Bay 12 Games Forum

Please login or register.

Login with username, password and session length
Advanced search  
Pages: [1] 2

Author Topic: Military Training EXP Analysis  (Read 25993 times)

Psieye

  • Bay Watcher
    • View Profile
Military Training EXP Analysis
« on: May 24, 2011, 06:25:24 pm »

Ok so following my previous thread, I will now study how Student/Teacher skills affect military training EXP rates. All training will not be subject to insta-acceleration from designations spam. I have 6 dwarves arranged into 3 squads as follows:

Squad 1:
Edem - Teacher 5, Student 5
Kadol - Teacher 5, Student 5

Squad 2:
Rimtar - Teacher 5, Sword 5
Avuz - Observer 5, Concentration 5

Squad 3:
Meng - Teacher 5, Sword 5
Udib - Student 5, Fighter 5

I will only record the progress of the Fighting and Swordsdwarf skills as a measure of how fast they're learning unless some other skill is the subject of demonstrations.

==================

10 Granite

11 Granite

12 Granite

13 Granite

14 Granite

15 Granite

16 Granite

17 Granite

18 Granite

19 Granite

20 Granite

21 Granite

22 Granite

23 Granite

24 Granite

I need to run a second batch of experiments to see if Leader and Organiser affect the training rates. I'll draw my conclusions after that, but first I need sleep.
« Last Edit: May 24, 2011, 06:26:57 pm by Psieye »
Logged
Military Training EXP Analysis
Congrats, Psieye. This is the first time I've seen a derailed thread get put back on the rails.

Psieye

  • Bay Watcher
    • View Profile
Re: Military Training EXP Analysis
« Reply #1 on: May 24, 2011, 06:27:09 pm »

In this second experiment, I will only pay attention to the Swordsdwarf EXP gains from Sparring and not explicitly record gains from Drills. Demonstration gains will be recorded as before, but will be abbreviated to how many "lessons" were given and how much was gained. All Teacher/Student/Concentration and the teacher's Taught Skill EXP gains were as established before (+10/+10/+10/+5 per lesson).


Squad 1
Lor 'S1a': Teacher 5, Sword 5
Libash 'S1b': Student 5, Conc 5

Squad 2
Logem 'S2a': Org 5, Sword 5
Udil 'S2b': Student 5, Leader 5

Squad 3
Ral 'S3a': Teacher 4, Org 4, Sword 2
Goden 'S3b': Leader 5
Morul 'S3c': Nothing

----------------

11 Granite
12 Granite
13 Granite
14 Granite
15 Granite
16 Granite
17 Granite
18 Granite
19 Granite
20 Granite
21 Granite
22 Granite
23 Granite
24 Granite
25 Granite
26 Granite
27 Granite
28 Granite
« Last Edit: May 25, 2011, 06:52:24 am by Psieye »
Logged
Military Training EXP Analysis
Congrats, Psieye. This is the first time I've seen a derailed thread get put back on the rails.

Sir Leonard III

  • Bay Watcher
    • View Profile
Re: Military Training EXP Analysis
« Reply #2 on: May 24, 2011, 08:29:26 pm »

Very interesting, thanks for the information.
Logged
Hush little baby, listen to my song,
Mamma's gonna tan you a troll leather thong.
And if that thong don't menace with spikes,
Mamma's gonna kill you whatever you like...

krenshala

  • Bay Watcher
    • View Profile
Re: Military Training EXP Analysis
« Reply #3 on: May 24, 2011, 09:57:57 pm »

I believe Organizer just affects the speed at which the event (demonstration, training, etc) is set up and started (i.e., how long the students have to wait before the training actually begins).
Logged
Quote from: Haspen
Quote from: phoenixuk
Zepave Dawnhogs the Butterfly of Vales the Marsh Titan ... was taken out by a single novice axedwarf and his pet war kitten. Long Live Domas Etasastesh Adilloram, slayer of the snow butterfly!
Doesn't quite have the ring of heroics to it...
Mother: "...and after the evil snow butterfly was defeated, Domas and his kitten lived happily ever after!"
Kids: "Yaaaay!"

JmzLost

  • Bay Watcher
    • View Profile
Re: Military Training EXP Analysis
« Reply #4 on: May 24, 2011, 10:26:45 pm »

Jumping in with some early analysis of my own.

Teacher, Student, and Concentration all seem to increase by a fixed amount for demonstrations.  The teacher gains a fixed amount in the skill being taught (half as much as they gain in Teacher), while the student gains a variable amount based on all 3 skills.  I wonder if the teacher-dwarf would get improved gains for teaching more than one dwarf at a time?

Observer trains during sparring (and real combat), and presumably gives a boost to EXP gains.  Can't be proven without detailed combat logs and more detailed tracking of EXP gains, though.

Conclusion-jumping:  Good teachers should go in the military early, good students/concentrators should join later.  Observation makes sparring (and combat) a better learning experience.  This was something I already believed, but now there is SCIENCE! to back it up.

@krenshala:  The point is to test our beliefs.  Sometimes we are proven wrong.  Just because "everybody knows" doesn't mean we're right.

Regarding Leadership, I've only ever seen gains on militia captains.  This doesn't mean other dwarfs can't gain it, I've just never seen it.  Having the leader of a squad gain Leadership makes sense, but sense and DF have a very tenuous relationship. 

JMZ
Logged
Also, obviously, magma avalanches and tsunamis weren't exactly a contingency covered in the mission briefing.
I can assure you that Ardentdikes is not the first fortress to be flooded with magma. What's unusual is that we actually meant to flood it with magma.

Psieye

  • Bay Watcher
    • View Profile
Re: Military Training EXP Analysis
« Reply #5 on: May 25, 2011, 07:38:20 am »

So first, sparring - I don't see any conclusive evidence that Observer boosts EXP gained from sparring. Definitely it goes up while sparring, but it doesn't seem to have an effect on learn rates. The distribution of EXP gains in weapon skill from one day of sparring had an average of ~73 with a standard deviation of 40 (i.e. it was a wide distribution).

Leadership didn't do anything for EXP gain rates.

Organiser didn't have any noticable effect on how fast the organisations were finished but maybe there'd be a difference at Legendary vs Novice. Neither did it affect how many lessons were given per day. Something I did notice at the very end was that a 3-man squad was simultaneously sparring in a 1v2 match, something I've never seen before (and I look at quite a few sparring logs). Possibly this is what Organiser does - more testing is needed though I'm not sure if I'll get to it quite so fast.

So, as has been mentioned before: Lessons give +10 in Teacher/Student/Concentration and +5 in the Taught Skill for the teacher. How much students gain is what's changed by all 4 of the above skills - no matter how good the teacher is at teaching and how good the student is at learning, if the teacher doesn't know anything about the subject matter then the student can only learn ~15 EXP per lesson. This did not change even in the 3-man squad where the teacher was giving lessons to 2 students at once.


Teacher 5, Student 5, Concentration 5, Teacher's Sword 5:
Average gain per lesson: 56
Standard Deviation: 11

Teacher 5, Student 5, Teacher's Sword 5:
Average gain per lesson: 36
Standard Deviation: 4

Teacher 5, Concentration 5, Teacher's Sword 5:
Average gain per lesson: 32.5
Standard Deviation: 5.5

Student 5, Teacher's Sword 5:
Average gain per lesson: 29
Standard Deviation: 12

Teacher 4, Teacher's Sword 2:
Average gain per lesson: 25
Standard Deviation: 5.5

Teacher's Fighter 5:
Average gain per lesson: 22
Standard Deviation: 4


So, each of the 4 skills mentioned above have some weight in how much EXP is gained per lesson. But I felt something wasn't quite right, S3b and S3c had fairly consistently different gain rates in Sword demonstrations. Then this page came into my memory. S3b had great analytical ability but S3c had great focus. Maybe that explains why S3c has better gains overall from demonstrations.

This is a tiny handful of data. Science (and thus, !!SCIENCE!!) wants cross-checking and further statistics (more data points) to consolidate these existing conclusions. Someone could probably try fiddling around in Runesmith or the raws to make sure all the attributes are equalised to make this a fair test.

Next I will investigate how live training (against animals) gives out EXP.

===================

Spoiler: "Live Combat EXP" (click to show/hide)

Thus a steady stream of live combat training for an embark teacher/manager in the early years of a fort might be the most beneficial for good organised training later on. I'm going to make a fort to test this out.
« Last Edit: May 25, 2011, 08:14:58 am by Psieye »
Logged
Military Training EXP Analysis
Congrats, Psieye. This is the first time I've seen a derailed thread get put back on the rails.

utuki

  • Bay Watcher
    • View Profile
Re: Military Training EXP Analysis
« Reply #6 on: May 25, 2011, 08:22:50 am »

http://i.imgur.com/mWRbZ.png
This might be useful (its somewhere on magmawiki but i cant find it there).
I did some similar testing about month ago, didnt get any solid results. It might be really good idea to set fixed attributes in raws.
Logged

Suihanki

  • Bay Watcher
    • View Profile
Re: Military Training EXP Analysis
« Reply #7 on: May 25, 2011, 10:17:03 am »

This is really interesting. Thanks for the research.
Logged
Think of it as Jenga: Mines of Moria edition, and you're not too far off.

Psieye

  • Bay Watcher
    • View Profile
Re: Military Training EXP Analysis
« Reply #8 on: May 30, 2011, 10:46:55 am »

So after making a fort based on these findings... the Organiser skill does have an effect on how fast formal classes start but it's hardly essential. At no skill they take about a day to get demonstrations going whereas at Master Organiser they took around 1/4 of that time. I didn't get to test out whether Organiser helps out with arranging bigger sparring sessions because most of the training time got spent on teaching not sparring.

Sparring can occur between two soldiers of fairly different skill but it's rare. They'd much rather do some demonstrations when there's a skill difference. Problem is, what skill they study is random and they'll waste a lot of time teaching less important skills like punching or biting.

I got to see a Teacher 7/Student 7/Concentration 7 demonstration underway when the teacher had Lv 12 in the taught skill. That was an average of ~60 EXP per lesson. Given you're unlikely to find a dwarf (in the early years) with good teacher skill and excellent levels in all combat skills, trying to make some brilliant army instructor is impractical (i.e. megaproject grade). I suspect when Toady gets to work on children there'll be a School system where dwarves can learn Teacher/Student/Concentration skills without having to sit around in a squad. Until then, trying to train up recruits into killing machines primarily through demonstrations is impractical.

In a previous fort, I had 2 soldiers get to Lord status (Lv 11 weapon skill I think) within about a year of sparring. There, they spent most of their time sparring instead of demonstrating as they both started out knowing nothing. Any demonstrations were simply to catch the other up on what they had learnt during sparring so their EXP differences in all military skills were minimised.

Immigrants tend to come with an evenly spread skill set if they get military skills. This is counter to how your soldiers will develop (lopsided towards weapon/Fighter and defensive skills going up much slower). While it would be nice if said migrants gave dedicated lecture courses on Dodge and such, they end up trying to learn your squad's weapon skill half the time and the other half is spent teaching something useless like Biter or Kicker.


So to conclude: demonstrations are a good way to keep your dwarves on an even footing - impractical for trying to teach a greenhorn how to fight like a master. Like all other skills in the game, dwarves prefer to learn by doing than by being shown. Sparring at least keeps a 2-man squad close in skill level so they can consistently rise to legendary status. The alternative is to never rely on formal group training and do all EXP harvesting from live combat (or danger rooms) - preferably against animals in confined spaces (so they have nowhere to run) while wielding training weapons. Further experimenting required to see how well a large squad of equally no-skill dwarves train in groups.
Logged
Military Training EXP Analysis
Congrats, Psieye. This is the first time I've seen a derailed thread get put back on the rails.

darkflagrance

  • Bay Watcher
  • Carry on, carry on
    • View Profile
Re: Military Training EXP Analysis
« Reply #9 on: May 30, 2011, 12:39:20 pm »

If the above is the case, this strategy might be ideal.

Start a fort with two military dwarves. Use them to mine out the initial fort and meanwhile set up farming and cooking. When the other  dwarves can handle the rest of the fort, they start training. You should probably start the fort with a ton of copper/tetrahedrite ore (depending on what's available) so you can start making copper arms immediately (but make shields from leather or wood).

Both of the military dwarves have two Teacher and identical mililtary skills, perhaps 2 armor, 2 shield, 2 fighter, 2 axe. They spar against each other and teach each other when one grows faster than the other. They fight whatever wildlife appears on the map. By the time the first few migrants arrive, they should be skilled enough to be teachers of other dwarves in their own right. From there, you multiply your military. The lives of your trainers are to be valued beyond those of other dwarves, because they are the key to a sustainable army.
Logged
...as if nothing really matters...
   
The Legend of Tholtig Cryptbrain: 8000 dead elves and a cyclops

Tired of going decades without goblin sieges? Try The Fortress Defense Mod

celem

  • Bay Watcher
    • View Profile
Re: Military Training EXP Analysis
« Reply #10 on: May 30, 2011, 12:51:50 pm »

This is how I run longterm games in high turnover situations like a FD game.  Candy armour goes to teachers not killers :)
Logged
Marksdwarf Pillboxes
I wish I had something cool to say about this.  Because it's really cool.

Reelyanoob

  • Bay Watcher
    • View Profile
Re: Military Training EXP Analysis
« Reply #11 on: May 30, 2011, 12:55:10 pm »

Watchin dis
Logged

Noir

  • Bay Watcher
  • Dwarven Evil Overlord
    • View Profile
Re: Military Training EXP Analysis
« Reply #12 on: May 30, 2011, 01:55:59 pm »

I guess I'm lucky then, my Rambo dwarf is also the best teacher. The first suit of masterwork candy is all his!

Anyway, great, informative post - thank you very much.
Logged
Naked dwarves remove the need to produce more clothing, which means more of your pig tails can be brewed into booze.
I think this is less a problem and more an expression of dwarven priorities.

Psieye

  • Bay Watcher
    • View Profile
Re: Military Training EXP Analysis
« Reply #13 on: May 30, 2011, 02:13:51 pm »

Both of the military dwarves have two Teacher and identical mililtary skills, perhaps 2 armor, 2 shield, 2 fighter, 2 axe. They spar against each other and teach each other when one grows faster than the other. They fight whatever wildlife appears on the map. By the time the first few migrants arrive, they should be skilled enough to be teachers of other dwarves in their own right. From there, you multiply your military. The lives of your trainers are to be valued beyond those of other dwarves, because they are the key to a sustainable army.
Suboptimal, because skill development is lopsided. Drop the weapon and fighter skills and put all the points into defensive skills. They'll learn weapon and fighter very quickly - 4 times faster than dodging per action and they get a lot more weapon/melee actions than dodging actions (they parry too, which counts as a weapon action). Making them attack wildlife is a double edged sword as while they develop skill even faster that way, they pick up useless skills like Biter. Thus they end up wasting time teaching those skills instead of the defensive skills and weapon skills you want them to teach.

I'm going to try an INVADERS:OFF fort to test out a new theory - specialised teachers.
Logged
Military Training EXP Analysis
Congrats, Psieye. This is the first time I've seen a derailed thread get put back on the rails.

DrKillPatient

  • Bay Watcher
  • The yak falls infinitely
    • View Profile
Re: Military Training EXP Analysis
« Reply #14 on: May 30, 2011, 02:19:28 pm »

It seems that teacher/student skills are the biggest part of training for me. I embark with two military dwarves, with offensive skills on one (3 weapon/2 fighter/1 striker), and defensive skills on the other (2 shield/2 armor/2 dodge). Both train each other regularly and also tend to get new soldiers' teaching skills up to an acceptable quickly. If I do lower teaching and and increase the other skills, training goes much slower.

Observing seems to increase when dwarves observe sparring matches, but student goes up during demonstrations. Concentration is high on the training dwarves too, but I'm not sure where that comes from.
« Last Edit: May 30, 2011, 02:24:42 pm by DrKillPatient »
Logged
"Frankly, if you're hanging out with people who tell you to use v.begin() instead of &v[0], you need to rethink your social circle."
    Scott Meyers, Effective STL

I've written bash scripts to make using DF easier under Linux!
Pages: [1] 2