Economic value of Warrior types

[Esthlos]

EDIT: It seems I forgot to actually state the goal of this thread: it is to find the accurate "economic value" of each warrior type, so that it can be used in the Test Fight map to compute and compare economically equivalent armies (as if coming from the towns built by two players of same skill but going for different strategies) yet made of different warrior types and troops.

Hi!

I've been trying to figure out a better way to compare the economic worth of warriors; up to now, I've been using the cost they'd have if they'd be entirely bought in the Market through Corn.
This sadly wasn't accurate at all.

I think I've figured out a (hopefully) better way: using the data collected about each building's production, I've calculated roughly how many cells are needed to fit an entire production chain to feed an Armor Workshop with Leather. [Number of buildings times (building's height +1) times (building's length+1)]

Then, I've simply seen how many production chains for every other military producer would fit in the same space, and compared the theoric production of these chains.

For raw materials' collectors' home's dimensions, I've used:

• Farm: 35 cells
  Mines: 64,09 cells
  Woodcutter's: 37 cells

These values are not random:
Farms are 15 cells for Fields (the optimal number) + 20 for the building itself [(building's height +1) times (building's length+1)]

Woodcutters' is tricky: theorically, each can cover an area of 254 cells, but they don't actually use it all. So I made one in the editor, let it run free for a while, then counted the number of trees (so 25 + 12 for the building itself) as soon as he started cutting them (because he then temporarily stopped planting).

For Mines I've calculated roughly the number of free cells in location 6 of "Across the desert" and the average number of cells where Gold and Iron mines could be built on average per loc.
Then I've estimated (with the mentioned data) how many production chains they'd be able to sustain, and seen how many theorical cells they're occuping for these production chains to fit the given loc space.

After all this was done, here's the end result:

values

And, attached, the test fight map that can optionally use these values to estimate an Army 2 economically equivalent to the given Army 1.

map

Opinions, please?

To me, these values don't "feel" right, but I can't find the error in the process... maybe mines should be "worth" less space, but how much?

[Esthlos]

Judging by the overwhelming response to this issue I'm unsure if there's a point in updating this info...

I've found a few inconsistences in the algoritm and used a modded KaM_Remake.exe to have the game print more detailed info without having to scavenge manually ( ) for it.

Also figured a better way to estimate the worth of ores, based on the notion that for a strategy to be a viable option instead of plain better, then they must be equivalent as far as the ratio "value/used space" goes.

Admittedly ores should actually be plain better because they are limited and their units are better, but I've yet to figure out a good way to do this adjustment... on the other hand, though, another thing to consider is that the space used for Iron is effectively wasted once you run out of ore, which makes Leather and Wood productions more valuable in the long run.

This update is more of a fix, using more precise data and more comprehensive calculations.

Anyway, here are the adjusted values:
Civilians: 7
Militia: 13
Axe Fighters: 22
Sword Fighters: 27
Bowmen: 17
Crossbowmen: 20
Lance Carriers: 17
Pikemen: 20
Scouts: 34
Knights: 39

P.S. Funny thing: Across the Desert has 23456 buildable tiles. Really.

EDIT: Found a typo in the formula I was using... sorry. The correct values are:

Civilians: 7
Militia: 13
Axe Fighters: 22
Sword Fighters: 31
Bowmen: 17
Crossbowmen: 23
Lance Carriers: 17
Pikemen: 23
Scouts: 34
Knights: 43

For comparation, here's the "Corn equivalent" for buying these units through the Market:
Civilians: 4 Corn
Militia: 11 Corn
Axe Fighters: 23 Corn
Sword Fighters: 37 Corn
Bowmen: 19 Corn
Crossbowmen: 26 Corn
Lance Carriers: 19 Corn
Pikemen: 26 Corn
Scouts: 36 Corn
Knights: 50 Corn

[Soldado con espada]

I think you should normalise all the values to the highest or lowest value to compare the two methods. It would help.
I will make a couple of plots with these values when I go back to Linux (if you don't make them before I do xD). I'll edit the post.


Edit: as I promised, here are the plots for an easier comparison. Remember to read the three Legends

Value_Corn.png

Value_Parcel.png

Those two are nice to visualise the values. Next is the comparison. The lower plot is the ratio of the values, Corn/Parcel:

Value_Comparison.png

You can see that the "expensive" warrior types (not militias ) are predicted to be more or less of the same cost in both methods WITH THIS NORMALISATION. Civilians and militias are very diferent, however.

The normalisation I used is: I divided each "corn value" by the sum of all the corn values for the different units. And the same for each "parcel value". The reasoning was "the sum of the cost of all the units should be the same in both methods". Of course you may not like this "reasoning" and I would accept it. It is probably false.

In any case I think the main idea is that your two methods predict wildly different costs for civilians but similar costs for warrior units.

Making these plots is what I've done for a living for a few years, but I couldn't be bothered to make them perfect... You can criticise me, no problem

If you are interested (I don't recommend you being interested in the way I do things, there are much better an easier ways) I attach here the python code I used. You need ROOT installed to run it

WarriorEconomicValues.zip

[Esthlos]

Someone else cares

Here are the updated test fight map and the spreadsheet with the used math, in case you find them useful:

Map

Spreadsheet

[Soldado con espada]

Updated post with plots!

Question (it will be easier than looking for it xD): do your values for warriors contain the value of a civilian?
It's a constant for all of them, so it doesn't really matter in the comparison, but it would be good to know it.

[Esthlos]

Updated post with plots!

Very interesting, thank you!

Question (it will be easier than looking for it xD): do your values for warriors contain the value of a civilian?
It's a constant for all of them, so it doesn't really matter in the comparison, but it would be good to know it.

Yes.

Also, there is another thing to be said: the new "warrior values" also include actually using the Gold to train a Recruit (thus additional space needed for the School, thus an higher value), while the Corn values do not: 4 Corn is just the cost of buying 1 Gold chest in the Market, since you can't buy Recruits.

[Esthlos]

I've tried to adjust the "occupied tiles" value of each building in every chain by applying them the ratio [(average distance from viable tiles)/(number of viable tiles)], obtained using values taken from Across the desert through a script, instead of the... old one I didn't write about here ( ), and that was actually pointless.

Normalizing them (assuming I did it correctly ), they should be pretty much equivalent to the Corn values.

Here are the values:
Civilians: 3
Militia: 9
Axe Fighters: 17
Sword Fighters: 31
Bowmen: 13
Crossbowmen: 22
Lance Carriers: 13
Pikemen: 22
Scouts: 27
Knights: 42

P.S. It seems I forgot to actually state the goal of this thread: it is to find the accurate "economic value" of each warrior type, so that it can be used in the Test Fight map to compute and compare economically equivalent armies (as if coming from the towns built by two players of same skill but going for different strategies) yet made of different warrior types and troops.