Loading...
  OR  Zero-K Name:    Password:   

How can the mex spot dedection fail so badly?

20 posts, 1044 views
Post comment
Filter:    Player:  
sort
12 years ago
GreenHaven-V3
Spring 91.0
Zero-K 1.1.3.6

Very clear separated, spaced, small quadratic mex spots.
But: too many mex spots and some a bit apart from where they should be.

Here is a screenshot:
+0 / -0

12 years ago
Does it still give you the designated metal income?
+0 / -0
12 years ago
maybe too small extractor radius value in your smd/config file?
+0 / -0
12 years ago
@MauranKilom> Does it still give you the designated metal income?

If you have to build 2 mexes to get same m/s, you spend 75m, 75e, 7.5s (+time for walking) more.
Also decreases OD

If you have more m per spot, it may be imba too


DErankKlon> maybe too small extractor radius value in your smd/config file?

I neither made the map nor the config file.
This issue is not related to me, but to whole zk.


@ Who-has-written-it: How can the code fail on such a 'simple' input?
+0 / -0
I'll call JPEG artifacts just for teh lulz.

Edit:
Lol AFK_Kitty, how is that "easy"?

1. It's a jpeg, so say hello (no pun intended) to artifacts.
2. There's random grey crap all over the metalmap. The fetcher picks up on that.
3. The "grey crap" accumulates to a lot of brighter spots, which you can't blame the algorithm for recognizing as small mex.
+0 / -0
12 years ago
You could sort the black-gray stuff out.


Just ignore every square with less than 10..20% of the max. value...

Each such square would be as big as a 4r*4r (r=radius of mex)
Even x/z: 0*4r, 1*4r, 2*4r
Odd x/z: 0*4r+2r, 1*4r+2r, 2*4r+2r
offset = 1: odd x, odd z & 2: even x, odd z & 3: odd x, even z & 4: even x, even z
for 1 to 4: insert into table: { x, z, metal } and update maxMetalFound if required.

Walk through generated table of { x, z, metal } and sort out every square with 10..20% metal or less.

Look in table-entries (=squares - or more exactly their offsets) for best metal coordinates.
Merge/Drop some overlapping mex-coordinates.
+0 / -0
Well what was the intention of the mapmaker by putting that grey stuff there?
Don't blame the algorithm for mistakes of the mapmaker. Ofcourse you can find some procedure for an individual corner case, but that doesn't help at all. There's cases where the grey stuff is supposed to be a mex. There's cases where it's not. There's cases where the input is completely nonsensical and even a human couldn't place mexes properly.

TL;DR: The mexspot fetcher is not the problem.
+0 / -0
12 years ago
I think the map maker's intention was to make the metal map looking more interesting...

> There's cases where the input is completely nonsensical and even a human couldn't place mexes properly

Then players would withdraw map anyway.

> There's cases where the grey stuff is supposed to be a mex.

If it pays back in time or not many other mex spots are close to it.
Maybe only sort out the squares which have low metal AND are close to 'real' metal spots, but max. 10% of map metal.

Because we can set the output manually that wouldn't be a big issue.
+0 / -0

12 years ago
If ZK still used my easymetal algorithm I bet it would detect it better, but instead it uses a different one.
+0 / -0
quote:
I think the map maker's intention was to make the metal map looking more interesting...
What? Where do you even see the metalmap? You obviously didn't notice how dirty it is ingame, so no benefit there. And who cares about the metalmap picture on the map page?

quote:
If it pays back in time or not many other mex spots are close to it.
Maybe only sort out the squares which have low metal AND are close to 'real' metal spots, but max. 10% of map metal.
And i could still create examples of input that would misbehave. Again, looking at a special case and saying "well, here this property is intended and should be treated this way" doesn't work out. Your suggestion ends up in kludging stuff together. What is "close"? What is pay back "in time" (how long)? Why 10% of max metal, what if there's a lot of mexes or very few?
And once you spent 2 hours figuring out reasonable equations for the above, i'll come and break your system with another kind of input.

The purpose of a metalmap is clear: Black where no metal is supposed to be, grey/white where metal/more metal is supposed to be. Don't put grey when you want no mex there.
Would you draw a smiley on the heightap and then complain that the terrain generator picks up on it? No, if you screw with the input for some reason that it's not meant for, it's your own fault (yes i know the map is not made by you).
+0 / -0


12 years ago
Carp 'easymetal' is unsuitable for our purposes. It is worse at discrete spots and would still fail to make a playable metal map in this case. Detection does not have to fail gracefully on maps with stupid metal maps.

It "fails so badly" because as others have pointed out your metalmap is shit.
+0 / -0
quote:
Well what was the intention of the mapmaker by putting that grey stuff there?
Don't blame the algorithm for mistakes of the mapmaker. Ofcourse you can find some procedure for an individual corner case, but that doesn't help at all. There's cases where the grey stuff is supposed to be a mex. There's cases where it's not. There's cases where the input is completely nonsensical and even a human couldn't place mexes properly.

TL;DR: The mexspot fetcher is not the problem.



You are so wrong it makes me puke...
Its ZK who uses some hacks which prevents building mexes anywhere, even more I almost guarantee that this map is older than this "mex something" "feature" which was added to ZK only so far, how you dare to blame mapper for this? Either it was to make it look cooler in F1, or whatever reasons(even if it is accidentally outlined! YES EVEN) with normal spring without any hacks it works as mapper intended so its absolutely ZK fault, mappers in ZK are already anally raped with constantly removed features from what engine offers so DONT DARE TO BLAME MAPPERS for ugly ZK hacks which doesnt work.
+0 / -0
12 years ago
+0 / -0
12 years ago
+0 / -0
12 years ago
sure we can't prepare for everything, but most common cases.
+0 / -0


12 years ago
Not all maps are supposed to be playable with all games.

If you wanna play on fucked up mups, shut up and bite the pillow.
+0 / -0
Actually with introduction of this mex thing some before playable maps mostly with cloudy metalmap become less playable when that gadget still trys to force mex places instead letting them to be built anywhere resulting in weird mex placement. Or when metal patches was intended to be very big and multiple mexes built at same area this mex thing makes only 1 mex which results in completely different gameplay and metal layout which was not intended by mapper.

For example: football field (you have only 1 mex per team)
Epic Shore maps, become less epic, instead having like 5 mex per big patch you have only 1.
Of course there is more maps which names I forgot.

Ahh forgot to mention also if mex'es are very near each other even if its not cloud, they are merged into one which can result in very unfair advantage for other team because of overdrive, there is some good map broken like this also forgot name, but it has discussion on its own map thread...
+0 / -0


12 years ago
What you call "hacks" we call "features."
+0 / -0


12 years ago
Feel free to provide metal configs for maps you think are broken. Any other action on your part is a waste of your time.

I don't care that a few maps were broken and I not supporting cloud metal is actually a feature. Overdrive interacts very unintuitively with cloud maps and maps on which it is possible to grab multiple spots with one mex. The possibility of multiple spot grabbing has not been increased and some effort was put in to maintain it's existence. Check Geyser Plains.

What we do is more than "forcing" metal extractors. Nothing in ZK extracts metal from the map, we have just defined a new resourcing system based on point control which takes data from the old metalmap method if no point config is available.
+0 / -0
12 years ago
KR thats exactly how I call them :P

Dunno how special metal config is made, but isint this going backwards, I mean what is easier to simply disable this feature on maps which it fails or write many metal configs for each map. So simplest solution would be allowing users to vote disable. Actually it should have been used on before broken maps to fix them, not to force on previously working maps, well that was my opinion.

"If it ain't broke, don't fix it."
Don't code fixes that nobody wants to problems that don't exist.

Straight from your book ;)
+0 / -0