EPT[ASL6] Ro8 Flash vs Mini - Page 8
| Forum Index > Brood War Tournaments |
|
YPang
United States4024 Posts
| ||
|
ShloobeR
Korea (South)3809 Posts
| ||
|
Freakling
Germany1529 Posts
On October 16 2018 21:53 rand0MPrecisi0n wrote: How would one go about analysing this type of stuff? Thanks! The root of the problem is that Scarabs are units that can only travel over passable ground. So to ensure that things like cliffs and other major terrain obstacles do not cause Reavers to shoot out Scarabs that will just run around erratically and then dud out (because there is no or only a very indirect path between the Reaver and the target) the Reaver attack is coded to use pathfinding distance instead of simple radial distance in its attack range check. The algorithm that determines ground traveling distance is just a very rough heuristic, though. Across multiple regions it seems to use pathfinding node-to-node distance instead of actual unit positions or linear distances, so if the nodes of semi-adjacent regions (regions in close proximity, but separated from each other by at least one interjacent region/node) are spaced out too far the engine considers that as a long travel distance. This is how far I understand it so far. For the exact algorithmic details I guess I'll have to delve into the OpenBW code at some point :\ Example for the instance that caused the bug in this particular game (I think what Bonyth is referring to is the scene at about 0:51:30, at least that's what caught my attention): + Show Spoiler [show image] + ![[image loading]](https://i.imgur.com/siKbQUH.jpg) I marked the approximate unit positions and highlighted the relevant region nodes. | ||
|
rand0MPrecisi0n
313 Posts
On October 16 2018 23:37 Freakling wrote: The root of the problem is that Scarabs are units that can only travel over passable ground. So to ensure that things like cliffs and other major terrain obstacles do not cause Reavers to shoot out Scarabs that will just run around erratically and then dud out (because there is no or only a very indirect path between the Reaver and the target) the Reaver attack is coded to use pathfinding distance instead of simple radial distance in its attack range check. The algorithm that determines ground traveling distance is just a very rough heuristic, though. Across multiple regions it seems to use pathfinding node-to-node distance instead of actual unit positions or linear distances, so if the nodes of semi-adjacent regions (regions in close proximity, but separated from each other by at least one interjacent region/node) are spaced out too far the engine considers that as a long travel distance. This is how far I understand it so far. For the exact algorithmic details I guess I'll have to delve into the OpenBW code at some point :\ Example for the instance that caused the bug in this particular game (I think what Bonyth is referring to is the scene at about 0:51:30, at least that's what caught my attention): + Show Spoiler [show image] + I marked the approximate unit positions and highlighted the relevant region nodes. Sick explanation man, but what I mean is which tools do you use/recommend if I wanted to do this type of analysis myself? Thanks a lot anyways | ||
|
Freakling
Germany1529 Posts
| ||
|
konadora
Singapore66155 Posts
| ||
|
iFU.pauline
France1538 Posts
| ||
|
O.P.
Sweden109 Posts
On October 17 2018 00:16 iFU.pauline wrote: Frankly I have never seen a protoss winning vs Flash doing 2 bases carrier. Is 2 base carrier even viable if terran scouts it? It feels very cheesy. | ||
BigFan
TLADT24920 Posts
On October 17 2018 00:28 O.P. wrote: Is 2 base carrier even viable if terran scouts it? It feels very cheesy. I personally want to say no since as a terran, you can scan it and find out what the protoss is up to, then strike right before the carrier count goes up. However, I don't play protoss so I'm not sure either. Can a protoss player give up more info on this? | ||
|
rand0MPrecisi0n
313 Posts
On October 16 2018 23:45 Freakling wrote: ScmDraft obviously (that's what I created that image with). Ah ok, I didn't know scmdraft did this sort of thing. | ||
|
plast1c
Germany99 Posts
On October 17 2018 00:16 iFU.pauline wrote: Frankly I have never seen a protoss winning vs Flash doing 2 bases carrier. Bisu won vs Flash with carriers in ASL 3; it was 3 base carrier though. | ||
|
Bonyth
Poland553 Posts
| ||
|
starkiller123
United States4030 Posts
| ||
|
Freakling
Germany1529 Posts
On October 17 2018 00:28 O.P. wrote: Is 2 base carrier even viable if terran scouts it? It feels very cheesy. If the map is very favourable to Carriers (i.e. has cliffs and long ground paths, limiting Goliath mobility, like Katrina, Pathfinder – or Transistor) it seems to often work. | ||
|
Brainojack
Canada195 Posts
it looks like Flash glances over at Mini to see whats going on then quickly remembers the the cameras are on right in his face. Youtube Jump | ||
|
Barneyk
Sweden305 Posts
And its to bad that the ASL youtube timestamp kinda spoils it. :/ | ||
|
prosatan
Romania7936 Posts
On October 17 2018 02:10 Brainojack wrote: I laughed at the end of the circuit breaker match, just prior to the fantasy gg timing. it looks like Flash glances over at Mini to see whats going on then quickly remembers the the cameras are on right in his face. Youtube Jump When I watch it now I feel so sad for Mini  | ||
|
JoinTheRain
Bulgaria408 Posts
And for game 1, I wonder, how is 12 nex still a thing, I have yet to see a decent terran lose to it in a map where the 2nd is not accessible from a ramp. | ||
|
BigFan
TLADT24920 Posts
As for 12 nex, I also wonder as well. | ||
|
Aminus
Bulgaria35 Posts
On October 16 2018 23:37 Freakling wrote: The root of the problem is that Scarabs are units that can only travel over passable ground. So to ensure that things like cliffs and other major terrain obstacles do not cause Reavers to shoot out Scarabs that will just run around erratically and then dud out (because there is no or only a very indirect path between the Reaver and the target) the Reaver attack is coded to use pathfinding distance instead of simple radial distance in its attack range check. The algorithm that determines ground traveling distance is just a very rough heuristic, though. Across multiple regions it seems to use pathfinding node-to-node distance instead of actual unit positions or linear distances, so if the nodes of semi-adjacent regions (regions in close proximity, but separated from each other by at least one interjacent region/node) are spaced out too far the engine considers that as a long travel distance. This is how far I understand it so far. For the exact algorithmic details I guess I'll have to delve into the OpenBW code at some point :\ Example for the instance that caused the bug in this particular game (I think what Bonyth is referring to is the scene at about 0:51:30, at least that's what caught my attention): + Show Spoiler [show image] + I marked the approximate unit positions and highlighted the relevant region nodes. Thats a thread that shouldve been discussed way more than it is, not only that but changes shouldve been made long time ago. Ive lost not one game because of the broken ai of the reaver, reaver bugs even on flat ground trying to melee attack... When we are talking about profesional scene and sports, its unacceptable spoiled game design causing bugs to be a win or lose factor in competition, it should be fixed. Its unfair that these flaws harms one race and spare others. No other race is affected more by bugs, than the Protoss does and its not like the Protoss race is the most dominant or has more titles in tournament than zerg or terran, its quite the opposite. Reaver AI needs to be fixed as much as dragoon ai. Ive seen jangbi lose game on stream, simply because his dragoons dont follow his commands and cant get up a ramp, he did his best to save them, but tank fire has no mercy, then he had to gg... Or the bug when you click dragoons to go to certain spot on an flat ground and just as they reach the destination, half of them starts going in every possible direction, one of them get a shot from a tank then all the group of goons going into tank fire. Its a common thing seen hundreds and hundreds of times in all situations and games. And player have to watch out if the goons going to bug or not rather than doing his macro or whatever he needs to do at that time. I dont know if these questions were risen by the korean progamers, but the game from today was serious alarm that something needs to be fixed. | ||
| ||
