So, I made use of my humble programming skills and created a script which takes every possible build and tests it in battle. There are many interesting informations which can be read from the data. But first, to be able to understand it, you must know how was it measured. So here:

I did NOT take account of weapon types or chassis. Chassis type is not relevant assuming both squads have the same one, the same with weapon types. Not only that, but testing every possible weapon, chassis and upgrade combination would take years to get done... So don't want me to do that. So only infantry with concussive weapons was used in the testing

Now, because I have to put these builds into battle with something else to measure it's efficiency in real battle... I was thinking about various builds, but because only one is the most popular one (10A;0D;20R - or some minor modification of that), I used only this one. However, what varies a lot in real encounters is army size. So I tested them against three builds in total: Whether it wins in a fair battle, how much does the build damage a squad double of it's size and how efficiently it crushes a squad of half of it's size.

I had to measure the damage done / taken somehow, so I chose the cost as the meter. Because it's hard to smash together two costs (metal and oil), I simplified that:
Armor unit has the cost of 1
Damage unit has the cost of 1.75
Range unit has the cost of 2.5

If anyone strongly disagrees by the values, i can rebuild the table based on the scores you suggest.

So, finally.. here is the table:
http://pastebin.com/86dMAHdC

You probably need some legend first:
First column is the actual build.
15A;5D;10R is a squad of 15 armored, 5 damage and 10 ranged units
Second column is the total cost based on the modifiers above.
Third to fifth column is the battle score earned against these builds:
5A;0D;10R
10A;0D;20R
20A;0D;40R
The score equals to 0 if tied, the cost of all the units left if won and the negative cost of all the units left of the opposing squad if lost. So, generally, the higher the value, the better the build was performing against that squad. If negative, it lost, if zero, it was a tie and if positive, victory!

The results!

... okey, it's not so surprising. No build was able to beat the mighty "pro" build on one on one battle. But there is still room for improvement! whole 53 other builds tie with it! Here they are, sorted by total cost: http://pastebin.com/SMy5khUf

So, that's a more interesting table! You can read so many things from that!
For example, the cheapest, while still more or less on par with the range build is damage build:
10A;20D;0R with it's total cost of only 45.0. But as you can see... It dies quickly against bigger army and suffers a lot of casualties when standing against a smaller army.

Another interesting build is 5A;0D;25R. Although one of the more expensive ones, it crushes weaker opponents with minimal loss. A quick glance to in-game simulator shows that only the easy-to-replace armor units died when standing against 5A;0D;10R. Interestingly, it performs pretty well against stronger opponents, too! The only downside is it's prize, 67.5, but i think it's worth it if you can afford it.

Generally, anything in range of 5 to 10 armored units, the more armored units you exchange for ranged ones, the better performing, but more expensive to maintain build you get. Where 5A;0D;25R is the "cookie-cutter rich man's" build and the classic 10A;0D;20R more like "jack-of-all-trades" without any serious downside, but definitely not the best one in terms of performance.

Interestingly, there is no build with more than 10 armored units in there.

Finally, here is the same table sorted by the total score (all the scores summed up) regardless of the cost: http://pastebin.com/T4VfqfbK


So there it is... It's up to you to what extent will you trust the computer... or stick with your own experience... Take it more or less like a disclaimer

If you want to know the best build based on some criteria, tell me the formula and I'll post it.

If you want to test the builds against other squads, post them here and I'll make a table for you.

If you want anything else... good luck getting it.

Nice simulation you did there

I know I'll stick to my 21/9, but it's fun to see the full calculations for this stuff.

I'll stick to my all armor build :3

so wat :3

I wanna see my 500 armor units died, wat's the problem?

What?! What about their families?! Those guys have wives and children! I want to see you explaining that to them!

I sent them to the battlefield too.

Problem?

Nicely done.

By how much does 25:5 have to outnumber its opponent (assume probuild) to be more effective than probuild outnumbering by the same?

Also, where did you get the battle formula? Care to share?

Good question... After about 150% of the size:

Multiplied the builds by 10 to get more accurate results.

About the battle formula.. there is no science about that... Got pretty much all i need from http://wiki.battledawn.com/index.php?title=Battle_System. Based on that, i made an algorithm and checked it against the in-game battle simulator... and the results seem to match! So, the way i did that:

For each round, until one of the army is dead:

• First compute damage done by both armies accounting for distance.
• Count the total hp per unit type for both armies.
• Iteratively subtract the damage done by opposing army from the total unit type hp starting from armored units, because they take the damage first.
• Finally, compute the number of units left of each type from the subtracted hp. The formula is: ceil(total unit hp / hp per unit)

I can post the actual code if you want (and you know Python).

meh python is close enough to C++. I'd appreciate the script, so yes please.

so what this tells us is that if we are confident of being able to outnumber enemy by 160% 25:5 would be better? Can we adjust for cost (i.e., same cost army) which would be a bit more realistic?

10+20*2.5=60
(1+2.5*5)=13.5
60/13.5=4.4repeating
So we have 22:4.4

Dolkar can you modify to use the new build 22:4.4 instead of 25:5 to get the percentages like you just did?