Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[Colonel Sanders Lite]

The purpose of these experiments is to test the efficacy of using a veteran soldier to train new recruits vs just throwing new recruits in a squad by themselves to spar.

The game version used for these experiments is 43.05.

Experiment 1:
5 Recruits with a teacher VS 5 Recruits without a teacher.

The setup:
10 recruits, split into 2 squads of 5.  Squad A, The Orbs Of Oiling, is lead by a veteran soldier, acting as a drill instructor.  Squad B, The Helmed Bowels, has no instructor.  Each squad has it's own training room and sleeps in a shared barracks.

TL;DR Results:
Skills after 2 years of training:

Spoiler (click to show/hide)

As can be seen, both squads are approximately on par in the axe, discipline, and fighter skills.
Squad A has a strong lead in biting, dodging, kicking and striking skills.
Squad A has a slight lead in the wrestling, armor, and studying skills.
Squad B has a slight lead in the shield user skill.
Squad B has a strong lead in the Teaching skill.

While these results clearly show the efficacy of using a veteran to teach your new recruits over a training period of two years, the results can be a bit muddier when considering a shorter time frame.  Squad B started Sparring sooner and did so much more actively for much of the test.  Squad B gained weapon, shield, strength, and agility much more rapidly at first, while squad A had gains in armor user, dodge, striking, and kicking.  Once squad A starting doing a lot of sparring, they caught up to and largely surpassed squad B by the end of the test.  See the full results below for details.

Very important note - Cog Atollorbam (squad A) was by far the weakest recruit.  At first, she had trouble getting training time in due to the weight of her gear slowing her down when it was time to get provisions, sleep, etc.  Towards the end of the third season, and in the beginning of the 4th season of the test, she managed to gain some armor skill from demonstrations and then improved her strength through sparring.  This allowed her to start keeping up with the rest of the group.  I strongly suspect that if she did not have a veteran teacher showing her how to wear her armor, she would still be lagging very far behind the rest of the recruits.





Additional Details:
Recruit Selection:
All recruits have a score of 15 in the "values martial prowess" trait to remove that desire as a variable.  From there, recruits where primarily selected by strength, to help reduce the variables relating to movement speed.  Squad B got the strongest recruit, Squad A got the second strongest, Squad B got the third strongest, and so on.  Looking at the strength, agility, and endurance stats, squad A has the a slightly worse batch of recruits than squad B.  This was done so that we know that, if Squad A pulls ahead of Squad B, it was not likely due to initially  better stats.

It's worth noting that most of the recruits have ranks in marksdwarf and kills.  This comes from untrained citizens being drafted, grabbing a crossbow and getting on the ramparts when goblins attack.  Besides just taking pot shots at goblins, they have no prior military training.  A few recruits have a few xp from dodging/blocking the occasional stray arrow during goblin attacks.  These recruits are evenly distributed across both squads.


Equipment:
All recruits are equipped with:
Battle Axe
Shield
Helm
Mail Shirt
Gauntlets
Greaves
High Boots

Dwarves are set to replace clothing and wear exact matches only.  All equipment is masterwork quality steel, to eliminate any variables caused by uneven equipment.

All Data:

Spoiler (click to show/hide)

The initial stats:


Seasonal Data:














Final Results:

[Colonel Sanders Lite]

Experiment 2:
This experiment will involve varying class sizes, to see if there is some ideal ratio of veteran teachers to raw recruits.  Details to come soon.

[Spriggans]

Loved it.

Could the fact that one squad was 5 large and the other 6 large be impactful ?

In experiment 1 :
Squad A was 5 + 1 teacher = 6 dwarves large.
Squad B was 5 large.

So maybe that caused less sparring in squad A, which lead to its slower learning in axeskill ?

EDIT : The difference in biter is very impressive !
Do you if a veteran in misc. combat skill will teach his students that ?

[Colonel Sanders Lite]

Could the fact that one squad was 5 large and the other 6 large be impactful ?

It's definitely possible that this had an impact.  My intuition, based on my observations, is that it might have been the other way around though.  I strongly suspect that squad A initially did less sparring because the Drill Instructor spent more time demonstrating all those other skills.  However, the way that Squad A had almost caught up to squad B by the end of the test might be because the larger squad size increased the amount of sparring towards the end.  Then again, the increased rate of sparring in squad A towards the end of the test might have been due to the higher armor skill increasing speed and reducing fatigue.

The next experiment might help shed some light on that.

EDIT : The difference in biter is very impressive !
Do you if a veteran in misc. combat skill will teach his students that ?

Excellent question.  I added Misc Object to the grid view in DT and here's the final result on that:


I don't have the hard data on this season by season, but we can see that it likely worked exactly the same as the biter skill did.

[Ulfarr]

This is very interesting . Compared to my results both of your squads have achieved in two years skill levels that my squads needed 3-4 years to achieve and in some* skills (biter,kicker) they would still  1-2 levels behind yours.

*This is, most probably, because of the teacher.

Could the rest of the differences between our sample be because of you using smaller squads (5 vs 7 members)? Or maybe because of choosing dwarves that "value martial prowess" (I didn't check any of their personalities for my experiment)? What was your training schedule?




 

[Colonel Sanders Lite]

Honestly, without having even seen your fort, there's just way too many variables for me to say.  That's the exact reason I used my own control group, instead of trying to compare to one of your squads.

I could honestly think of dozens of potential unaccounted for variables.

That being said, if you like, I can post a save for you to look through.


Edit:
I was thinking...  If I had to take a wild guess, the culprit would be that the way you have 6 set to train, with 1 off duty.  I just used train with 10 minimum for both squads.

[Ulfarr]

Honestly, without having even seen your fort, there's just way too many variables for me to say.  That's the exact reason I used my own control group, instead of trying to compare to one of your squads.

I could honestly think of dozens of potential unaccounted for variables.

That being said, if you like, I can post a save for you to look through.


Edit:
I was thinking...  If I had to take a wild guess, the culprit would be that the way you have 6 set to train, with 1 off duty.  I just used train with 10 minimum for both squads.

For starters, I think we can rule out the personality traits. I have 4 dwarves that don't value martial prowess (their values are -10,-7,4,6), yet their skills are in line with the rest of their squad.

You are probably correct about scheduling. Assuming the dwarves rotate their breaks evenly, my schedule will result in 2 off duty months every 14 months of total training. That means that by the end of the 3rd year, each individual dwarf would have about 30 months of actual training and 6 months of off duty  work ( fortress work, socialising,preying or individual drills). I 'm assuming that with your schedule, your dwarves even when not sparring (from my experience they tend to spar in pairs, though 3 at a time isn't unheard of)  they would still focus on individual drills which can explain the difference between our results.

The reason I used that schedule was because of this quote from the wiki, which, since you report no such problem, I assume is outdated:

Spoiler (click to show/hide)

 

http://dwarffortresswiki.org/index.php/DF2014:Scheduling

Warning: Dwarves who are permanently on-duty with no downtime have been observed to begin to starve themselves keeping to the rigorous schedule, and thus grow unhappy. Do you really want to find out how much damage that practice spear can do? To allow some of your dwarves to go sleeping when they need, and also eat and drink (even from carried rations!), you need to lower the minimum number of dwarves of the squad that need to follow the current order at any time in the order criteria as listed above. You want the criteria to be at least one or two dwarves less than the current number of dwarves in your squad. This will change the limit for the current month only, so use the copy/paste feature to update every month. It should be emphasized that dwarves who are stationed or on patrol will not eat or drink from their backpacks and flasks. Backpacks and flasks are more to shorten break times than to eat in the field. If there are no opportunities for breaks, carried rations will not be utilized.

[Insert_Gnome_Here]

I could honestly think of dozens of potential unaccounted for variables.

 
 
It'd be interesting to gather a load of longitudinal training data from a load of forts and do a statistical analysis. 
‼Big Data‼

[Colonel Sanders Lite]

You are probably correct about scheduling. Assuming the dwarves rotate their breaks evenly, my schedule will result in 2 off duty months every 14 months of total training. That means that by the end of the 3rd year, each individual dwarf would have about 30 months of actual training and 6 months of off duty  work ( fortress work, socialising,preying or individual drills).

There's another factor that you might be overlooking too.  Travel time.  Suppose Urist McRawRecruit is off duty, hauling rocks down in the depths of the fort.  The scheduler tells him to get back on duty.  As dwarves tend to do, he insists on carrying that rock until he reaches his destination.  Strictly speaking, he is on duty, but just hasn't arrived yet.  You might have a squadmate sitting in the barracks, organizing a striking demonstration, just waiting for Urist McRawRecruit to get back...  In the meantime, Urist McOtherRecruit decided to go ahead and go on leave.

I would imagine that this is especially a problem when the recruit is still pretty green.  Low on armor skill and strength with all that gear slowing him down.

I 'm assuming that with your schedule, your dwarves even when not sparring (from my experience they tend to spar in pairs, though 3 at a time isn't unheard of)  they would still focus on individual drills which can explain the difference between our results.

Yep.  It wasn't at all uncommon to see two dwarves in a squad sparring, with the rest involved in a skill demonstration or doing individual combat drills.

The reason I used that schedule was because of this quote from the wiki, which, since you report no such problem, I assume is outdated:

Spoiler (click to show/hide)

 

http://dwarffortresswiki.org/index.php/DF2014:Scheduling

Warning: Dwarves who are permanently on-duty with no downtime have been observed to begin to starve themselves keeping to the rigorous schedule, and thus grow unhappy. Do you really want to find out how much damage that practice spear can do? To allow some of your dwarves to go sleeping when they need, and also eat and drink (even from carried rations!), you need to lower the minimum number of dwarves of the squad that need to follow the current order at any time in the order criteria as listed above. You want the criteria to be at least one or two dwarves less than the current number of dwarves in your squad. This will change the limit for the current month only, so use the copy/paste feature to update every month. It should be emphasized that dwarves who are stationed or on patrol will not eat or drink from their backpacks and flasks. Backpacks and flasks are more to shorten break times than to eat in the field. If there are no opportunities for breaks, carried rations will not be utilized.

I can assure you that this information is wildly inaccurate.  I will also admit that it has been a *while*, so my memory cannot be relied on fully, but I don't think that any of that was accurate in the 40d days either.

First, dwarves on duty will eat and drink their carried provisions.  I know that this is true whether set to defend a burrow or set to train.  I'm not as certain about station orders, but I'm still pretty sure that it's true there as well.  I do not use patrol orders, so I can neither confirm or deny that aspect.  I have no real reason to believe it's any different.

Second, in the scheduling, you can set dwarves to sleep in their rooms at will, in the barracks at will or in the barracks at need.  These orders apply to dwarves that are on duty and do work correctly, with the same certainty in various situations as the provisions stuff above.

Third, dwarves with backpacks, waterskins, and quivers will leave their posts to collect provisions once they are empty.  I see this happen routinely.  An important note is that dwarves that do not carry backpacks or waterskins will not get the collect provisions job and will remain at their post if hungry/thirst.  This is at least true in the case of defend burrow orders.  In the case of a training order, I'm not as sure, but I strongly suspect that it works in the same manner as the defend burrow order.

Fourth, I'm not 100% sure it still works, but at least in 43.03, dwarves with the feed patients/prisoners labor enabled would feed hungry on duty dwarves, if they can be reached.  Of course, as always, this is a relatively low priority job.  A soldier could starve to death in the specific case of being tasked to stay on duty without a backpack when everyone is too busy hauling rocks.


It's certainly possible that these where legitimately encountered bugs at one point.  I haven't played *every* version of DF ever.  Besides which, my memory for little details that go too far back isn't the best.  If I had to guess though, all the problems observed by that player where caused by his dwarves not equipping backpacks and waterskins for one of a multitude of reasons, combined with a lack of idlers as detailed just above.


I can say that there are specific downsides to keeping a dwarf constantly on duty though.  A dwarf that is on duty: Won't go to the temple.  Won't eat in your legendary dining room. Will receive negative thoughts from being on duty for too long.  Will receive negative thoughts for being made to sleep in the barracks instead of his room.  Won't be able to see family and friends.  Etc.  Essentially, there are negative thoughts and unmet needs that will definitely build up and the positives are less numerous for an on duty soldier.

That being said, I have never had a problem with an initially happy dwarf during a soldiers basic training period.  After their training, I tend to give them a season off to go and meet their needs, and then put them into the main duty rotation.

[Sanctume]

An idea to test using first 6 as training.

Dwarf 1. Wrestler+5, Teacher+1 to +5.
Dwarf 2 - 6. Recruits.

Dwarf 7. Fortress Builder.

Calm biome. Invasion off. Pop cap: 19. Embark with 200 Drinks, 100 Food, and stuff to make initial uniform.

Test 1. Uniform: Helm + Mail Shirt. 
1. How soon can the recruits reach Wrestling+5?
2. How much Discipline is gained?
3. How many other skills were learned from Individual Drills: Fighter, Dodger, Biter, Kicker?
4. How soon does sparring start?

More test: Repeat test set increasing initial Teacher skill.

5. Are there any Armor User skill gained?  Will using more armor help gained Armor User, i.e. +Greaves +High Boots + Gauntets?  How about Leather/Bone vs Metal?

[Colonel Sanders Lite]

Working on experiment 2 detailed above.  It's early, but I already have an interesting preliminary result:



It's far too early to tell anything else yet, but observe squad A with one DI and just one recruit (kulet).

Kulet started as the weakest of the recruits, with an overall ranking of 37, but is now the strongest of the recruits.  She is also the first to hit legendary in her weapon skill by a very wide margin.

She did all this in 9 months and 15 days.

As far as I can tell after careful observation, this boils down to one simple reason:  Due to the way the squad members current status and job selection interact, 2 person squads rarely do demonstrations.  They do some individual combat drills, and *lots* of sparring.  If, as I believe, this hypothesis is correct, it has some substantial ramifications on embark setups that use two soldiers.



As for the effects on training with other squad sizes, I haven't finished collecting all the data yet.

[Colonel Sanders Lite]

@Sanctume
It might be worth checking out, but I anticipate problems with that test.  In particular, the recruit pool isn't very large, meaning that there isn't really any opportunity to reduce the variables caused by the characteristics of the dwarves themselves.

Beyond the dwarves themselves, doing a new embark for each trial also introduces a lot of chaos as well.  Even if each embark is in the same location.  As a few examples:  A test could be spoiled by a were-horse attack in the middle of the 10th month.  Variables can be introduced in the form of uneven equipment quality owing to the limited time frame.  How would you account for the rhesus macaque that stole a battle axe?

I think that you can get much more controlled circumstances in the depths of a well run stable fortress.

[Sanctume]

dfhack, liquids, and set obsidian walls and floors as immediate cover, then user up/down hatch constructed by the 7th dwarf. 

food and drinks in wagon do no spoil, so it can serve as QSP. 

dfhack create-items 6 sets of steel armor of masterwork quality can be done. 

I am not sure if dfhack gui/gm-editor can set individual dwarf preference to values martial prowess. 
Maybe if DF Tools is working, all starting dwarf stats can be modified, so it's possible to have 11 dwarf start, 10 with identical stats, and 11th as builder.

I'll read up on Dftools if it can be updated to 44.05

[Colonel Sanders Lite]

If you're going to hack, why even bother doing the multi embark thing?

Build a small sealable fort with lots of food and drink and all the equipment you need.

Pick out your guinea pigs, and get them set up into their squads.

Set their stats to the baseline.

Save and backup the save.

Run Test. Record results.  Restore backup.

Edit parameters.  Run Test.  Record results.  Restore Backup.


Etc.


Since DF is multithreaded, you could even run multiple tests on a multicore cpu at the same time with no perfromance loss.

[Sutremaine]

Dwarves set to permanently train and not carry backpacks will eat as appropriate. Some of them might become distracted over time, if they have a strong need to pray or do other location-based activities, but sparring does fulfil several different needs by itself.