|
Units attacking from lower ground to higher ground only have a 70% chance to hit Units that are in trees are considered to be "in cover". Any units that are in cover only have a 70% chance of being hit
Or so says Blizzards official strategy guide. But is it really true? It often feels impossible to attack units on high ground, and -30% doesn't seem like that big a deal right?
So since I recently had courses in statistics I thought it would be a fun idea to test the hypothesis about 70% hit-rate.
Gathering data
I made a map where 10 dragoons can attack a missile turret on high ground. Missile turret gets 1000 hp, no armor, and dragoons get to make 1 damage.
I then shot 10 volleys from the goons synchronized and noted the turrets hp.
This was repeated 30 times (restarting map when needed).
Analysis
The goons fired a total of 3000 shots. 1587 hit the target, giving a hit-ratio of 52.9%
What can we then say about the real value p (=probability to hit)?
+ Show Spoiler [Math] +
Well 1587 is an observation of a random variable X from Bin(3000,p).
Since n is 3000, an approximation with the normal distribution will be very accurate.
So X is~ N(3000p,sqrt(3000*p*(1-p)))
So p* = X/3000 is~ N(p,sqrt(p*(1-p)/3000))
and so on...
We can now contruct a ~99.9%-confidence interval for p:
(0.499, 0.559)
This means that the probability that p=(probability for an attack to hit from low to high ground) is somewhere between 0.499 and 0.559 is 99.9%.
Conclusion
It's definatly not a 70% hit rate, at least not on my computer.
It might also be something wrong with starcrafts randomizer in my computer or something 
I also tried the same thing with a tank on high ground and a tank behind a tree. I didn't do 3000 shots, but both cases gave hit-rates around 50% just like with the turret.
I'll upload the map so you can try it yourselves.
>Map from rapidshare
Thanks for reading!
Update
Since many of you thought it might be different for other units I tried the same thing with marines and hydras but with smaller sample sizes, only 500 shots each.
Out of 1000 shots 522 hit giving a hit-ratio of 52.2%
So there's no significant difference between units which is what you'd expect from the programming point of view.
|
I always thought something wasn't quite right with the '70% hit' stat, thanks for confirming. :D
|
I thought it was 70% chance you miss. I've seen some recent pro games where a goon is shooting a mine on a ramp and it takes roughly 8-12 shots just to hit it once (obviously just a few games).
|
Ditto, it always seemed to be much less then 70%. 70% would at least mean if you fire, you have a good chance to hit the unit; I've always seen it around 50% as well.
|
I always thought that the unit shooting from low ground has 33% chance of hitting the one higher o.O
and never knew about the trees thing, heh.
|
The 70% is definitely not an accurate statistic. Sometimes units rarely miss high ground targets, sometimes they miss almost all their shots. Almost seems like Blizz left that to a random number generator or something to that extent.
|
|
|
Nice OP. It does feel like less than 70% when playing at times. Blizzard, fix the high ground advantage or change the statistic!
edit- I'm thinking maybe if vision plays some factor, too. Like when you lose vision of the high ground and you can't see your goon shot hit the tank because it disappeared. For example, when a nuke is being dropped, they scan the area so they have vision. Does vision play any role in getting an effective hit?
|
It probably would have worked better if you used mutas.
|
Highground factors in for everyone except Reach
|
I've always felt like that number was a bit high, and it's nice to see someone prove just how ridiculously important high ground is. Chiming in to say great work, OP!
|
On February 12 2009 08:49 arb wrote:
Highground factors in for everyone except Reach
Haha~~
+ Show Spoiler +That game was the reason I did this
|
On February 12 2009 08:35 ZoW wrote:
The 70% is definitely not an accurate statistic. Sometimes units rarely miss high ground targets, sometimes the miss almost all their shots. Almost seems like Blizz left that to a random number generator or something to that extent.
uh. how else would you do this apart from using a form of RNG?
Interesting OP btw, the sample size and math seem to check out.. looks pretty good that it's not 70%. Maybe it was 70% at one time and changed in a later patch with the guides not updated? i find it hard to believe that blizzard either randomly pulled out a number or just decided to misinform people for fun.
|
Nimue would be proud to have you as a student :p
|
On February 12 2009 08:52 JeeJee wrote:Show nested quote +On February 12 2009 08:35 ZoW wrote:
The 70% is definitely not an accurate statistic. Sometimes units rarely miss high ground targets, sometimes the miss almost all their shots. Almost seems like Blizz left that to a random number generator or something to that extent. uh. how else would you do this apart from using a form of RNG? Interesting OP btw, the sample size and math seem to check out.. looks pretty good that it's not 70%. Maybe it was 70% at one time and changed in a later patch with the guides not updated? i find it hard to believe that blizzard either randomly pulled out a number or just decided to misinform people for fun.
I'll check this in the original, non-updated starcraft.
Apparently the 1.00 exe runs at 1.15.3 for me =/ 1.09b has 50% hitrate as well, I don't have any older ones here.
|
On February 12 2009 08:52 jtan wrote:Show nested quote +On February 12 2009 08:49 arb wrote:
Highground factors in for everyone except Reach Haha~~ + Show Spoiler +That game was the reason I did this
what game? Link please
|
On February 12 2009 08:59 LemOn wrote:Show nested quote +On February 12 2009 08:52 jtan wrote:On February 12 2009 08:49 arb wrote:
Highground factors in for everyone except Reach Haha~~ + Show Spoiler +That game was the reason I did this what game? Link please
+ Show Spoiler +http://www.youtube.com/watch?v=oqSwXOdv2uE&eurl=http://www.teamliquid.net/forum/viewmessage.php?topic_id=50875¤tpage=247
Reach vs Young from today, nothing special but at one point reach attacks uphill and kicks ass
|
+ Show Spoiler +I was amazed too but those storms really dealt a lot of damage.. Plus I think young sort of took the high ground safety for granted
|
Netherlands13554 Posts
I always thought it was 50%. Never knew about the 70% statistic but I know that can't be right.
|
Nice post! Very cool.
Blizzard starcraft guides also mention wraith rushing in TvP is good because you can harass their probe lines. lul.
|
did anyone else not know that on the new 'broken' ramps, the one that go the wrong directions and stuff, theres patches of ground that count as high ground, so they dont get the % disadvantage?
never knew that till a couple of weeks ago.
|
Hong Kong20321 Posts
there's patches that count as high so they don't get the disadvantage...? o_ O
[edit]
oh right i get what u mean now
|
United States12607 Posts
Notice: this data put the hit rate at slightly above 50%. The OP's observed mean is ~.53, based on the CI he gives.
53% sounds like a pretty arbitrary number for Blizzard to set as the hitrate when shooting uphill...but your sample size is massive enough that I doubt the "true" hitrate is much different (to quantify that, your 99.9% CI only gives a range of only .03 on either side of .53. Based on this, we can say conclusively that the hitrate is not 50%).
Nice report btw, this is exactly the kind of thing I love :D
|
Having played a billion 2v2's involving breaking someone's ramp with MM, I can also agree that it is definitely nowhere neaaaaaaaarrr 70%, sometimes it even feels like 20% when 8 vults miss a zealot 3 times each and a cannon finishes because of it
|
On February 12 2009 09:26 jwd241224 wrote:
Doesn't this data put the hit rate at slightly above 50%? Your mean must be ~.53, based on the CI you give.
53% sounds like a pretty arbitrary number for Blizzard to set as the hitrate when shooting uphill...but your sample size is massive enough that I doubt the "true" hitrate is much different.
Nice report btw, this is exactly the kind of thing I love :D
Good point, my data suggests that it is also very unlikely that the number is 50%, but like you say it would be weird to set it to 53% by blizzard. A few things could have gone wrong though. Perhaps some goons close to the edge glitch and think they are on high ground, or perhaps the random number generator has some sort of minor bias.
edit:also my observed mean was exactly 1587/3000=0.529
|
I actually think it might vary from unit to unit. I once heard that Tanks and marines are best at shooting uphill, while dragoons are worst (could have been a misunderstanding, maybe it was referring to the Tanks' splash or smth)
Anyway, great post and thanks alot! Starcraft never ceases to amaze
|
On February 12 2009 09:22 IdrA wrote:
did anyone else not know that on the new 'broken' ramps, the one that go the wrong directions and stuff, theres patches of ground that count as high ground, so they dont get the % disadvantage?
never knew that till a couple of weeks ago.
I think anyone who created maps with inverse ramps knows this, or at least they should have guessed it. Yeah it sucks, but I don't think it fixable
|
Why don't you try this with other ranged units too?
Tanks
Rines
Hydras
etc.
|
On February 12 2009 09:33 minus_human wrote:
I actually think it might vary from unit to unit. I once heard that Tanks and marines are best at shooting uphill, while dragoons are worst (could have been a misunderstanding, maybe it was referring to the Tanks' splash or smth)
Anyway, great post and thanks alot! Starcraft never ceases to amaze
PLay some TvT and tell me if tanks are better at shooting uphill than other units...
Maybe it only seems that marines can shoot better uphill because a large number, stimmed, appears to connect more often. This is only an illusion in my experience.
On February 12 2009 09:22 IdrA wrote:
did anyone else not know that on the new 'broken' ramps, the one that go the wrong directions and stuff, theres patches of ground that count as high ground, so they dont get the % disadvantage?
never knew that till a couple of weeks ago.
Could you give an example of this? Are you talking about ramps like at 11 on Othello?
|
maybe you should test your theory with other ranged attack units as well. Maybe the 70% Blizzard stated was through all gound ranged units and different units have different percentages.
The more likely reason though that it isn't 70% is that there has been many patches which messed with the hit ratio since they published there strategy guild.
|
I wasn't too sure anyway, it's more like I was seeking confirmation
But you make a valid point NoobsOfWrath, and come to think about it, I think I actually did read what I wrote above in a Korean interview or smth, but I never realized that they could have been taking about the fact that in large numbers, and with very fast shooting speed, marines are best at shooting uphill.
There shouldn't logically be any exception from the rule, silly me I guess
|
On February 12 2009 09:52 NoobsOfWrath wrote:Show nested quote +On February 12 2009 09:33 minus_human wrote:
I actually think it might vary from unit to unit. I once heard that Tanks and marines are best at shooting uphill, while dragoons are worst (could have been a misunderstanding, maybe it was referring to the Tanks' splash or smth)
Anyway, great post and thanks alot! Starcraft never ceases to amaze PLay some TvT and tell me if tanks are better at shooting uphill than other units... Maybe it only seems that marines can shoot better uphill because a large number, stimmed, appears to connect more often. This is only an illusion in my experience. Show nested quote +On February 12 2009 09:22 IdrA wrote:
did anyone else not know that on the new 'broken' ramps, the one that go the wrong directions and stuff, theres patches of ground that count as high ground, so they dont get the % disadvantage?
never knew that till a couple of weeks ago. Could you give an example of this? Are you talking about ramps like at 11 on Othello?
all ramps not allowed in the original editor, afaik
i found out because of the 6 ramp on destination
2 unranged goons broke 4 rines on a ramp without losing one and i was like o.o?
|
That happens because the ramp is created by manually copy-pasting little patches of tileset to make it
a)walkable and
b) visually appealing
In the process, mappers use patches from high ground and/or low ground, whichever looks best. The only thing they try to respect is to make it as large as a normal ramp, and often a ramp like this is made out of patches of ground that would normally be 'low ground', so it doesn't give a high ground advantage.
|
Yeah, it does seem more like 50% and not to mention that you need to get vision too. High ground is indeed quite an advantage.
|
On February 12 2009 08:49 arb wrote:
Highground factors in for everyone except Reach
|
I always assumed it was 70% hit rate, as that was what I'd read...
But after reading this, yeah, 50% sounds a lot more right. I dunno, I never really thought that deeply into it.
Thanks to the OP for this, though. Very interesting.
|
I always thought it was 50%, and since that's pretty much what happened it made sense
|
United States12235 Posts
Nice work jtan. I always spouted the 70% hit statistic as gospel without ever checking to back it up.
Possible reasons which would fit within or near the margin of error:
- Double counting (70% hit rate counted twice or 49%, or possibly the reverse of 51%)
- Double counting out of 256 (179/256 = 70%, 179/256 * 179/256 = 48.9%, or possibly 51.1%, everything else in the game is /256 so this may be no different)
- 70/128 instead of 70/100 (54.6%, or if out of /256 it could be double counting additively)
More uncommon possibilities:
- 100-30/70 (57.2%)
- 70/130 (70/(100+100-70) = 53.8%, close result, but the reason for the numbers makes no functional sense)
- 69/130 (53.0%, this factors in a possible 1% chance a ranged shot will miss as we sometimes observe, but instead of 1/100 it is a 1/130 less chance)
I contend that it's probably a bug, but the reason for the bug is unclear since we have no way of viewing the code.
|
On February 12 2009 09:22 IdrA wrote:
did anyone else not know that on the new 'broken' ramps, the one that go the wrong directions and stuff, theres patches of ground that count as high ground, so they dont get the % disadvantage?
never knew that till a couple of weeks ago.
yes and some maps like python 6 have 1 tile that is low ground on the top of the ramp so ur unit may not even get vision (like he's been optical flared).
And this thread is very interesting. After 10 years no one actually bothered to check the stat that we all know by heart and even realize the extreme value of high ground.
Now the question is what should be fixed, the stat on their website/game manuals etc. or patch it to actually be 70% .
The game would seem to change so drastically.
it's like playing holdem and being put all in with Top pair/flush draw normally then all the sudden someone hands you trips.
PS- This needs to be tested a little more imo, more shots and more units. Also do buildings have a different hit rate than a unit? Do different buildings have different hit rates (T/P/Z etc)? Do large/small/medium units effect anything?
|
The sad thing is that I can see some Korean site having this investigation years ago.
;_; Foreign scene compared to Korea
|
lol -_-x no wonder when i TvP its impossible to break a natural
|
Nice OP, always thought that the 70% number was a little high.
|
On February 12 2009 12:06 Ellis wrote:
The sad thing is that I can see some Korean site having this investigation years ago.
;_; Foreign scene compared to Korea
lol what a retarded statement
seriously....stupid and ignorant
|
On February 12 2009 10:36 Excalibur_Z wrote:
Nice work jtan. I always spouted the 70% hit statistic as gospel without ever checking to back it up.
Possible reasons which would fit within or near the margin of error:
- Double counting (70% hit rate counted twice or 49%, or possibly the reverse of 51%)
- Double counting out of 256 (179/256 = 70%, 179/256 * 179/256 = 48.9%, or possibly 51.1%, everything else in the game is /256 so this may be no different)
- 70/128 instead of 70/100 (54.6%, or if out of /256 it could be double counting additively)
More uncommon possibilities:
- 100-30/70 (57.2%)
- 70/130 (70/(100+100-70) = 53.8%, close result, but the reason for the numbers makes no functional sense)
- 69/130 (53.0%, this factors in a possible 1% chance a ranged shot will miss as we sometimes observe, but instead of 1/100 it is a 1/130 less chance)
I contend that it's probably a bug, but the reason for the bug is unclear since we have no way of viewing the code.
I actually think it'd be a fairly interesting endeavor to find the code that handles these random numbers. Where to begin in reverse engineering that, however, I have no idea. I'll think on it for a while, maybe I can figure something out
|
I would really like to see this tested with other range units - maybe it's subjective, but I always feel that Tanks do not suffer all that much when firing uphill when compared to Dragoons - at least, it always feels nearly impossible to assault a Terran sieged on high ground.
|
On February 12 2009 12:38 Funnytoss wrote:
I would really like to see this tested with other range units - maybe it's subjective, but I always feel that Tanks do not suffer all that much when firing uphill when compared to Dragoons - at least, it always feels nearly impossible to assault a Terran sieged on high ground.
Tanks probably fire a lot better uphill because of splash damage?
|
As I understand it, the way replays can accurately recreate games where 'random' events like shooting uphill take place is by storing a seed number that is created (randomly, I assume) when the game is made. There may be some variance from game to game in hit rates, averaging ~50%.
|
Bill307
Canada9103 Posts
On February 12 2009 12:38 Funnytoss wrote:
I would really like to see this tested with other range units - maybe it's subjective, but I always feel that Tanks do not suffer all that much when firing uphill when compared to Dragoons - at least, it always feels nearly impossible to assault a Terran sieged on high ground.
Tanks deal splash damage with a radius big enough to hit almost any unit for 25% or 50% damage, even if the tank blast misses. This is the same for units on high ground, behind trees, and under dark swarm.
I think burrowed units are the only units with small enough hitboxes that the tank splash misses them completely.
jtan: huge thanks for finally testing this statistic.  I always thought it was 70%, but that % always felt too high.
|
always thought it was about 50% and i think most others accepted that as well
|
Thanks for the positive feedback from all of you!
Some of you requested stats from other units, so I updated the OP with stats for marines and hydras. It was a lot harder to get the number of shots right so I used smaller samples. I won't do tanks because of the splash damage.
|
Ah, so splash damage is what really negates the high ground. (well, kind of) This is quite an interesting thing to finally find out, great work!
|
50% does seem more accurate. What about when like a zealot attacks another zealot? is that 50^% too?
|
On February 12 2009 20:47 Racenilatr wrote:
50% does seem more accurate. What about when like a zealot attacks another zealot? is that 50^% too?
no no, it's only for ranged attacks
|
Burrowed units don't take splash damage (it's not a hitbox thing).
That's why swarm + lurker works..
|
|
|
is awesome32274 Posts
On February 12 2009 10:12 minus_human wrote:
That happens because the ramp is created by manually copy-pasting little patches of tileset to make it
a)walkable and
b) visually appealing
In the process, mappers use patches from high ground and/or low ground, whichever looks best. The only thing they try to respect is to make it as large as a normal ramp, and often a ramp like this is made out of patches of ground that would normally be 'low ground', so it doesn't give a high ground advantage.
I believe this is the same reason why when you block one of these ramps with a marine, scv, probe, etc sometimes you lose complete vision around it. It's like your unit is on a low ground patch surrounded by high ground patches, so it can't see anything.
|
On February 12 2009 23:30 IntoTheWow wrote:Show nested quote +On February 12 2009 10:12 minus_human wrote:
That happens because the ramp is created by manually copy-pasting little patches of tileset to make it
a)walkable and
b) visually appealing
In the process, mappers use patches from high ground and/or low ground, whichever looks best. The only thing they try to respect is to make it as large as a normal ramp, and often a ramp like this is made out of patches of ground that would normally be 'low ground', so it doesn't give a high ground advantage. I believe this is the same reason why when you block one of these ramps with a marine, scv, probe, etc sometimes you lose complete vision around it. It's like your unit is on a low ground patch surrounded by high ground patches, so it can't see anything.
Precisely. That thing never happens with normal ramps.
|
On February 12 2009 23:23 errol1001 wrote:
Burrowed units don't take splash damage (it's not a hitbox thing).
That's why swarm + lurker works..
Are you sure about the burrowed and splash?
I'm sure I played some UMS where the reaver was killing 2-3 burrowed units with one hit.
|
demon forest: now even shittier than it already was.
|
On February 12 2009 23:49 arbiter_md wrote:Show nested quote +On February 12 2009 23:23 errol1001 wrote:
Burrowed units don't take splash damage (it's not a hitbox thing).
That's why swarm + lurker works.. Are you sure about the burrowed and splash? I'm sure I played some UMS where the reaver was killing 2-3 burrowed units with one hit.
I think the reaver has melee damage though so swarm doesn't affect it.
But you can still kill burrowed units under swarm with tanks by focusing fire on another unit a bit over the first one.
|
On February 12 2009 23:23 errol1001 wrote:
Burrowed units don't take splash damage (it's not a hitbox thing).
That's why swarm + lurker works..
On February 12 2009 23:49 arbiter_md wrote:Show nested quote +On February 12 2009 23:23 errol1001 wrote:
Burrowed units don't take splash damage (it's not a hitbox thing).
That's why swarm + lurker works.. Are you sure about the burrowed and splash? I'm sure I played some UMS where the reaver was killing 2-3 burrowed units with one hit.
Of course burrowed units take splash.
Thats why siege tank & SCV or Archon & zealot against lurks works.
They just take no splash damage from tanks under swarm.
Does anyone know, do burrowed units under swarm take damage from Archons or reavers? I'm guessing archons no, reavers yes but i cant check right now...
|
70% chance of hit isnt like, it will hit in 7 out of 10 shots
for each shot 70%
so it means that it may never land a hit, or land all hits in 100 shots, it is lucky based
|
No, it's probability based and the probability that a large simple will differ significantly from the probability is tiny.
|
yeah if you just look at the numbers in the first post you'll realize the odds of it actually being 70% is vanishingly small (actually, impossible because games use pseudo-random number generators, not actually random numbers)
|
On February 13 2009 00:13 stet_tcl wrote:Show nested quote +On February 12 2009 23:23 errol1001 wrote:
Burrowed units don't take splash damage (it's not a hitbox thing).
That's why swarm + lurker works.. Show nested quote +On February 12 2009 23:49 arbiter_md wrote:On February 12 2009 23:23 errol1001 wrote:
Burrowed units don't take splash damage (it's not a hitbox thing).
That's why swarm + lurker works.. Are you sure about the burrowed and splash? I'm sure I played some UMS where the reaver was killing 2-3 burrowed units with one hit. Of course burrowed units take splash. Thats why siege tank & SCV or Archon & zealot against lurks works. They just take no splash damage from tanks under swarm. Does anyone know, do burrowed units under swarm take damage from Archons or reavers? I'm guessing archons no, reavers yes but i cant check right now...
In regard to your question, I believe you're guessing right.
|
Ya this has always bothered me, Im glad someone finally did the math on this topic.
I always thought it was a 66.6% (1/3 miss) rather than 70 but even then when I would attack a ramp and notice every now and then my goons miss 3 shots in a row before making a hit made me wonder, was it really 2/3 hit rate?
Nice writeup I BELIEVE YOU!...
Maybe it has to do with the patches and how the game has changed over time, either way I still think its fair the way it is... usually when u need to use your highground to your advantage you really need it so... more power to the 52% IMO.
maybe we can try to figure out why when you have a ton of units and they hit 1 building/unit and it obviously would of died within the first few hit, yet each unit gets their punches in and maybe twice... anyone else ever notice how it seems that units "miss" on lowground every now and then?
EDIT: This usually happens when there is a lot of units attacking 1 or very few units, without ground height advantage/disadvantage.
anywho nice write up. Hope to see some more math related topics from you.
|
This was a very good test.
But I don't think it's complete unless some other units are tested (how about tanks without siege?) and/or different targets are tested. Perhaps the results will be different if you attack a unit and not a building.
|
Good work. I also didn't believe the 70% lol.
|
Have you tried using the lower high ground or raised or whatever it's called?
|
United States12607 Posts
On February 13 2009 00:23 Kim_Hyun_Han wrote:
70% chance of hit isnt like, it will hit in 7 out of 10 shots
for each shot 70%
so it means that it may never land a hit, or land all hits in 100 shots, it is lucky based
Sorry, but this.... holy shit...
|
On February 13 2009 02:01 jwd241224 wrote:Show nested quote +On February 13 2009 00:23 Kim_Hyun_Han wrote:
70% chance of hit isnt like, it will hit in 7 out of 10 shots
for each shot 70%
so it means that it may never land a hit, or land all hits in 100 shots, it is lucky based Sorry, but this.... holy shit...
Yea, this is like a request for an algebra 1 textbook in my eyes.
|
On February 13 2009 00:13 stet_tcl wrote:
They just take no splash damage from tanks under swarm.
Thats wrong, tank splash damage under swarm is a bit out of place, if you manually target the lurker behind(in direction from your tank) others, the front one will get damaged by the splash.
For example: X² X³ T
You shoot with Tank T at Lurker X² and X³ will take the damage from splash. (under swarm, and of course close to each other)
|
On February 13 2009 01:27 Scamp wrote:
This was a very good test.
But I don't think it's complete unless some other units are tested (how about tanks without siege?) and/or different targets are tested. Perhaps the results will be different if you attack a unit and not a building.
As I say in the op, I already tried 3 kinds of units, and I tried attacking a tank instead of a turret but with the same results.
|
On February 12 2009 15:43 Bill307 wrote:Show nested quote +On February 12 2009 12:38 Funnytoss wrote:
I would really like to see this tested with other range units - maybe it's subjective, but I always feel that Tanks do not suffer all that much when firing uphill when compared to Dragoons - at least, it always feels nearly impossible to assault a Terran sieged on high ground. Tanks deal splash damage with a radius big enough to hit almost any unit for 25% or 50% damage, even if the tank blast misses. This is the same for units on high ground, behind trees, and under dark swarm. I think burrowed units are the only units with small enough hitboxes that the tank splash misses them completely. jtan: huge thanks for finally testing this statistic. I always thought it was 70%, but that % always felt too high.
It's not hitboxes, they are the same, it's just that burrowed units are completely immune to 50% and 25% splash(but not 100%), even if it's coming from a nuke.
It's not true that splash doesn't work against swarm, that's because of splash immunity from burrow, while swarm moves the damage area a bit towards the attacker, so combined with burrow the lurkers become invulnerable to direct tank and archon attacks, but you can still damage a lurker, if you target something behind him, so the 100% splash zone appears on top of him. Lurkers and Firebats fire directly at the target, so they can't miss and Reaver Scarabs are melee kamikaze units, so they attack normally under swarm(and their 100% splash zone is pretty big).
I guess I'll just c/p a post of mine concerning the most commonly asked questions about game mechanics:
+ Show Spoiler +
Here's how the damage calculation is done:
1. Reduction from splash zone(damage is multiplied by 0.25/0.5/1.0)
2. Armor amount reduction
3. Armor type reduction
If it's a protoss unit and:
a) Currently has no shields, just the armor type and amount of the unit are used.
b) It has shields, the shield upgrade amount is used and the armor type takes 100% from everything. If this attack depletes the shields, the leftover damage after depleting the shields is calculated again as in case a)
The minimum damage from an attack is 0.5 and this does not apply twice for the separate shield/hp calculations for protoss units. Damage amount is rounded to 1/256.
Units with special attacks:
Zealot - 2 attacks by 8 + 1 per upgrade
Scout Air - 2 attacks by 14 + 1 per upgrade
Goliath Air - 2 attacks by 10 + 2 per upgrade
Firebat - 3 attacks by 8 + 1 per upgrade, even burrowed lurkers can be hit by all 3 attacks, it just depends on the angle of attack(the amount shoed is hardcoded to be the damage of 2 of the attacks added up, but all three are upgraded and each deals 8 damage base)
Lurker - several attacks by 20 + 2 per upgrade, with the restriction that only 1 attack can hit a single unit and all of them have only 100% splash(the 50% and 25% areas are the same as the 100%, so a unit can't take partial damage from the lurker's attack). A bug can override this rule, this happens when the lurker dies the exact moment it attacks, then a unit can be hit by several of the attacks, it's easily noticable with full hp marines dying to a single lurker.
Valk - 8 attacks by 6 + 1 per upgrade, if there are multiple units in the 100% splash zone, only one will take 100% damage and the others will take 50%. The same aplies to corsairs.
Mutalisk bounces deal 1/3 of the previous bounce, for example +1 damage muta against 3 targets with 1 armor will deal 10 - 1 = 9 on the first hit, 10 / 3 - 1 = 2.333...(rounded to the nearest 1/256) on the second hit and 10 / 9 - 1 = 0.111... upped to 0.5, which is the minimum damage, on the third hit.
Devourer spores increase damage from each attack by 1 per spore, this includes spell damage.
Not an unit attack, but since it affects attacks: Defensive matrix reduces attack damage to the minimum - 0.5, this includes spell damage. The total damage prevented is 250, but this does not acount the minimum damage inflicted or the unit's armor, so a zergling with 5 damage will perish the shield in 50 hits and still deal 25 damage, no matter the target's armor and the end effect will be 225 damage absorbed on a unit with 0 armor and even less on a unit with several points of armor(if the target has 5 armor, for example, having Defensive Matrix will not make any difference).
An exception to these 2 is the plague, which doesn't deal damage, but directly removes hp, so it doesn't interact with defensive matrix or acid spores.
Nukes deal:
Against units/buildings with 750 or less total maximum HP + shields -> 500, 250, 125 explosive damage.
Against units/buildings with 750 or more total maximum HP + shields ->
2/3, 1/3, 1/6 of the total maximum HP + shields.
The three values are for the 100%, 50% and 25% splash zones ofc.
|
What if you have cover under a tree pluss you're on higher ground? Do the miss chances stack?
|
Sounds very rigorous. Nice work!
|
|
|
interesting
i wonder if it was 70% when the strategy guide was printed? quick, someone repeat the test on version 1.1! the game has gone through many changes, obviously, and the guide hasn't exactly kept up - it still tells you to rush to scouts iirc.
|
Just one comment about this
On February 12 2009 08:30 jtan wrote:It might also be something wrong with starcrafts randomizer in my computer or something
It's definitely not something only on your computer. Randomization in Starcraft (or actually in computer systems in general) is deterministic - means it's random in a pre-determined way.
In a multiplayer game, a neutral critter, though moving around randomly, will move the exact same way on all players' computers (there's no communication exchanged among the computers about movement of neutral critters). Similarly, there's no communication needed among players' computers about whether the shot from the dragoon has missed the target - it's all determined "randomly" the same way.
So yeah, I'm sure it's not something wrong in your computer
|
perhaps when they tested their RNG they came up with a result of about 70%? Who knows, when the hits are completely random like that, you have to do thousands of trials to nail down a good average. This is why they have super computers.
|
On April 07 2009 20:02 ghermination wrote:
perhaps when they tested their RNG they came up with a result of about 70%? Who knows, when the hits are completely random like that, you have to do thousands of trials to nail down a good average. This is why they have super computers.
Lol no, they can just run to the next room to ask the game coders.
|
United States41 Posts
|
I recently came back to Starcraft to play with some people from my Diablo 2 guild (The Amazon Basin) and we had a little confusion about the chance to miss in general, so I set up some tests for it. My results agree with the OP but using Siege-mode Tanks instead of Dragoons: 3819 misses and 4400 hits when attacking a high ground target for a ~53.5% chance to hit. This test used two spider mines separated by one grid square with the tank targeting the far mine, so that a hit would damage the far mine for 1 point and a miss would damage the near mine for 1 point.
Some additional info: for targets without cover, Siege-mode Tanks had a ~0.337% chance to miss (31 misses, 9206 hits), Marines had a ~3.5% chance to miss (384272 hits out of ~398000 shots), all ranged units I looked at had a visible chance to miss attacks (Tank, Marine, Mutalisk, Vulture, Dragoon), and melee units seem to have a random chance to pause between attack animations similar to a miss chance.
|
On April 08 2009 06:19 SirNukes wrote:
Some additional info: for targets without cover, Siege-mode Tanks had a ~0.337% chance to miss (31 misses, 9206 hits), Marines had a ~3.5% chance to miss (384272 hits out of ~398000 shots), all ranged units I looked at had a visible chance to miss attacks (Tank, Marine, Mutalisk, Vulture, Dragoon), and melee units seem to have a random chance to pause between attack animations similar to a miss chance.
This is also an interesting trivia - most units have random pauses when they attack. The reason is if, say a group of marines, attack a building with the same rate (without random pauses), there will be a shooting pattern that make it look and sound really unnatural. I think I read it somewhere that the Blizzard programmers realized that from Warcraft 2, so they decided to add the random pauses in.
|
On April 08 2009 08:35 mrdx wrote:
This is also an interesting trivia - most units have random pauses when they attack. The reason is if, say a group of marines, attack a building with the same rate (without random pauses), there will be a shooting pattern that make it look and sound really unnatural.
Ya, I noticed this as well. However, rather than random pausing, it seems to be a slight cooldown offset or somesuch that causes units to spread out their attacks. I call it an offset because it did not affect attack rate over time; at the end of one of my siege tank tests with 20 separate tanks I checked how many attacks each one did: 17 tanks had taken 462 shots while 3 had taken 461 shots. However, the melee unit pausing I mentioned did cause attack differences over time: a pack of 20 Ultralisks varied by about 5 attacks out of 100 and 20 Dark Templar varied by about 5 attacks out of 1000.
|
|
|
|
|
|
EPT![[image loading]](http://img18.imageshack.us/img18/2729/10goonsfm4.jpg)
