#Production cost and return

1 messages · Page 1 of 1 (latest)

mild pagoda
#

Updated version here: #1130816687311179846 message

This is an unbiased spreadsheet for information on unit production cost, stat score and stat throughput, and part eco planner.

  • Use the blue boxes at the top for input.
  • Stats used for score calculation may be adjusted with the checkboxes.
    Selecting none of the stats gives all units a score of 1, and is useful to compare costs and throughput without stats interfering.
    Checking single stats is also useful, as well as HP & DPS, or Speed & Range. I recommend checking multiple aspects and using your own judgement about the results.
    This sheet will not give you a final answer.

Build power available: Set the amount of build power you wish to have. Increasing build power decreases build time, but increases costs.

Walk-off time: Time a unit takes to walk off the production ramp so next unit production may start.
Walk-off time directly affects unit throughput (units/time). I suggest 0,5 - 1,5 sec, or use 0.

Target metal/energy income & margin: The income you plan to have available for unit production* + margin (%). Margin is added to metal/sec & energy/sec columns scale and checkmarks. Margin is added because of walk-off time giving room for producing costlier units.
Default values are production supported on 1 AFUS.
*income needed for other builds not included.

Score: Geometric mean of selected stats (HP, DPS, Range, Speed).
Score/cost: Amount of stats returned for metal & energy spent.
Score*throughput: Amount of stats returned for build power spent. (How fast do you get stats from production of this unit).
Score*throughput/cost: Both the above combined.

More detailed data on units in other tabs if you wish to dive into the numbers.
The main table has some notes for info, but just ask if anything is unclear!
Suggestions for additions and improvements welcome!

image.png image.png
#

I can post it to Google Docs as well, just let me know.

fading thistle
#

took me a while to sift through, but it's a really good representation, love the color coding and all. will need a closer look to gain some insights, but suprised how high the incisors are on the list & i thought they would be ranked in several things a bit lower.

mild pagoda
#

Yes, the tier 1 "scouts" (Rascal, Grunt, Incisor) pretty much always wins out on throughput and cost.
But high speed, low range makes it hard to spend their stats effectively.

#

Might be useful to score by just HP and DPS to compare.

#

Thanks for the feedback!

#

I was surprised how high the Fiend scored overall and on HP+DPS score. Seems it might be cheap for its stats and build time.

mild pagoda
#

As you scale up build power the scouts will still win out if you have walk-off time set to 0.
With walk-off the end-game units will win out with sufficient build power, so you have a reversal as eco scales up.

#

Compare with "infinite" build power:

image.png
#

Also, this does not account for AOE, traversability, stealth and other things which impact unit effectiveness.

thorn prawn
#

This thread's sheet might be useful for additional ideas related to sorting options #1062929118775230614 message

mild pagoda
#

Thanks, I checked it briefly before posting.
At first I tried giving some bias to the scores which is what that spreadsheet seems to do.
But I could not make it work properly without skewing the results, so I ended up simply comparing the stat spreads against each other for scoring. Will check it out more when I have time.

thorn prawn
#

Bias probably won't work particularly well no matter how you attempt it, but it has things like "value excluding movement speed" for defensive circumstances

mild pagoda
#

I have a selector for which stats to include in score, but you won't be able to view data from both at the same time yes.

thorn prawn
#

I should've looked more closely at your work before saying anything but am short on time at the moment and hoped it might provide something of value for ideas of what could be added or improved

mild pagoda
#

Suggestions are very welcome :)

fading thistle
mild pagoda
#

I have some time next week so I can look at adding the rest of the units

mild pagoda
#

It has confirmed some of my hypothesis, like how build power worked, or the reversal of "best" unit as a function of build power and walk-off time.

#

I was surprised how little energy unit prod. costs, while metal is quite high. Seems to confirm that converters are needed.
At least if we go by the conventional wisdom measures 25/500 and 50/1500 M/E I think I heard somewhere..

#

Typical usecase is ex. I want to play Core Vehicles, and know I will get 6 mex before T2. So I put in 6 mex, and adjust metal income to see which T1 vehicles I can produce in what amounts, and I can see how many converters & energy I need to keep up production in the top.

#

Now, I see for score/cost the best is actually the Poison Arrow, but it gives me 1,3 units/min which will not do. (And we may not have T2 yet anyway).
Score/cost for T1 is Rascal, Lasher, Brute. Which is interesting, so I've tried playing a bit with the Lasher.

#

Throughput score the best is the Incisor, and throughput/cost is the Rascal.
I never tried spamming Rascals, but low HP looks hard to put to good use, so I will start with the Incisor while I get my eco up then look to go Lasher or Brute.

#

Then adjust build power and eco up for T2 etc.

#

Or, you want to spam a particular unit, look up its costs and adjust accordingly. See how much build power you can cram into it.

#

So you can use it to set some targets and plan accordingly.
Or simply compare units. Compare Vehicles to Bots, T1 to T2 etc.

rustic stirrup
mild pagoda
#

I use the geometric mean of the 4 "main" stats which I found was the best. Mean/median/geomean is in a different tab with comparisons, and you can switch to another measure.

rustic stirrup
#

How do you define "the best"?

mild pagoda
#

With the mean and median, if I selected just e.g. HP the ranking was different for score than if I ranked just HP stat, which is wrong I think.

#

Geo mean gave the same order for single stat and HP&DPS.

#

I am not a mathematician tho so I may be wrong :D

#

My rationale was simply to bring the stats into the same range or size to be able to compare them, but without squishing the top/bottom too much. Geo mean seemed to do that.

mild pagoda
rustic stirrup
#

I have severe doubts about this kind of scoring, honestly... It just feels like it doesn't convey any useful information. There is no universal single best unit for all situations, so compressing unit stats to a single scalar value like that just sounds... off 😦

thorn prawn
#

At the very least, severe outliers from conventional knowledge can be tested to see if we're undervaluing a unit based on false assumptions

#

It does sound like this method of assigning value is potentially valuing longish-range units too highly though

#

Non-quantifiable factors can be quite heavy as well, but maybe some other units end up jumping out to people regardless. I myself have been tempted to make incisors in slightly larger numbers as of late but have been hesitant due to damage dropoff and inability to fire over each other. The sheet can't quantify those when ranking them highly, but i still find the idea more tempting than before

#

In that particular case it mostly comes down to being tired of pounders falling way behind whatever they're supposed to be helping with handling cheap balls of units, though, let alone how far they fall behind the cheap units themselves

#

Also an only-somewhat-quantifiable factor

#

Actually this probably quantifies that part well, considering it values speed relative to other units' speed

mild pagoda
#

Yes, even if I could include all quantifiable factors .. If we want a perfect simulation we have to play the game.
Unit efficiency is situation dependent. A score will never show that.
But it may inform player decisions.

mild pagoda
#

v1.1
Armada units added, and some adjustments to Cortex stats. All dps and range numbers are for ground weapons only - no anti-air or sub-sea.
Hound is split, one unit for each firing mode. Thor is only lightning cannon dps, no missile included. Thor missile buildtime seems to be 66,67 sec, so I suggest adding 1200 dps to its stat if you want to include missile damage.

I will look to combine the spreadsheets into one next, not sure when I will get around to it.

📎 UNBAR_Armada_-_UNit_Balance_And_pRoduction_v1.1.xlsx 📎 UNBAR_Cortex_-_UNit_Balance_And_pRoduction_v1.1.xlsx
mild pagoda
charred seal
#

Cool!

mild pagoda
#

v1.2
Combined Armada & Cortex units in the same sheet.
Updated some formatting.
Added ability to select none of the stats for score calc. All scores will then be 1, and ranking columns for cost and throughput will only reflect those numbers.

Updated description in main post.

📎 UNBAR_-_UNit_Balance_And_pRoduction_v1.2.xlsx
#

I am not satisfied with the walk-off / cost numbers.. Cost is correct, but the delay between production gives a good margin.
The margin addition will have to do for now.

#

I believe walk-off should be a function of unit speed and build time. More speed = lower walk-off, lower build time = more stops in production.

rustic stirrup
#

also acceleration

mild pagoda
#

Ah, true. Probably what I'm missing in my current hunt for a formula.

mild pagoda
#

Shoot, maybe I should've specified amphibious units in the type column. Might be relevant.

mild pagoda
mild pagoda
#

Took a stab at calculating walk-off time by speed, thanks to Perfi reminding me acceleration exists. Unfortunately it was a miss.. Delay between productions doesn't seem to be tied to speed and accel.
After testing in-game, I found varying distances needed to walk resulting in varying times. E.g. Fiend walk-off was 0,94s, while Sumo which is slower and bigger was 0,3s, surprisingly. This equates to a distance walked of 70 by the Fiend and 5,4 by the Sumo .. Size doesn't seem to matter either, so I'm at a loss.
It seems easier to estimate the time by checking # units produced over 10 mins and calculating the difference from build time, like I started with for controlling the numbers.

test2.PNG test1.PNG
molten hatch
#

@mild pagoda could it be using colVol's to check if the build area is clear?

#

To enable/disable: /debug colvol

mild pagoda
fading thistle
rustic stirrup
#

Did you include the unit's radius in the calculation?

mild pagoda
#

Mammoth and Termite is larger than the Fiend, but further back.

#

And the tables I set up if any interest

image.png
#

straight line in two parts, first for accel to max vel, then to walk remaining distance.
I used the times I measured in-game to solve for d and got different distances

mild pagoda
molten hatch
#

archon, your effort into this is epic, much respect :)

mild pagoda
mild pagoda
# fading thistle

Thanks a bunch for the code! I adapted it to use game frames and spit out a log file.
I wasn't aware build times vary, but they seem to vary by +- 1 frame for both build time and delay. May have to check this further - it may just be rounding to nearest frame.

Worse yet, the times I measured are consistently higher than my calculated build times.. so there must be something more I'm missing.

📎 production_timer.lua 📎 production_timer_log.txt image.png
mild pagoda
#

Might be a bug in the code too.. got pretty late tinkering on the widget :)

fading thistle
#

did you manage to convert the code to track just the difference it takes to go off the ramp @mild pagoda , now that I think off it, I used it more to calculate how much units i can make with different setups.

good stuff on the frames, iirc game runs at 30 simframes per sec.

#

ramp should be difference between UnitFromFactory (start timer), until next UnitCreated (meaning lab can work)

#

or other way, idc who callin did what

mild pagoda
#

Yes, I used the Created event to mark start time, and Factory event to mark end time, then calc the diff each time

fading thistle
#

ahh, perfect

mild pagoda
#

wondering if I have an off by 1 error somewhere, tinkering some more with the widget

fading thistle
#

i'll load up your widget see, if maybe can draw on the screen each time it gets, instead of spring.echo log, i bet gpt can cook up text on screen as personally when I tested back then in 2021 it was pretty hard to read from the logs, but your export file looks pretty cool

mild pagoda
#

you'd need to queue n+1 units to get the last delay

#

like so

image.png
#

I believe this should be better.

mild pagoda
#

This time I think it's right.. Nope

mild pagoda
#

Con turrets tripped me up. There is a delay from when production starts until they start contributing. For the Grunt which I was testing with, that meant as much as 7% was built by factory alone before cons activated. Will have to go by factory build power alone.

thorn prawn
#

/nocost?

mild pagoda
# thorn prawn /nocost?

Yes, that's what I'm testing now! Ended up being interesting I think -- both build time and delay vary still, but their sum is strangely neat. I'm wondering if it's not a coincidence.

thorn prawn
#

I feel like at some point reading through patch notes in the past i remember seeing a nerf to something's walk-off time, but i have no idea which value(s) in the code it would be tied to or whether i'm even remembering correctly

mild pagoda
mild pagoda
thorn prawn
#

as in, the game running in simframes of 30 fps like LSR said?

mild pagoda
#

no like, thug buildtime = 6.00, delay = 24.00, sum = 30.00.
or the mammoth bt = 1,05 + d = 58,94, sum = 59,99 ~= 60

#

since the build cost is 1 it should be 1 frame or close to nothing, so its not in the equation

#

or maybe I'm thinking about this too much lately

#

but it would mean set values like 1,0s or 1,5s, 2s

thorn prawn
#

you've glimpsed a piece of some eldritch truth, need to discover how everything fits together

mild pagoda
#

So yeah.. I'm about out of ideas for finding the true time of unit production. Might have to do without walk-off for now.

thorn prawn
#

Well, wouldn't the values you found which included time needed for con turrets to start working be the most applicable for real situations?

mild pagoda
#

And the delays I logged are too different I feel, when comparing /nocost on with off.

#

The game is doing something I'm missing.

thorn prawn
#

would it work to use a formula combining nocost walkoff time, the amount of building done by lab before con turrets start helping, and only adding the nanos' bp during time when neither of those 2 conditions are active?

mild pagoda
#

Not sure, the numbers I logged were without con turrets to eliminate that variable.
I still can't find any consistent numbers.

image.png
#

1 - 4 frames (or ticks) is 0,033s - 0,13s, so maybe that is an acceptable margin of error.
But I'm afraid that small number will be enough to skew the times when scaling up.

#

And the delays are way different from each other.

#

This is what I found interesting with /nocost earlier. Even if the buildtime and delays make no sense, they add up in neat divisibles of 30 ... mostly. :)

image.png
earnest verge
#

@mild pagoda i just saw this
so colvols do not play any factor here
its footprints and movedeffs that do

mild pagoda
#

Production cost and return

mild pagoda
#

Data in this sheet is now outdated and not maintained as I have satisfied my own curiosity for now. Ping me if you need anything.

alpine cargo
#

@mild pagoda

If you want to, then you can dig into the next rabbit hole.

Here is a sheet of mine.
May I present: the 10min afus.

https://docs.google.com/spreadsheets/d/1UHBmVmeEqJG5m9ntDE-P7c0N8Kw0w7KH4Sxt18g3t9s/edit?usp=sharing

Google Docs
10min Afus

Timesteps

Time step (sec),time step / 10 (for graphs),stored metal,stored energy
Energie,Max E,available BP,max BP / M,max BP / E,used BP,used M per sec,used E per sec,produced M per sec,produced E per sec,availabel Bp ,energy excess (unused by builders),exess E after advanced metal maker (amm),...

mild pagoda
alpine cargo
charred seal
#

Time to an afus on what map?