Combat Calculator redux

[Darkbrady]

During my testing I also noticed that scouts not only don't seem to gain the Intelligence bonus but they were taking more damage than expected..

Reverb found a similar problem some time ago, that units with less than 50 defence somehow reversed on themselves and took incorrect damage.
I think s/he wound up workin' out that treatin' the 50- defence units (only workers and scouts, iirc) as if they had 50 defence.
Not sure if this applies to what you're talkin' about, far too tired to compare two different people's note, but seems like it's worth a mention if it makes your life any easier.

[Birtles]

Aha, thanks for that DB. I already knew about the 50 def minimum but you saying 'reversed on themselves' made me think...
Just reran the numbers from my tests and the results work if both the Intelligence and Iron Working bonus are removed from the formula. I'll do some more testing to confirm it.

[Onslaughtx]

I don't see the big deal. Even without a hero, the cavalry would have enough damage to kill each group they attacked, each round. The only difference is the randomness that the game adds in - such as mirrored attacks having different casualties every time.

[Birtles]

I don't see the big deal. Even without a hero, the cavalry would have enough damage to kill each group they attacked, each round.

You've missed the point. I sent enough cavalry to make hero attack a non-issue, the focus is on how Intelligence reduces casualties. This is useful to know throughout every stage of the game from conquering your first archer-defended valley to facing massive one-round suicide waves against your city.

The only difference is the randomness that the game adds in - such as mirrored attacks having different casualties every time.

You have any examples of this randomness?

[Darkbrady]

No one who works with code and mechanics likes to think about a random factor; it makes all the work about a hundred times harder for no real benefit in the end...especially when there's such a complicated mathematical code.
If there was a heavy random element then wouldn't excercise results vary?

Birtles, set up another similar test. Get your two heroes to attack two or three of the exact same defences in the same location. If there's a random factor then all six reports should wildly vary. If there's no/very little random factor then the results should only vary between the two different heroes.

[Neokarasu]

Fact: there is NO RNG in combat calculations.
You can do a very simple test of throwing the exact same attack with the exact same hero at exactly the same Valley. You will get exactly the same losses both times.

So the fact that the casualty count is different points to a different factor. Int giving a buff to Defense is a very solid explanation of why losses calculated by hand is higher than the actual losses in combat.

[Birtles]

Birtles, set up another similar test. Get your two heroes to attack two or three of the exact same defences in the same location. If there's a random factor then all six reports should wildly vary. If there's no/very little random factor then the results should only vary between the two different heroes.

Unfortunately I can only do small scale battles as the account that I 'borrowed' has just a few thousands troops. Here's another selection of results from yesterday's testing :-

Heroes (Intelligence); Wanda (19), Maxwell (75), Ed (138) also guest star Roderick with an attack of 75 (woot).

Supporting cast; 5000 pikemen, 100 cavalry.

Baseline losses, without the Intelligence factor, would be 62 with no defending hero (NH) and 94 with a defending hero (H).

Wanda's casualties were 60(NH) and 91(H)
Maxwell's were 53(NH) and 81(H)
Ed's were 45(NH) and 69(H)

Relevant Techs were all 10 for the pikemen and 9 Military Training for the cavalry.

pics (NH)




pics (H)




Formulae

Attackers' damage

B/100(H + T + 100)N

B base attack, H hero attack, T Military Training (percentage), N number of attackers

Defence

1 - (((I + T + 100)/100)D)/1000)

I intelligence, T Iron Working (percentage), D base defence

Life

L x 1.T

L base life, T Medicine (percentage)

Effective Life

L/D

L Life (as derived from above formula), D Defence (ditto)

Casualties

A/E

A Attackers' Damage (as derived from above formula), E Effective Life (ditto)

[Neokarasu]

Defence

1 + ((I + T)/100)D

I intelligence, T Iron Working (percentage), D base defence

Nitpick:

Isn't it 1 - (I + T) * D / 100,000 ?

Because from your casualty formula:

losses = A / E = A * D / L
and it would make no sense that you would lose more when D goes up.

[Birtles]

Edited, thanks. That'll teach me to read formulae from spreadsheet cells. Maths looks plain ugly when linear, this is much prettier.



Strictly speaking it needs another pair of brackets. Speaking of which, you should really add some to your own formula, seeing as we're picking nits

[Birtles]

Did another couple of larger scale tests (still relatively small scale). Just report links this time, seeing as they're pretty much duplicating the first post.

http://battle10.evony.com/default.html?logfile/20091128/b6/6c/b66cf8fd5198abd932eeee4865bd74ca.xml

http://battle10.evony.com/default.html?logfile/20091128/f2/36/f23635d87578d4a303decc3a26a241d8.xml

Both attacks on the same level 10 valley. In regards to valleys, I have a strong suspicion that the valley hero stats are re-rolled upon capture.