|
On December 31 2008 13:15 food wrote:Show nested quote +On December 31 2008 13:08 SonuvBob wrote:
I think you need to check each individual build command with a * to see if it's on minerals/gas/unbuildable terrain. Most *'s will be false positives. he says that if you not hacking you wont get those actions recorded, so when i was placing my hatchery at an unbuildable terrain( which happens a lot) it did get recorded. Check any known zerg replays, im sure everyone has this. In his opinion tho, this only gets recorded if youre hacking, which seems untrue so far, thats my point. edit: ok sec ill link the replay in a sec
I just tested this and it didn't show the actions in the replay... -_-
|
Me too, tried building a hatchery in the middle of luna, on minerals, on top of an enemy hatch, etc, not a single build command in the rep.
|
well look up the replay i posted and tell me what my actions marked as * are then
action 3995
4005
4560 - for extractor
he has many too, supply, bunker etc, but check mine
|
They're whatever bwchart marks with a *, dunno how it determines them. The point is that trying to build on minerals/unbuildable terrain only shows up with Oblivion, so if you can find one of those (which would be among the *s), it means hack.
Seems like finding them without an automated tool would be really hard.
|
see the action that i just listed, i believe i was building a hatch/extractor, nothing else.
|
On December 31 2008 13:23 food wrote:
well look up the replay i posted and tell me what my actions marked as * are then
I just checked a number of them on python and none of the ones I checked so far are on unbuildable terrain, therefore they are not hack detections, you must have had a unit in the way or something.
It's only a detection of a hack when it was on unbuildable terrain.
|
On December 31 2008 13:28 food wrote:
see the action that i just listed, i believe i was building a hatch/extractor, nothing else.
As I said, the hatch one (one of the ones I checked) was not on unbuildable terrain so there is no hack there, and the one with the extractor doesn't have a star...
|
oh shit, maybe it does work then^^
|
On December 31 2008 13:30 food wrote:
oh shit, maybe it does work then^^
|
Ah, *s normally come from sending multiple build commands when you have the money, where only the last one, if any, actually results in a building (select an scv and spam bsclickbsclickbsclickbsclickbsclick)
|
On December 31 2008 13:32 SonuvBob wrote:
Ah, *s normally come from sending multiple build commands when you have the money, where only the last one, if any, actually results in a building (select an scv and spam bsclickbsclickbsclickbsclickbsclick)
Aha, cool to know. I wonder what the occasional ?'s come from (they're in the same column as the *'s). Who actually made bwchart? It'd be nice to be able to ask the creator questions about things in it that were never documented 
|
|
|
why the fuck would you want to build something on terrain you haven't seen yet? that's just plain retarded...
|
On December 31 2008 13:50 Xeris wrote:
why the fuck would you want to build something on terrain you haven't seen yet? that's just plain retarded...
I don't think it's whether you can see it or not, this is whether it's unbuildable terrain or not, like if you try to build a building and you misclick and hit a cliff or ramp, or if you go to build your assimilator and you miss the gas geyser, or if you go to build something like a cannon and you hit the minerals instead of hitting the ground.
|
On December 31 2008 13:50 Xeris wrote:
why the fuck would you want to build something on terrain you haven't seen yet? that's just plain retarded...
whoops nvm
|
Ah yeah, I guess gas would be the best thing to look for.
I'm not sure if *s are the only actions you'd have to look at, bwchart doesn't * all the failed build commands. So we need a program that checks the location for all non-refinery/assim/extractor build commands against mins/gas/invalid terrain (taking into account the size of the building) and checks all refinery/assim/extractor build commands to make sure they're over a geyser.
|
Alrite with the 30% estimation we can make the appropriate models in how many replays would you have to test on average before you can catch a * mark.
Suppose we're dealing with a bunch of replays of a suspecting hacker, and each of them has a .3 chance of showing a * mark. How many replays do we need to check, on average, before we can catch someone?
This is a "geometric distribution", so we're looking at the expected value of the geometric distribution. Which is:
1/0.3 = 3.333
So this is a very powerful tool in catching them, as on average, you only need to check 3-4 replays
|
Germany2896 Posts
So there are three checks if a building can be created:
1) In the GUI before the command is recorded
2) At the time the command is received
3) At the time the unit arrives at the target location
Only the second one indicates hack. And only if in the time between 1) and 2) nothing changed to block the build.
Do I understand this correctly? And does this also catch hackers using oblivion in safemode which deactivates all features zynastor considers detectable. And I guess build anywhere would be considered detectable.
|
Calgary26003 Posts
Keep fighting the good fight. I love you.
|
On December 31 2008 16:30 evanthebouncy! wrote:
Alrite with the 30% estimation we can make the appropriate models in how many replays would you have to test on average before you can catch a * mark.
Suppose we're dealing with a bunch of replays of a suspecting hacker, and each of them has a .3 chance of showing a * mark. How many replays do we need to check, on average, before we can catch someone?
This is a "geometric distribution", so we're looking at the expected value of the geometric distribution. Which is:
1/0.3 = 3.333
So this is a very powerful tool in catching them, as on average, you only need to check 3-4 replays
It's not about just finding a * mark, it's about finding a * mark that is due to unbuildable terrain as opposed to other buildings/units being in the way.
Jeez I hope flag having released this method doesn't create a slew of people checking replays and accusing others of hacking because their build actions have a * next to them 
|
|
|
|
|
|
EPT
