/3/2009 7:43:22 PM)


quote:

ORIGINAL: mr.godo

open your logfile. parse for "roll". find all the numbers rolled for.

i just played a solo game with 2406 rolls from 1 to 20. This is interesting.

Count of Roll
Roll   1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 Grand Total
Total 541 217 219 213 211 200 131 139 136 166 23 26 13 27 18 16 27 30 29 24 2406

This is the *basic* unmodified roll. Why does "1" show up so often???


Because the program's random function is not random. I reported this problem almost a year ago, and everybody who responded said it was because of a failing in the logging or my math. I still believe it is a problem, especially with forage rolls. 1s come up >10 times as often as they should.



mr.godo -> RE: better leader - combat results (4/3/2009 8:30:22 PM)

there is obviously no math involved here.



NeverMan -> RE: better leader - combat results (4/3/2009 8:40:52 PM)

It would be nice for Marshall to chime in since I'm not sure how this could be a problem.

If he's simply seeding a rand function and calling that function, there's no reason for it to come up 1s 10x more often than any other number.

If he tried to implement his own psuedo-random algorithm then yes, it's probably all jacked up.



easterner -> RE: better leader - combat results (4/4/2009 1:06:17 AM)

When does need for d10 & d20 come up?



Ted1066 -> RE: better leader - combat results (4/4/2009 9:09:00 AM)

where is zero in all these rolls? Is it being lumped in with 1? I'm wondering if the log file is attributing a zero in combat as a 1 in the log. If there were an even split between 0 and 1 for the 541 1 rolls (i.e ~270 each), that is not an abnormal distribution (it has variance of roughly 5-10% and is less than 1 standard deviation - not bad by stats rules IF 0's are being counted as 1's in the log).

As to the 10 to 20, that shows up in reference to the minor diplomacy rules I think. You have to roll higher on d10 plus your bonuses vs the comp rolling a d20.

Cheers,

Ted



mr.godo -> RE: better leader - combat results (4/4/2009 2:54:44 PM)

that must be when the die falls off the table. i always roll into box lids to prevent that from happening.

0-6 is technically 1d7. i have seen 0-7 for combat. that would be 1d8. the rules make no mention of these varied dice substitutions.

i am sure there is an explanation somewhere for the inconsistency.




borner -> RE: better leader - combat results (4/4/2009 4:30:58 PM)

this would be something I would suggest people keep an eye out for. the total adjustment to the roll is supposed to be +1/-1. sure seems that the system wants to take it further.



NeverMan -> RE: better leader - combat results (4/4/2009 4:42:03 PM)

Zero should not be a problem.

The random function probably looks something like: (rand()%6)+1, with the rand() function being seeded by time something like: srand(time(NULL)).

I've very curious as to how Marshall decided to implement this.



Ted1066 -> RE: better leader - combat results (4/5/2009 4:58:44 AM)


quote:

ORIGINAL: mr.godo

that must be when the die falls off the table. i always roll into box lids to prevent that from happening.

0-6 is technically 1d7. i have seen 0-7 for combat. that would be 1d8. the rules make no mention of these varied dice substitutions.

i am sure there is an explanation somewhere for the inconsistency.



mr.godo

0-7 occurs from a 1d6 die roll being modified by +/-1, depending on combat circumstances (good vs bad leaders mainly, but cav superiority can supply a side with +1).

Ted



borner -> RE: better leader - combat results (4/5/2009 2:49:41 PM)

Ture, but the the max a roll is supposed to be able to be modified is +1/-1



mr.godo -> RE: better leader - combat results (4/5/2009 5:16:35 PM)

ted,
so the combat reports the modified die roll, not some funky 0-7 1d8? it is purely the reporting of a modded die roll. the original roll is 1d6 based. but you suggest the forage could have an unmodified 0 die roll? the numbers i pulled from the log are unmodified. why would there be a 0-6 1d7 roll?

if there was a log addition for the battle results, we could see for ourselves if there was a pattern in the die rolling.





easterner -> RE: better leader - combat results (4/5/2009 6:29:55 PM)

I would assume the mod is ignored in calculation otherwise 7 would be skewered result. 2-6, 7-10, 11-20 are all within probability of one another with only '10' a little high  +30 over 7-9.  But nothing like '1' 



Mardonius -> RE: better leader - combat results (4/6/2009 1:48:02 PM)


quote:

ORIGINAL: Jimmer


quote:

ORIGINAL: mr.godo

open your logfile. parse for "roll". find all the numbers rolled for.

i just played a solo game with 2406 rolls from 1 to 20. This is interesting.

Count of Roll
Roll   1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 Grand Total
Total 541 217 219 213 211 200 131 139 136 166 23 26 13 27 18 16 27 30 29 24 2406

This is the *basic* unmodified roll. Why does "1" show up so often???


Because the program's random function is not random. I reported this problem almost a year ago, and everybody who responded said it was because of a failing in the logging or my math. I still believe it is a problem, especially with forage rolls. 1s come up >10 times as often as they should.


There was an issue in earlier versions with autoforages being reported as "1"s. Moreover, I have recently seen forages that were manually done but assured being reported as "1"s. Therefore, to get a representative data set is cumbersome. One must extract all autoforages or automatic success forages. This data set will give a result much closer to the expected mean.

Also, please note that the above data set extracted from Mr. Godo's sample takes input from multiple variables (1d6, 1d10, 2d6, and 1d20) each with variable modifiers so can not return a statistically viable result unless parsed out. Time consuming, but doable.



Marshall Ellis -> RE: better leader - combat results (4/6/2009 2:01:34 PM)

FYI: Not all rolls are rolls. Mardonius is correct. Autoforages were always being reported as "1s" so tracking all rolls would be difficult and would reveal little since they are not all rolls.



NeverMan -> RE: better leader - combat results (4/6/2009 3:46:07 PM)


quote:

ORIGINAL: Marshall Ellis

FYI: Not all rolls are rolls. Mardonius is correct. Autoforages were always being reported as "1s" so tracking all rolls would be difficult and would reveal little since they are not all rolls.



Well, this doesn't seem like a good idea since it makes possible bugs harder to decode. Why not make them "-1" or some value specific to "auto forage" so that we (your playtesters) can help you debug this game better?



Marshall Ellis -> RE: better leader - combat results (4/6/2009 4:49:02 PM)

Actually, I believe they are being reported as "Autoforages" in 1.06 so this should go away.



mr.godo -> RE: better leader - combat results (4/6/2009 9:15:48 PM)

fine. the log data is crap and cannot be used for analysis. a die roll may be reported where no roll is made or it may be modified or unmodified. makes sense. one more place to avoid.

so, can anyone say definitively whether this game can roll dice fairly?



Thresh -> RE: better leader - combat results (4/6/2009 10:48:33 PM)

quote:

so, can anyone say definitively whether this game can roll dice fairly?


Seems to be par for the course where most computerized die rollers are concerned...

Thresh



no_dice -> RE: better leader - combat results (4/6/2009 11:32:31 PM)

Hmm, sounds like the "random" number system they use in Las Vegas. :)



Marshall Ellis -> RE: better leader - combat results (4/7/2009 1:47:00 AM)


quote:

ORIGINAL: mr.godo

fine. the log data is crap and cannot be used for analysis. a die roll may be reported where no roll is made or it may be modified or unmodified. makes sense. one more place to avoid.

so, can anyone say definitively whether this game can roll dice fairly?



Yes. I can :-)



NeverMan -> RE: better leader - combat results (4/7/2009 1:56:36 AM)

mr. godo,

You should have a better look at what's going on in 1.06 when the numbers aren't all skewed by the extra 1s. You can then take the log and find out the percentages for each number over a large dataset.



Marshall Ellis -> RE: better leader - combat results (4/7/2009 3:22:11 PM)

That would reveal more accurate results.



mr.godo -> RE: better leader - combat results (4/7/2009 6:10:27 PM)

looking forward to it. am working strictly on solo games as the pace for multi is too slow for me and i enjoy shellacking the ai.



Page: [1] 2   next >   >>

Valid CSS!




Forum Software © ASPPlayground.NET Advanced Edition 2.4.5 ANSI
2