The Status of the Republic of Test (#41)

The Basics
Turns Left 106
Turns Taken 1
Turns Stored 100
Rank 37
Networth $5591
Land 120
Money $33,453
Population 934
At War No
GDI Member Yes

Land Distribution
Enterprise Zones 0
Residences 0
Industrial Complexes 0
Military Bases 0
Research Labs 0
Farms 0
Oil Rigs 0
Construction Sites 1
Unused Lands 119

Technology
Military 0 100.0%
Medical 0 100.0%
Business 0 100.0%
Residential 0 100.0%
Agricultural 0 100.0%
Warfare 0 0.200%
Military Strategy 0 100.0%
Weapons 0 100.0%
Industrial 0 100.0%
Spy 0 100.0%
SDI 0 1.00%
Total 0

Current Status
Tax Revenues $5753
Tax Rate 35%
Per Capita Income $17.60
Expenses $427
Net Income $5326
Cashing $6477
Food 112
Production 48
Consumption 28
Net Change 20
Oil 0
Production 0
Building Rate 5
Explore Rate 54
Tech Rate 3

Military Forces
Spies 0
Troops 0
Jets 0
Turrets 0
Tanks 0
Nuclear Missiles 0
Chemical Missiles 0
Cruise Missiles 0

Expense Breakdown
Expenses $427
Military $0
Spies $0
Troops $0
Jets $0
Turrets $0
Tanks $0
Alliance/GDI $0
Land $427
Corruption $0

Anyone care to point out where my math is wrong? How does the Networth equal $5591?

(Troops * 0.5) + (Jets * 0.6) + (Turrets * 0.6) + (Tanks * 2) + (Spies * 1) + (Tech * 2) + (Land * 45) + (Buildings * 35) + (Money / 20000) + (Food/1000) + (Missles * 2500) + (Pop/6) + (Oil/100)

(0) + (0) + (0) + (0) + (0) + (0) + (120 * 45) + (1 * 35) + (33453 / 20000) + (120/1000) + (0) + (934/6) + (0) = 5592

I guess they're rounding down...

In my professional experience, its relatively common to only use whole dollars in financial reporting for larger amounts. While cash transactions often have cents/decimals, they can clutter up reports. The thought process is that $5,591.99 is still not actually $5,592, so just report to the whole dollar you have.

So could be something similar here due to decimals

The Basics
Turns Left 123
Turns Taken 5
Turns Stored 119
Rank 2512
Networth $4989
Land 100
Money $37,585
Population 828
At War No
GDI Member No

Land Distribution
Enterprise Zones 6
Residences 0
Industrial Complexes 0
Military Bases 0
Research Labs 0
Farms 0
Oil Rigs 0
Construction Sites 4
Unused Lands 90

Technology
Military 0 100.0%
Medical 0 100.0%
Business 0 100.0%
Residential 0 100.0%
Agricultural 0 100.0%
Warfare 0 0.200%
Military Strategy 0 100.0%
Weapons 0 100.0%
Industrial 0 100.0%
Spy 0 100.0%
SDI 0 1.00%
Total 0

Current Status
Tax Revenues $5241
Tax Rate 35%
Per Capita Income $18.09
Expenses $347
Net Income $4894
Cashing $5943
Food 161
Production 36
Consumption 25
Net Change 11
Oil 0
Production 0
Building Rate 6
Explore Rate 54
Tech Rate 3

Military Forces
Spies 0
Troops 0
Jets 0
Turrets 0
Tanks 0
Nuclear Missiles 0
Chemical Missiles 0
Cruise Missiles 0

Expense Breakdown
Expenses $347
Military $0
Spies $0
Troops $0
Jets $0
Turrets $0
Tanks $0
Alliance/GDI $0
Land $347
Corruption $0

OK what about this scenario? I've tried with 18000 and 90000. Neither comes close to the PCI of $18.09
22.5 * (1 - Taxrate) * (1 + ((Networth/Land)/18000)) * (1 + (2 * (Ent/Land))) * BusTech * GvtPCIBonus
22.5 * (1 - 0.35) * (1 + ((4989/100)/18000)) * (1 + (2 * (6/100))) * 1 * 1.2 = 19.710479879999998

I've never tested it, but what do you get when you sell something on the market, and it is 32.32? What if it's 32.62?

Does it round to the nearest whole number or always round down on market transactions?

The formula you're using only determines MAX PCI not CURRENT PCI. But the modifier is correctly changed from 90k to 18k. You have that right. Current PCI growth is I think this:

=Round(If(MaxPCI>=CurrentPCI,MIN(pciA,-pciC),-MIN(pciB,MAX(-pciA,pciC))),2)
pciA =(1+0.03*(1-Taxrate))*(1-0.5*PopGrowth/NEWPop)*CurrentPCI-CurrentPCI
pciB =(0.035+0.3*Taxrate)*CurrentPCI
pciC =CurrentPCI-MaxPCI

PCI can decrease even if Max PCI is greater than or equal to CurrentPCI. This anomaly occurs when there's rapid population growth. At least....that's what I have written down in a random notepad and I get the feeling that's not all the information you need. I did math this whole thing out once, but it all looks like chicken scratches from 10 years ago to me now. Might as well be Greek. Hope this helps.

Edit: I'm also pretty sure your PCI starts at 20.00 rather than the number for MAX PCI, and floats using that formula to what it should be.

Also, re: the beginning part of this discussion, it isn't just NW and population growth formulas where 1.999=1, but those are the two that come to mind first. There's a few formulas that work that way.

Yeah. Seems he uses floor() on selective functions.

BTW, thanks for that formula. I'm going to math it out & hopefully it works. You're 100% correct about the 20 PCI,. I noticed the inaccuracy earlier today.

What's "NEWPop"? Is that same as current population?

Yeah so when you take a turn, CurrentPop + Pop Growth = NewPOP

In other words, it's what your pop will be after your next turn taken, and the formula shows how much your PCI changed within the turn based on said growth. The PCI change is based off the land/pop/building change that occurred in the turn, so the formula requires basically how much your pop will grow, and how much pop you will have to determine the turns PCI change.

And the output of the formula is usually around +/- 0.04. It doesn't output the new PCI, but rather just the change. You have to go turn 1 PCI + the formula = New PCI.

For example, if you are building 45 Ents, you would want your +45 in the the formula as well.

Here's an example spreadsheet which will probably need some converting/love, but I've tested my current countries and it seems to be working fine. I think the information is correct.

https://docs.google.com/...Upf8-wPo/edit?usp=sharing

For some reason my population growth and loss formulas are landing +/- 1% civis off. Idk if they have been changed but my testing isn't exactly nailing pop growth on the nose. For example, I just got 372 in my country that the growth formula says I should have gotten 377 for. Donno why. I could see qz giving max pop and pop growth a small change somewhere throughout time...so feel free to make edits if I have something off. Otherwise it seems to be a totally working PCI calculator, though I'm showing your example test country needing about a 0.45 PCI correction and a 9 civilian loss on your next turn, so a few values might be a hair off, particularly max pop.

It's also TRUNC() not FLOOR() (not that it probably matters). For example, pop growth:

Min(BioFactor*Trunc(Round((MaxPop-CurrentPop),0)/3,0),BioFactor*Max(40,Round(0.03*(1-Taxrate)*CurrentPop),0))

and loss:

Min(Round(BioFactor*(0.05+0.15*Taxrate)*CurrentPop,0),Trunc(BioFactor*Round((CurrentPop-MaxPop),0)/3,0))

Pang has lots of junk in his Trunc. The NW formula is just =Trunc((Troops*0.5)+(Jets*0.6)+(Turrets*0.6)+(Tanks*2)+(Spies*1)+(Tech*2)+(Land*45)+(Buildings*35)+(Money/20000)+(Food/1000)+(Missles*2500)+(Pop/6)+(Oil/100))

Update, made some small adjustments in the spreadsheet and fixed a small mistake I'd made. Seems to be working for me perfectly now after a couple tests in alpha.

Thanks Derrick., checking it out.

Something is still slightly off..

The Status of the Republic of Yyy (#2616)

The Basics
Turns Left 120
Turns Taken 5
Turns Stored 120
Rank 2530
Networth $4989
Land 100
Money $37,585
Population 828
At War No
GDI Member No

Land Distribution
Enterprise Zones 6
Residences 0
Industrial Complexes 0
Military Bases 0
Research Labs 0
Farms 0
Oil Rigs 0
Construction Sites 4
Unused Lands 90

Technology
Military 0 100.0%
Medical 0 100.0%
Business 0 100.0%
Residential 0 100.0%
Agricultural 0 100.0%
Warfare 0 0.200%
Military Strategy 0 100.0%
Weapons 0 100.0%
Industrial 0 100.0%
Spy 0 100.0%
SDI 0 1.00%
Total 0

Current Status
Tax Revenues $5241
Tax Rate 35%
Per Capita Income $18.09
Expenses $347
Net Income $4894
Cashing $5943
Food 161
Production 36
Consumption 25
Net Change 11
Oil 0
Production 0
Building Rate 6
Explore Rate 54
Tech Rate 3

Military Forces
Spies 0
Troops 0
Jets 0
Turrets 0
Tanks 0
Nuclear Missiles 0
Chemical Missiles 0
Cruise Missiles 0

Expense Breakdown
Expenses $347
Military $0
Spies $0
Troops $0
Jets $0
Turrets $0
Tanks $0
Alliance/GDI $0
Land $347
Corruption $0

PCI is 18.09
The calculator reads 18.08

Hmmmmmm.

You'd assume it's rounding somewhere. The long PCI without any truncating or rounding is 18.084...so something in the thousandths isn't getting rounded correctly.

My thought is this...and it might be a long shot, but it's a thought nonetheless. That the Advisor is displaying a rounded number and that the calculation is actually to a much further digit. In other words, when your PCI rose to 17.60 the turn prior, it was actually 17.601 or more. Though it would require that you know your current PCI to the further digit to begin with.

I've got to believe it's something with the rounding anyways for it to be 0.001 off like that. My thought is that PCI change isn't actually rounded, just the display is. And like I say that 17.60 Current PCI number was somewhere in the thousandths more. I'll move the rounding in the spreadsheet to show you what I mean.

I know from testing that tech percentages at least go to a much further decimal than the game displays as well, and the pop change formulas vary from in game experience because I don't have the full formulas for tech entered in the spreadsheet... Perhaps it's just that the rounding in the formulas is only for display.

I can confirm what Celphi says that PCI has further hidden decimals (I dont know how many)

The reason LaF says the PCI growth formula isnt known is that this PCI growth formula is the one for E2025 and pciA component is wrong in some instances (or there might be a pciD component that we dont know about that takes effect in some extreme cases).

This PCI growth formula is attributed to those who reverse engineered E2025 (Revelat0r, slagpit, QZ along with the help of the ones that found formulas) and those formulas were usually correct or correct-ish for e2025. All formulas were not in fact found out. And not all formulas were copied straight into EE, some were changed or have been changed along the way. For instance the PCI formula (not PCI growth) had a notable change in one of the terms, and the spying formulas were never found out and have had additions to them along the way.

I believe also that BPT is rounded differently now than in E2025, but thats just from my memory and I can't test that theory. Not sure if anyone remembers.

It doesnt appear to be a rounding issue,. because as I continue to take turns and compare no matter how many turns it's always off by .01 pci. It's almost like there's a component (like Gerdler said) within pciA formula that's being missed that's only applied with pci != maxpci AND only after you build ENTs.

Yeah I'm definitely not smart enough to reverse engineer the difference without a pile of testing... You'd assume it's something in the pciA like you both note.

It is insightful that PCI actually goes to a further decimal than display tho. That does help clean some of the rounding (or lack therein up).

I'd be curious and maybe could test the 0.03 just being a hair more, because the beginning part of that formula at .35 tax rate, it wouldn't change much. I'll maybe mess around with 0.05 and such at 70% and 0% rates and see if there's just a small change to it.

If it seems more complicated than that, tho, I might have to give up and say that's as close as we've got haha. I will make notes of instances that it's further off to help someone reverse engineer it further in the future.

You're right that there's definitely some sort of rounding with some of them,. because even if you build 4CS and 6Res all the numbers match up except,. the cashing appears to be .01 greater than 20% of revenue.

I messed around with some stuff. It's definitely not just a flat increase to any factors, but realistically could still be just a balanced factor change within pciA (in testing, I've also had it go down 0.01 off as well as go up 0.01 down). The return to Max PCI seems to act faster in the current formula in general, and it might just be a slight tweak to each side of the integers in pciA

Just totally a random thought here but what if the rounding was done differently. I am a certified Michigan Aggregate Technician and we don't round numbers like math should.

Any odd number followed by a 5 is rounded normally. However any even number followed by a 5 is rounded down.

Example: 0.5 is rounded down to 0, but 1.5 is rounded up to 2.

That is only when there are no other numbers following said 5. 0.501 would round up to 1. I'm trying to find a link to the Michigan test method that explains slightly better I don't have my manual on me.

This is the best I can do without paying for the fluff. I'll post more about it when I get to work if anyone thinks this may be significant.

Most standards recommended for science and technology, led by ASTM E29, make use of the “five-even” rule. By this rule, for example, the value 76.50, rounded to the closest unit, becomes 76, to make the last significant digit of the outcome even. In the same manner, 89.50, rounded to the closest unit, becomes 90.

Hmm. I wouldn't doubt it because of the way the formula is acting, however, knowing pang and slag, these are generally one of three or four Javascript commands. The formulas themselves can be complicated but it's usually min, max, trunc or round and a formula. Just from experience. Usually it's a spreadsheet formula on these functions.

ee is programmed in php so the rounding will be same for all languages

I spoke with qzjul and he provided me the cashier calculator. Even the calculator doesn’t match the game.

He even showed me the code for the pci growth which is slightly different than what is up above but even after I made those changes it’s still off. It’s more than just the PCI that’s causing the issue.

I’m kinda done with trying to figure this out- I was hoping to try to find the best strat using machine learning newest algo but I can’t do it unless I can exactly match the environment.

Do you mind if I look at his calculator? I feel like knowing what we know about the formulas, there's likely a changeset that's throwing it off, and we are magic at adjusing. I feel confident in my ability to correct his PCI formulas.


.......and even if I can't, man. This is a game we love without active dev. It wouldn't hurt us to, doomsday scenario, reverse engineer this. 2025 is soon.

And also I don't feel like 0.01 in PCI growth would change an AI's movement....just being honest. For machine learning, I think the provided information is close enough for success...

It makes a huge difference. It throws off a ton of decision making.

Contact me via irc for link