[Freeciv-Dev] Re: Bombers/Fighters no longer obsolete (PR#1628)

[Raahul Kumar <raahul_da_man@xxxxxxxxx>]

--- Gregory Berkolaiko <Gregory.Berkolaiko@xxxxxxxxxxxx> wrote:
> On Fri, 28 Jun 2002, Raimar Falke wrote:
> 
> > On Fri, Jun 28, 2002 at 04:52:26PM +0100, Gregory Berkolaiko wrote:
> > > On Wed, 26 Jun 2002, Ross W. Wetmore wrote:
> > > 
> > > > Of course this implies that your attack power formula actually 
> > > > means something rather than being one of many heuristic formulae
> > > > none of which describes reality in any accurate way :-).
> > > 
> > > I would like to point out that the formula used by Raahul does describe 
> > > reality in a pretty accurate way.  And I can give you a solid 
> > > justification why it is so.  As a matter of fact, I'll give you few hints
> 
> > > and can fill in the rest.  Since excercise is good for you ;)
> > > 
> > > We have an attacker A with stats (S_A, HP_A, FP_A) and 
> > > a defender D (S_D, HP_D, FP_D).  We want to decide which one is stronger.
>  
> > > And of course by "stronger" we mean "which one has larger probability to 
> > > win a fight when A attacks D".
> > > 
> > > Lets assume FP_A = FP_D = 1.  Otherwise we can substitute 
> > >   HP_A <- HP_A / FP_D
> > > etc.
> > > 
> > > 1. You can compute the win_chance and try to analyse it.  But that is too
> 
> > > tough :(
> > > 
> > > 2. Alternatively you can say, lets count how many successful hits will A
> > > score, _on_average_, before it dies.  If this _average_ number of hits
> > > will be greater that HP_D, then the defender will die, on _average_,
> > > before the attacker does.  Which implies that the attacker will succeed
> in
> > > more fights than the defender.
> > > 
> > > 3. So we are running a random process giving us 0s and 1s.  0 means A won
> 
> > > the round, 1 means D won the round.  The question we have to answer is
> how 
> > > many (on avergae) zeros will come up before we see HP_A ones. 
> > > 
> > > 4. Probability of having 0 is 
> > >   S_A / (S_A + S_D),
> > > probability of getting 1 is
> > >   S_D / (S_A + S_D).
> > > The rounds are independent.
> > > 
> > > 5. Now, instead of inventing 20 + 20 formulae, you, Ross, should get 
> > > yourself a "probability for kids and gamblers" book and calculate the 
> > > answer. :)
> > 
> > This is all clear but misses the point. The hypothesis is the military
> > power/value of a unit is "S_A * HP_A * FP_A". And I don't see a line
> > in code in combat.c which multiplies the three except
> > get_defense_rating which as the name suggests is a rating.
> 
> You think it misses the point because you didn't do the calculation.  Or 
> maybe even didn't read the email ;)
> 
> > Note that all these leave out other factors like increased move_rate.
> 
> I am afraid the increased move_rate cannot enter the win_chance 
> calculations.
> 
> G.

Thanks for the explaining to the unenlightened heathen Greg ;). By the way, I
fully agree that attackpower and defencepower should be calculated seperately.
The engineer unit is a good example. With an attackpower of 0, no amount of
engineers will ever kill any defending unit. This is explicit in all my calcs.

This simple point is what I have been attempting to hammer into Ross/Raimar.
The 

hp * fp * attack encapsulates one very important bit of info. 

If fp = 0 or attack =0 the unit is worthless for attacking. It also shows the
very strong equivalence between all 3 stats.  Of course flags like Aegis,
IGWALL, etc matter, and so do defensive effects but we ignore that for a
straightforward unit comparision. 


Aloha,
RK.

History teaches us that men and nations behave wisely once they have exhausted
all other alternatives. -Abba Eban

__________________________________________________
Do You Yahoo!?
Yahoo! - Official partner of 2002 FIFA World Cup
http://fifaworldcup.yahoo.com