.com.unity Forums

.com.unity Forums (http://forum.shrapnelgames.com/index.php)
-   Dominions 3: The Awakening (http://forum.shrapnelgames.com/forumdisplay.php?f=138)
-   -   Patrolling and Unrest (http://forum.shrapnelgames.com/showthread.php?t=49096)

krpeters August 31st, 2012 10:24 PM

Patrolling and Unrest
 
Well this is interesting.

I invite you all to try this yourselves. Ramp up taxes, then put a patroller with a medium army (say, 30 troops) to catch the rebels. But one twist: give the army orders "guard commander" and watch what happens.

Mine become instantly pathetic. I think there's some sort of combat mechanic involved, which is why flyers and troops with rapid movement are more effective. I wouldn't be surprised if there's a hidden battle happening somewhere.

parone September 1st, 2012 06:57 AM

Re: Patrolling and Unrest
 
pathetic how?

John_Madlock September 1st, 2012 09:34 AM

Re: Patrolling and Unrest
 
I think you are wrong. I set guard commander order, even retreat, and I see no difference.
I even created 2 players game and I set for both 200% tax and patroling commands. I didn't see any battles resolving between turns.

krpeters September 1st, 2012 11:45 AM

Re: Patrolling and Unrest
 
after further experimentation, I may be totally wrong. Just an odd coincidence possibly.

I'm just trying to figure out why sometimes 30 harpies are enough to slaughter 30 rebels and other times they can't kill 5.

momfreeek September 1st, 2012 12:02 PM

Re: Patrolling and Unrest
 
I think its covered here:
http://dom3.servegame.com/wiki/User:...fect_unrest.3F

"Basically, every patroller will catch 0,5 brigand, but the graph is linear (with 30 patrollers you have the same chance to catch, say, 0, 10, 20 or 30 brigands). So, its much more random than if each patroller has 50% chance to succeed."

It also has the calculation to compare patrollers:

"Faster units are better patrollers: an unit with a tactical (battlefield) move of 20 (cavalry) count as 1,3 foot patroller. Flyers count as 2 patrollers. Some units, like the Forester (Man) have a bonus. Mindless units have a penalty."

JonBrave September 6th, 2012 04:12 PM

Re: Patrolling and Unrest
 
Quote:

Originally Posted by momfreeek (Post 810423)
I think its covered here:
http://dom3.servegame.com/wiki/User:...fect_unrest.3F

"Basically, every patroller will catch 0,5 brigand, but the graph is linear (with 30 patrollers you have the same chance to catch, say, 0, 10, 20 or 30 brigands). So, its much more random than if each patroller has 50% chance to succeed."

(Excellent link BTW.) Try as a might, I still can't get this (the "linear", the 0.5 catching not being the same as 50%, basically the whole thing). Care to clarify?

momfreeek September 6th, 2012 05:11 PM

Re: Patrolling and Unrest
 
I did a quick search for "dice bell curve"; the first few paragraphs here seem like a decent explanation:
http://oldguygaming.com/probability-and-dice-rolling

it applies to coin flips as well as dice rolls (a coin being a 2 sided dice). with 30 coin flips there are many combinations with 15 heads + 15 tails but only one with 30 heads.


so given 2 ways of calculating the result:

- flipping 30 coins and adding all the heads
- rolling a single 31 sided dice (0-30)

it could be said that the second method is much more random (as extreme results are far more likely)

jotwebe September 6th, 2012 05:22 PM

Re: Patrolling and Unrest
 
Linear simply means that extreme outcomes are just as likely as middle-of-the-road ones. As opposed to something like a bell curve where the average is also the most common.

So, for example, imagine you have a bunch of 40 militia with and indy commander patrolling away on your overtaxed cap, with a destealth strength of about 50. Say taxes are 150%, so 10 points of unrest. Now your 50 points of destealth mean that up to 50 points of unrest can be patrolled away - but possibly none at all, too. You're rolling a d50.

So on any given turn, there should be a 20% chance that your actual destealth value is below 10, and unrest ensues.

That's from the manual and the wiki. Makes me wonder if having two smaller patrolling groups instead of one big one gave two rolls and so more consistent results?

Edit: ninja'd

John_Madlock September 7th, 2012 04:19 AM

Re: Patrolling and Unrest
 
I think it is even more like reverse bell curve. In my tests extreme results are very common and average results are very few.

momfreeek September 7th, 2012 09:27 AM

Re: Patrolling and Unrest
 
0-7, 8-22, 23-30

8-23 covers the median 50%, even though 8 feels quite extreme to me. It would take a lot of tests to clean out coincidental randomness but I'm not enough of a statistician to quantify that. Results or it never happened!

John_Madlock September 7th, 2012 11:31 AM

Re: Patrolling and Unrest
 
Quote:

Originally Posted by momfreeek (Post 810723)
Results or it never happened!

OK, so here is my test.
Dominium 10, Order 3 (less events), Grow 3. Events rare (even less ugly events XD), Independents 9 (so easy AI Caelum will stuck in his capitol and I have a lot of time). EA Ermor (because has nice all 10 points units).

Patrol: Pontifex (10 precision, 10 AP), 9 Hastatus or Accensus (10 precision, 10 AP). 1 star experience gives +1 precision, I know this, so I check every unit and replace them with fresh troops if needed.
So this is always 10 points patrol

My methodology:
1 - patrol with 200% tax, if no increase/decrease events - note how many kills
2 - patrol with 0% tax, just to drop down unrest to 0. I didn't note kills here, because 0% tax make unrest to 0 and it is zero kills anyway. Those all zeros below (and all kills) are from point 1.
3 - if unrest is 0 - back to step 1.

Results:
9, 4, 0, 8, 0, 9, 0, 1, 9, 3, 10 (wow, so 9 was not the maximum), 7, 0, 0, 10, 0
OK, stop, I'm bored XD
No events happened.

Bad formula, bugged RNG, bad test, my bugged hardware?

momfreeek September 7th, 2012 01:09 PM

Re: Patrolling and Unrest
 
Yeah, that is quite convincing. Out of 16 tests there's 6 results of 0 and only 3 results between 1-6. I'm at a loss..

Its not clear where the data in Lch/Sunray FAQ came from, perhaps a guess based on an even smaller test.

jotwebe September 7th, 2012 02:08 PM

Re: Patrolling and Unrest
 
Well, I did a test with LA Man. I've patrolled one province wih a magister (destealh should be 26) and another with two foresters (destealh two times 16.4, so make that 32) for about 40 turns. Both provinces taxed a 150%, so unrest generated was 10 points every turn. I've assumed unrest reducing destealh power by one point per 2 points of unrest, but that can't be exactly right, since the magister managed to reduce unrest a bit even when it was over 50 points. Generally unrest does seem to impact patrolling negatively, though.

Another problem of course is that while you can see the low rolls, when the patroller misses some unrest, but not how much better than the 10 points needed.

Anyway, I've expected the magister to succeed about 0.6 (16/26) of the time, and the two foresters 0.7 (22/32). What I got was 0.67 for the magister and 0.63 for the foresters. that's close enough to what's expected for me.

tl;dr:

the wiki is right

jotwebe September 7th, 2012 02:19 PM

Re: Patrolling and Unrest
 
Edit: ninja'd again.

Anyway according to the wiki commanders get +20 to prec and move points before division by 20, so your pontife shuld have been wrh 3 points and your 9 dudes 1 each, for 12 vs. 10 points of unrest. Maybe the wiki is wrong about those commander bonuses?

John_Madlock September 7th, 2012 03:52 PM

Re: Patrolling and Unrest
 
I tax 200% to unrest 18-20, so even with 10 kills I had some unrest and I had to low tax to lower unrest to 0 (step 2). That is why I used only 9 dudes, to see real kills. I know that superfluity of kills are not displayed (1000 guys vs 1 unrest will kill only 1 bandit).

Edit:
Lots of 9 and few of 10, maybe this mean some "re-rolls" in this place?
Or maybe troops (9) and commander (1) are tested separately?

krpeters September 9th, 2012 01:00 PM

Re: Patrolling and Unrest
 
So it seems that there are two things going on -- first, the likelihood of finding the rebels, and second, the number that can be killed.

The "extreme" results then are probably caused by the interaction of these two. If you're patrolling with 50+ troops, you probably have enough troops to kill all the rebels they find each round. But you still have a significant risk of not finding any. This creates the tendency for the "all-or-nothing" results we tend to observe.

I guess the best strategy then is to have enough patrollers to take out *two* turns of unrest -- whatever they miss the first turn, they'll likely crush the next.

Soyweiser September 9th, 2012 01:41 PM

Re: Patrolling and Unrest
 
Quote:

Originally Posted by jotwebe (Post 810743)
Edit: ninja'd again.

Anyway according to the wiki commanders get +20 to prec and move points before division by 20, so your pontife shuld have been wrh 3 points and your 9 dudes 1 each, for 12 vs. 10 points of unrest. Maybe the wiki is wrong about those commander bonuses?

The wiki isn't wrong. Your reading comprehension is :). (I wrote that part of the wiki, and tested it using the debuggin tools from dom3, the log files then list the destealth values. Which baffled me for some time, until I figured out that commanders get double their skills).

Commanders get a double bonus. Not +20. So if their prec is 1 they count as having prec 2. Not prec 21.

The patrol bonusses count just as one point of patrol (which means they count as 20 before dividing by 20 ;) ).

jotwebe September 9th, 2012 05:52 PM

Re: Patrolling and Unrest
 
Yeah, I had a raging headache when I was writing that, and it was an off-the-cuff reply based on what I remembered from reading the wiki a couple of hours earlier. So it got a bit muddled.


All times are GMT -4. The time now is 06:43 PM.

Powered by vBulletin® Version 3.8.1
Copyright ©2000 - 2025, Jelsoft Enterprises Ltd.
Copyright ©1999 - 2025, Shrapnel Games, Inc. - All Rights Reserved.