Combat outcome calculation (berserk)

[SlowThinker]

In past I played SX maps with BfW 1.4 and noticed an extremely long computation if two very strong units faced one another.
Then I had heard the algorithm was improved, but now I have heard the outcome computation is still very slow.

The slowness of the algorithm is suspicious for me, I think it should be quite fast in case it was done in C++.
Can anyone tell how is the algorithm done roughly?

[Max]

http://forums.wesnoth.org/viewtopic.php?f=21&t=36141

[SlowThinker]

That thread is related to the topic, but it doesn't explain the algorithm that computes the outcome matrix...

[Crab]

roughly: create hp distribution matrix, set to initial hitpoints, 'simulate' all attacks, changing the hp distribution matrix in the process, after each attack. see
http://svn.gna.org/viewcvs/wesnoth/trun ... iew=markup

[Max]

In attack evaluation phase it builds a matrix of every possible outcome, and you can see this also in the damage calculations preview. If a unit has thousands of hitpoints, drain and/or berserk then it will cause a real slowdown.

it just doesn't scale well here. has nothing to do with c++...

http://mirror.linux.org.au/linux.conf.a ... gaming.ogg
(~1h50)

[SlowThinker]

... 'simulate' all attacks ...

There is no need to simulate anything:

I attached a file with a simple algorithm that counts a chance to kill for each specific strike. The distribution matrix may be computed from the spreadsheet this way:
for each cell in columns R (chance attacker kills now) and S (chance defender kills now) one can easily count a partial hp distribution matrix (rather hp distribution *vector*, because one side always dies), then add distribution matrixes of individual cells together.

the algorithm contains a mistake. Will you be the first one who finds it out?
Edit: the attached file contains formulas in cells now.

[Drakefriend]

... 'simulate' all attacks ...

There is no need to simulate anything:

I attached an Excel file with a simple algorithm that counts a chance to kill for each specific strike. The distribution matrix may be computed from the spreadsheet this way:
for each cell in columns R (chance attacker kills now) and S (chance defender kills now) one can easily count a partial hp distribution matrix (rather hp distribution *vector*, because one side always dies), then add distribution matrixes of individual cells together.

But you also have to calculate the hitpoints for any combinition of hits, which is important for the surviving unit, and for normal combat.

[SlowThinker]

But you also have to calculate the hitpoints for any combinition of hits, which is important for the surviving unit, and for normal combat.

Why would I need it? I believe all I need is the aggregate hp distribution matrix.

____________________________________________
Edit: My apologies, the original spreadsheet didn't contain formulas, only values.

But why nobody complained? I said "I attached a file with an algorithm", but there was nothing resembling to an algorithm ... 6 people downloaded it ...