How to calculate the battle odds?

[TR000]

Hi guys,

Do you know how can I know how are the battle odds calculated?

Thank you.

Regards.

[Kaiser]

It's a secret code that admin has never revealed. Though I believe those silly Grämlins may have cracked it

[Lord Brendan]

It's a secret code that admin has never revealed. Though I believe those silly Grämlins may have cracked it

I'm sure (DAC)Syzygy will show up any minute now and taunt us all with his secret formulas.

[Noob Cake]

That he will not share with us

[o ya baby]

I've seen it somewhere.

[Deep Blue C]

This is something I'm working on at the moment. Can anyone remember what the tech bonus is for the battle strength calcs.

I'm working around the idea that

Attacking battle strength =(2*soldier efficency +25*tanks)*1.01 for daytime attack* Defcon + Tech bonus(*1.05 for night-time)

Defending battle strength = (2 * soldier efficency +30*tanks)*1.01 (daytime) *defcon +Land Bonus +Infra bonus +Tech bonus *(1.05 - night)

Then Attacker odds = Attacking strength /(Attacker strength +defender strength) * 100%

Defender odds, similarly =Defender strength /(Attacker strength +defender strength) * 100%

There are some guesses in here -e.g. that defcon only affects military bonus, not tech etc. - this on evidence that when defcon was altered by spying and no other changes were made, the changes in odds was not linear.

Now I'm sure the tech bonus was discussed when the changes were made to tech and NS a few months ago, but I can't find the threads now.

[Lord Brendan]

That he will not share with us

Well if he shared them with us, they wouldn't be secret would they?

[Deep Blue C]

Attacking tank infrastructure bonus reduction = 1% reduction per 100 tanks up to 7500 tanks

Any ideas what that actually means or refers to?

Edit..

NVM

Defenders in ground battles receive a bonus to their battle odds based on the level of infrastructure of the defending nation. (This bonus is reduced based on the number of attacking tanks. See the Tanks description for more information.)

C'mon Deep , RTFM!

Edited April 30, 2008 by Deep Blue C

[Spartan519]

Can you put all this Math in English please?

[Decomposition]

If the various factors are linearly or quadratically related to battle odds then it should be possible to calculate the relative weights - assuming you have enough data points which any decent sized alliance should be able to put together. Ah.... the wonders of regression.

[LeVentNoir]

Deep blue, lets ask every body to post our odds on our boards and lork it our from there.

[Deep Blue C]

Spent some time trying this with some data, and I couldn't get it to work. Realise now it can't be the above as that would give battle strength to defenders with 0 troops - you wouldn't get 100% odds ever. I have a couple of other theories - All products , or add the bonuses then mulitiply by the tank/soldiers.

LVN - thread already there. Also, for the way I'm working, one person supplying repeated oddds as things gradually change is better at present.

Spartan - it's not worth translating until it is close to working - but don't worry, if I get something that works I'll share it (with NATO at least. )

decomposition - I was struggling to do something like that because the odds are only a ratio. Once I crack the format of the formula, the individual factors should come out as you suggest.

[unbiased mod]

If the various factors are linearly or quadratically related to battle odds then it should be possible to calculate the relative weights - assuming you have enough data points which any decent sized alliance should be able to put together. Ah.... the wonders of regression.

regression is good, but searching faults and 95% probability's on those regressed numbers, now THAT is

[Magicman657]

Spent some time trying this with some data, and I couldn't get it to work. Realise now it can't be the above as that would give battle strength to defenders with 0 troops - you wouldn't get 100% odds ever. I have a couple of other theories - All products , or add the bonuses then mulitiply by the tank/soldiers.

LVN - thread already there. Also, for the way I'm working, one person supplying repeated oddds as things gradually change is better at present.

Spartan - it's not worth translating until it is close to working - but don't worry, if I get something that works I'll share it (with NATO at least. )

decomposition - I was struggling to do something like that because the odds are only a ratio. Once I crack the format of the formula, the individual factors should come out as you suggest.

Keep in mind that this is not a formula; it's programmed code. There could be exceptions via if statements for situations like that which override whatever the general formula would give.

Focus on making it work for as many possibilities as you can and if it doesn't work for special circumstances, it could merely be the result of logic code taking precedent over the formula.

[Deep Blue C]

OK, I'm pretty close with the following...

Attackers battle Strength (S) = (Troop efficiency * 2 +Tanks * 25)*Technology bonus.

Defenders Battle strength (D) = (Troop efficiency * 2 + Tanks * 30)*(Technology bonus +Infra bonus+Land bonus)

Attacker odds = S/(S+D) Defender odds = D/(S+D)

Tech bounus = tech level * (?) around 15 to 20

Infra Bonus = Infra * ? (less 1% for every 100 attacking tanks up to 7500) around 4

Land bonus = Land * (?) no idea but n0t large

The missing factors still need to be worked out, but it should be something to work on.

Edited May 3, 2008 by Deep Blue C

[Zangmonkey]

If the various factors are linearly or quadratically related to battle odds then it should be possible to calculate the relative weights - assuming you have enough data points which any decent sized alliance should be able to put together. Ah.... the wonders of regression.

Here's where you logic hits the wall:

You have to first guess what relevant order that parts account for in the formula.

For example, having inputs A B anc C, it may be that the result is proportional to the square or cube of an input.

Without making testing the exponential orders, regressions will be difficult to use.

For an easy example, try and crack the land purchase formula. Once you've properly guess that order of the input values a regression works very well. My regression's t^2 is .989, so the reverse engineered formula is pretty accurate.

Lastly, you are not going to extract the precise formula using regression, typically you'll just get a closely related function.

Edited May 2, 2008 by Matthew PK

[Syzygy]

you can get the exact formula with very large series of tests and by doing these tests 'semi-professional'. As example: If you deploy the same amount of troops in the same DEFCON against the same enemy nation which only is buying fixed amounts of land (lets say 100miles) between the attack-attempts (you dont perform real attacks, just look how the odds change), you will get a pretty exact factor for land-influence over time. At least if you do that for dozens of nations.

Same is for tech, infra and so on. Important is, that you run large series of tests with changing only ONE variable at a time.

That being said: That is the exact reason why I see absolutely no need in giving away formulae which our research division has encountered during hard work and spent hours and days to get them. That has nothing to do with arrogance, but with the simple stance not to support laziness. Everyone can encounter them, if he just want it hard enough.