Hi all! For some days I test various terran's units ingame for their DPS and... I found that the attack rate of Battlecruiser and Reaper are lower than that in their tooltips. Tests were done in "Unit test map" at "Faster" game speed.
At first, I will explain the method of calculation (sample test): target - full HP and shields Nexus without any upgrades for armor or shields. attacker - Marauder without any upgrades for attack or stimpack. Base damage = 20, attack cooldown = 1.5 seconds.
Nexus doesn't have any shield armor, so Marauder will hit for full 20 damage every 1.5 seconds until shields ends (from 1000 to 0). The calculations are 1000/20*1,5=50*1,5=75 seconds.
As you can see from screenshot 2 it takes exactly 75 seconds (as calculated before). So, go next!
Battlecruiser test: #1 - Air-to-ground attack: Base damage is 8 to ground, and tooltip says attack rate is 0.23 sec. But on teamliquid wiki there is more detailed information (http://wiki.teamliquid.net/starcraft2/Battlecruiser). It says 0.225 sec (I check it with SC2Editor - 0.225).
target - Nexus with +3 graded shields. attacker - Battlecruiser w/o grades for attack. The point of calculation - the time needed to destroy 1000 shields.
The test was done in 58 seconds, but BK hit one extra time. There are 1000/(8-3)=200 hits needed to remove shields, and 1 extra attack was done, so 201 attack in 58 seconds. Time between attacks - 58/201=0,288557 seconds. That's far from stated 0.225 seconds. The difference is felt. DPS of Battlecruiser is 8/0.225=35.55 (rounded up to 35.6 on teamliquid wiki). Real DPS of Battlecruiser is 8/0,288557=27.72. That's 22% lower. DPS per limit is 27.72/6=4,62... so low. Stimpacked marine deals 6*1,5/0,86=10,46 dps!
#2 - Air-to-air attack: target - Battlecruiser with +2 armor grades, it totaly has 2+3=5 armor. attacker - Battlecruiser without attack grades, damage =6 and the attack rate is the same = 0.225. The purpose of the test - damage BCruiser up to 50 HP.
The time needed - 144 seconds. Number of attacks done = (550-51)/(6-5)=499/1=499. Time between attacks = 144/499=0,288557! It's equal to previous test's result! DPS (for 0.225 attack rate) - 6/0,225=26,66 Real DPS - 6 /0,288557 =20,7931 20,7931/26,66=0,7799 - 22% lower dps, again.
Reaper test: Attack to structures (with bombs): target - Nexus w/o any grades. attacker - Reaper, 30 damage to structures, 1.8 seconds attack cooldown from game tooltip. The purpose of the test - damage shields for 990 points. 990/30=33 attacks done.
It takes 69 seconds. Time between attacks - 69/33=2,09. One more time it's not as in tooltips! DPS (for 1.8 attack rate) - 30/1,8=16,66 Real DPS - 30/2,09=14,35. 14,35/16,66=0,8613. ~14% lower from stated amount.
Some words after The difference between reference and test data can not be explained by random +/-1 second or +/-1 attack. This affect results a little! Tests of other terran units grants perfectly matching values or very close ones to the stated data.
On July 17 2012 15:13 JustPassingBy wrote: Well, the bomb is a special attack, so it does not fall under the rule of the regular attack speed. no clue with the bc.
Wow, if these numbers are accurate that's insane. 22% less DPS than stated? That's a huge disparity. Hopefully we can have some other people check on the numbers/method and see if this is valid.
So you discovered BC is even more shit than it's supposed to be?
But joking aside, is bc "balanced" around the stats, or actual game play dps? Because if those calculations are correct, -22% is a huge difference.
edit: i mean lets say those calculations are correct. How would Blizzard correct the problem. Increase the real attack speed to match the stats, or change the stated attack speed of the unit.
Interestingly, the random attack delay maximum amount is higher for these weapons than most weapons. Not sure if this is to blame, but it certainly could be. Most weapons have a random delay maximum of 0.125, while BCs have 0.1875.
Reaper's anti-structure, however, is more interesting. It has a random delay maximum of 0.5 (very big) but also a random delay minumum of 0.1, a positive number. Most weapons, including that of BCs, has a minimum delay of -0.0625.
You could add anywhere from -0.0625 to 0.125 to the time between attacks for a regular weapon. A BC should average to be a higher because it's between -0.0625 and 0.1875. The reapers, though, don't even have a chance to fire as fast as they should because you will always add at least 0.1, with as much as 0.5.
That is insteresting! Have you tried on other units than the Nexus for the reaper and the Ground to Ground attack for the battlecruiser? Maybe the Nexus has a hidden armor value, or the upgrades grants more than +3 in a hidden way, it would be very weird though Or have maybe the Maurauder do a dps test against a +3 shield Nexus, because for the maurauder you did it without shield upgrades, but for the others, you did
O_O I have to assume Blizzard knew about this but failed to update the tooltips.... but that seems terrible... even worse if they had no idea though D:
This is a very interesting find.. I hope someone can use the map editor to modify the BC to see what it should function as. To be honest, the unit needs a buff. This is even coming from a zerg. x.x
you're completely disregarding the random attack delay. BC and reaper D8 are the only 2 attacks that have a different random attack delay from the default. real attack speed of a BC is 0.1625 to 0.4125 which is 0.2875 average probably same with d8.
oh nvm i see you mentioned random delay at the end. the problem here is that the random delay doesnt average 0. also it is pretty small compared to majority of the units' attack speeds. BC is an extreme case because of its small attack speed value and bigger random delay resulting in a big disparity between the base value and the real value. same with D8 because of its special huge random delays
On July 17 2012 15:40 kazie wrote: you're completely disregarding the random attack delay. BC and reaper D8 are the only 2 attacks that have a different random attack delay from the default. real attack speed of a BC is 0.1625 to 0.4125 which is 0.2875 average probably same with d8.
That doesn't change the problem that tooltip states incorrect values.
I hope the unit test map has nexus shield in fight regen disabled then this would be an awesome thread for the bnet forum to get answeres from the staffs
On July 17 2012 15:40 kazie wrote: you're completely disregarding the random attack delay. BC and reaper D8 are the only 2 attacks that have a different random attack delay from the default. real attack speed of a BC is 0.1625 to 0.4125 which is 0.2875 average probably same with d8.
That doesn't change the problem that tooltip states incorrect values.
tooltips are correct in that blizzard chose to state base values, and that random delays are public knowledge. you can disagree with their decision i guess.
only thing i would say needs to be changed is liquipedia. it should at least have an attack speed page explaining the matter. if op went to another wiki like sc2pod, he would have known the cause of this.
On July 17 2012 15:40 kazie wrote: you're completely disregarding the random attack delay. BC and reaper D8 are the only 2 attacks that have a different random attack delay from the default. real attack speed of a BC is 0.1625 to 0.4125 which is 0.2875 average probably same with d8.
That doesn't change the problem that tooltip states incorrect values.
tooltips are correct in that blizzard chose to state base values, and that random delays are public knowledge. you can disagree with their decision i guess.
only thing i would say needs to be changed is liquipedia. it should at least have an attack speed page explaining the matter. if op went to another wiki like sc2pod, he would have known the cause of this.
Random delays doesn't seems to effect other terran units in any meaningful way, as their test results match theoretical values, I see no reason why BC and reapers should be different.
On July 17 2012 15:40 kazie wrote: you're completely disregarding the random attack delay. BC and reaper D8 are the only 2 attacks that have a different random attack delay from the default. real attack speed of a BC is 0.1625 to 0.4125 which is 0.2875 average probably same with d8.
That doesn't change the problem that tooltip states incorrect values.
tooltips are correct in that blizzard chose to state base values, and that random delays are public knowledge. you can disagree with their decision i guess.
only thing i would say needs to be changed is liquipedia. it should at least have an attack speed page explaining the matter. if op went to another wiki like sc2pod, he would have known the cause of this.
Random delays doesn't seems to effect other terran units in any meaningful way, as their test results match theoretical values, I see no reason why BC and reapers should be different.
On July 17 2012 15:40 kazie wrote: you're completely disregarding the random attack delay. BC and reaper D8 are the only 2 attacks that have a different random attack delay from the default. real attack speed of a BC is 0.1625 to 0.4125 which is 0.2875 average probably same with d8.
oh nvm i see you mentioned random delay at the end. the problem here is that the random delay doesnt average 0. also it is pretty small compared to majority of the units' attack speeds. BC is an extreme case because of its small attack speed value and bigger random delay resulting in a big disparity between the base value and the real value. same with D8 because of its special huge random delays
hope they finally get rid of the random delay, it's just one small step towards wc3, where half the game was dealing with random dice rolls... it screws up terran micro so much
On July 17 2012 15:40 kazie wrote: you're completely disregarding the random attack delay. BC and reaper D8 are the only 2 attacks that have a different random attack delay from the default. real attack speed of a BC is 0.1625 to 0.4125 which is 0.2875 average probably same with d8.
That doesn't change the problem that tooltip states incorrect values.
tooltips are correct in that blizzard chose to state base values, and that random delays are public knowledge. you can disagree with their decision i guess.
only thing i would say needs to be changed is liquipedia. it should at least have an attack speed page explaining the matter. if op went to another wiki like sc2pod, he would have known the cause of this.
Random delays doesn't seems to effect other terran units in any meaningful way, as their test results match theoretical values, I see no reason why BC and reapers should be different.
As was previously stated by another poster, most units' attack delays are +X/-X, where both values are the same. That means that the average delay is 0, and over a large number of attacks the units' DPS is unaffected.
BCs' delay is -.0625/+.1865, which means that the average delay (assuming all values are equally possible) is .0625. Given a base attack speed of 0.23, that means the actual average attack speed is 0.2925, a 21.4% increase. That roughly matches the results given by the OP.
On July 17 2012 16:53 aRyuujin wrote: hope they finally get rid of the random delay, it's just one small step towards wc3, where half the game was dealing with random dice rolls... it screws up terran micro so much
ye i agree this is definately a huge factor and looses whole games ... also definately the reason why im not gm yet .... -.-
Wow i didnt even know about random attack delay. I always believed that sc2 was a game purely of numbers and no luck involved if controlled perfektly. I was proven wrong... So if i battle drone vs drone i just have to "hope" that i win. This really sucks. Also this can be the deciding facor in super close marine vs marine or ling vs ling battles. Why is this even in this game?
Great find, it looks pretty solid but I'd rather have someone independently confirm the results (I'll try to do it myself when I find the time): If it is really confirmed we should reach blizzard with this so they can fix the issues (most likely just update the tooltips, but one can always wish for a BC buff xD)
On July 17 2012 17:52 Qgelfich wrote: Wow i didnt even know about random attack delay. I always believed that sc2 was a game purely of numbers and no luck involved if controlled perfektly. I was proven wrong... So if i battle drone vs drone i just have to "hope" that i win. This really sucks. Also this can be the deciding facor in super close marine vs marine or ling vs ling battles. Why is this even in this game?
If I remember correctly, it's so that the game looks "good". I would imagine seeing a group of 20 Marines shooting at the same rate, would look awkawrd. This isn't my personal opinion, but I remember Artosis talking about this in a gsl game a few weeks ago.
Random delay actually explains a lot of things that's been happening to me... I go into probe vs probe battles getting a slightly faster attack off at the beginning but end up losing the fight..... thank you random delay, now I know to NEVER fight probe vs probe despite knowing the opponent's probe had the slightly slower first attack...
On July 17 2012 18:36 eScaper-tsunami wrote: Random delay actually explains a lot of things that's been happening to me... I go into probe vs probe battles getting a slightly faster attack off at the beginning but end up losing the fight..... thank you random delay, now I know to NEVER fight probe vs probe despite knowing the opponent's probe had the slightly slower first attack...
I fail to understand why we need any random factor in a strategic game. Also one more thing, 'Random' is sometimes f*cked up. You may spawn as a race more than 3 times in a row. That's flawed imho although there's really small chance something like this happens. Like... 1/12 to be one race 4 times in a row?
On July 17 2012 18:36 eScaper-tsunami wrote: Random delay actually explains a lot of things that's been happening to me... I go into probe vs probe battles getting a slightly faster attack off at the beginning but end up losing the fight..... thank you random delay, now I know to NEVER fight probe vs probe despite knowing the opponent's probe had the slightly slower first attack...
I fail to understand why we need any random factor in a strategic game. Also one more thing, 'Random' is sometimes f*cked up. You may spawn as a race more than 3 times in a row. That's flawed imho although there's really small chance something like this happens. Like... 1/12 to be one race 4 times in a row?
That is actually exactly what "random" is in reality. If it is not like that, the programmers have rigged it to make it feel better to players.
On July 17 2012 15:07 Dgotto wrote: Hi all! For some days I test various terran's units ingame for their DPS and... I found that the attack rate of Battlecruiser and Reaper are lower than that in their tooltips. Tests were done in "Unit test map" at "Faster" game speed.
At first, I will explain the method of calculation (sample test): target - full HP and shields Nexus without any upgrades for armor or shields. attacker - Marauder without any upgrades for attack or stimpack. Base damage = 20, attack cooldown = 1.5 seconds.
Nexus doesn't have any shield armor, so Marauder will hit for full 20 damage every 1.5 seconds until shields ends (from 1000 to 0). The calculations are 1000/20*1,5=50*1,5=75 seconds.
As you can see from screenshot 2 it takes exactly 75 seconds (as calculated before). So, go next!
Battlecruiser test: #1 - Air-to-ground attack: Base damage is 8 to ground, and tooltip says attack rate is 0.23 sec. But on teamliquid wiki there is more detailed information (http://wiki.teamliquid.net/starcraft2/Battlecruiser). It says 0.225 sec (I check it with SC2Editor - 0.225).
target - Nexus with +3 graded shields. attacker - Battlecruiser w/o grades for attack. The point of calculation - the time needed to destroy 1000 shields.
The test was done in 58 seconds, but BK hit one extra time. There are 1000/(8-3)=200 hits needed to remove shields, and 1 extra attack was done, so 201 attack in 58 seconds. Time between attacks - 58/201=0,288557 seconds. That's far from stated 0.225 seconds. The difference is felt. DPS of Battlecruiser is 8/0.225=35.55 (rounded up to 35.6 on teamliquid wiki). Real DPS of Battlecruiser is 8/0,288557=27.72. That's 22% lower. DPS per limit is 27.72/6=4,62... so low. Stimpacked marine deals 6*1,5/0,86=10,46 dps!
#2 - Air-to-air attack: target - Battlecruiser with +2 armor grades, it totaly has 2+3=5 armor. attacker - Battlecruiser without attack grades, damage =6 and the attack rate is the same = 0.225. The purpose of the test - damage BCruiser up to 50 HP.
The time needed - 144 seconds. Number of attacks done = (550-51)/(6-5)=499/1=499. Time between attacks = 144/499=0,288557! It's equal to previous test's result! DPS (for 0.225 attack rate) - 6/0,225=26,66 Real DPS - 6 /0,288557 =20,7931 20,7931/26,66=0,7799 - 22% lower dps, again.
Reaper test: Attack to structures (with bombs): target - Nexus w/o any grades. attacker - Reaper, 30 damage to structures, 1.8 seconds attack cooldown from game tooltip. The purpose of the test - damage shields for 990 points. 990/30=33 attacks done.
It takes 69 seconds. Time between attacks - 69/33=2,09. One more time it's not as in tooltips! DPS (for 1.8 attack rate) - 30/1,8=16,66 Real DPS - 30/2,09=14,35. 14,35/16,66=0,8613. ~14% lower from stated amount.
Some words after The difference between reference and test data can not be explained by random +/-1 second or +/-1 attack. This affect results a little! Tests of other terran units grants perfectly matching values or very close ones to the stated data.
That's an interesting finding, let's hope it will be backed up by other separate "experiments". To be honest , I always had the feeling that BCs were too weak compared to other related top tier units
That's an interesting finding, let's hope it will be backed up by other separate "experiments". To be honest , I always had the feeling that BCs were too weak compared to other related top tier units
Because the carrier is doing so much better, right?
At least the BC has its niche uses in all 3 matchups, and has gone through various buffs and is getting some more in HOTS.. carrier not so lucky though.
On July 17 2012 18:36 eScaper-tsunami wrote: Random delay actually explains a lot of things that's been happening to me... I go into probe vs probe battles getting a slightly faster attack off at the beginning but end up losing the fight..... thank you random delay, now I know to NEVER fight probe vs probe despite knowing the opponent's probe had the slightly slower first attack...
I fail to understand why we need any random factor in a strategic game. Also one more thing, 'Random' is sometimes f*cked up. You may spawn as a race more than 3 times in a row. That's flawed imho although there's really small chance something like this happens. Like... 1/12 to be one race 4 times in a row?
Otherwise unit will overkill their targets. Imagine 20 siege tanks hitting a single dropped marine wasting 90% of their dps.
So I ran two tests in the unit test map and it confirms what the OP said. (The first test against the Hydra begins at 43 seconds and ends at 45, the second against the Nexus begins at 1:45 and ends at 2:21)
First, the BC against a Hydra (low hp and no armor) with no upgrades yields the following: It takes the BC 11 shots to kill the Hydra and it takes the BC 3 seconds to kill the Hydra (and there is actually one more shot on the way toward the Hydra when it does).
So 3/11 = 0.2727 attack speed
Second, lets look at the BC vs 1000 Nexus shields with no shield upgrades: It takes 125 shots to kill the Nexus shields and it takes the BC 36 seconds to do so (and again there is one more shot on the way to Nexus when 1000 damage is done).
DPS per limit is 27.72/6=4,62... so low. Stimpacked marine deals 6*1,5/0,86=10,46 dps!
So i guess that proves what we've been saying all these years... + Show Spoiler +
marines are OP.
In all seriousness - I wonder how many other units could be affected by such bugs, it would be interesting to do further testing and see. I think you will however find out that unit balance is conducted based on in game performance ahead of numerical statistics and thus such glitches are ultimately accounted for anyway. Also, when you consider that marines cant float and have nowhere near the same HP, it probably makes sense its DPS is lower/limit.
On July 17 2012 19:22 BronzeKnee wrote: I find it odd that you decided to add armor to the Nexus and Battle Cruiser when doing the damage test. My first reaction is you are trying to hide something, and my second after thinking about it is that it magnifies the results of your test. Because it takes longer (ie more shots) for the Battle Cruiser to do the stated damage you end up with more overall random delays, and thus the Battle Cruiser does less and less DPS over time. .....
It's done to make the HP an multiplier of the damage done to easier be able to measure it.
Also measuring on something with more HP evens out the randomness.
On July 17 2012 19:22 BronzeKnee wrote: I find it odd that you decided to add armor to the Nexus and Battle Cruiser when doing the damage test. My first reaction is you are trying to hide something, and my second after thinking about it is that it magnifies the results of your test. Because it takes longer (ie more shots) for the Battle Cruiser to do the stated damage you end up with more overall random delays, and thus the Battle Cruiser does less and less DPS over time. .....
It's done to make the HP an multiplier of the damage done to easier be able to measure it.
Also measuring on something with more HP evens out the randomness.
I think my tests were pretty clear. Read my whole post. Download and watch the replay I provided.
On July 17 2012 19:22 BronzeKnee wrote: I find it odd that you decided to add armor to the Nexus and Battle Cruiser when doing the damage test. My first reaction is you are trying to hide something, and my second after thinking about it is that it magnifies the results of your test. Because it takes longer (ie more shots) for the Battle Cruiser to do the stated damage you end up with more overall random delays, and thus the Battle Cruiser does less and less DPS over time. .....
It's done to make the HP an multiplier of the damage done to easier be able to measure it.
Also measuring on something with more HP evens out the randomness.
I think my tests were pretty clear. Read my whole post. Download and watch the replay I provided.
Your test is clear, but it is less acurate than the OP. The longer the time, the better the average measurement
On July 17 2012 19:22 BronzeKnee wrote: I find it odd that you decided to add armor to the Nexus and Battle Cruiser when doing the damage test. My first reaction is you are trying to hide something, and my second after thinking about it is that it magnifies the results of your test. Because it takes longer (ie more shots) for the Battle Cruiser to do the stated damage you end up with more overall random delays, and thus you could find that the Battle Cruiser does less and less DPS over time.
So I ran two tests in the unit test map. (The first test against the Hydra begins at 43 seconds and ends at 45, the second against the Nexus begins at 1:45 and ends at 2:21)
First, the BC against a Hydra (low hp and no armor) with no upgrades yields the following: It takes the BC 11 shots to kill the Hydra and it takes the BC 3 seconds to kill the Hydra (and there is actually one more shot on the way toward the Hydra when it does).
So 3/11 = 0.2727 DPS
So in that case the BC actually does more damage than stated!
Second, lets look at the BC vs 1000 Nexus shields with no shield upgrades: It takes 125 shots to kill the Nexus shields and it takes the BC 36 seconds to do so (and again there is one more shot on the way to Nexus when 1000 damage is done).
36/125 = .288 DPS
Again the BC actually does more damage than stated!
So while in your scenario the BC does 22% less damage, in most real scenarios it will do much closer to it's stated DPS, and in some scenarios it even appears to do more damage, and thus I think you are misrepresenting the facts by picking unlikely scenarios.
However, your tests and mine do show problems with the damage system in SC2.
I also thought it seemed odd that the BC testing was done with +3 shields... i guess it simulates somewhat "expected" upgrades for that point in time, but to the same effect you would also assume attack upgrades were also completed probably +3. I'm not sure whether due to scaling +3 attacks is EXACTLY mitigated by +3 armour, (i always thought it was) in which case testing should just be done with no upgrades on either. Also, real world performance which is what balance is based on is unfortunately performed with real world scenarios in mind (at the highest end of play)... It's not everyday you see BC's in PvT, or shield upgrades for that matter.
On July 17 2012 19:22 BronzeKnee wrote: I find it odd that you decided to add armor to the Nexus and Battle Cruiser when doing the damage test. My first reaction is you are trying to hide something, and my second after thinking about it is that it magnifies the results of your test. Because it takes longer (ie more shots) for the Battle Cruiser to do the stated damage you end up with more overall random delays, and thus the Battle Cruiser does less and less DPS over time. .....
It's done to make the HP an multiplier of the damage done to easier be able to measure it.
Also measuring on something with more HP evens out the randomness.
I think my tests were pretty clear. Read my whole post. Download and watch the replay I provided.
But your test doesn't contradict what I said. (I answered to your first statement with him trying to hide something)
The tooltip displays attack rate (period) WITHOUT the random delay.
You're measuring the attack rate (period) WITH the random delay.
Of course they're going to differ! But somehow everybody is concluding that there must be some kind of bug. There isn't! You're just misinterpreting the tooltip.
So in that case the BC actually does damage faster than stated!
36/125 = .288 attack
Again the BC actually does damage faster than stated!
So while in your scenario the BC does 22% less damage, in most real scenarios it will do much closer to it's stated DPS, and in some scenarios it even appears to do more damage
If by stated you mean the attack rate displayed in the game it's 0.225 so your results are much higher than that and much more similar to the OP results(~0.288) also there is really no need to test because just looking at the random delay which is between -0.0625 and 0.1875 we can get the real rate of fire:
average delay = (0.1875 - 0.0625)/2=0.0625
actual rate of fire = stated + delay = 0.225 + 0.0625 = 0.2875
Guys, i dont understand why some of you are saying something like "omg the bc is so bad now, that this is discovered". No. If the attack speed would be 0.2875 without a random delay, nobody would complain in years. But now that it is 0.2875 with the random delay and the tooltip says 0.225 its a big deal?
Its not like the BC got worse because this was discovered t-t
On July 17 2012 18:36 eScaper-tsunami wrote: Random delay actually explains a lot of things that's been happening to me... I go into probe vs probe battles getting a slightly faster attack off at the beginning but end up losing the fight..... thank you random delay, now I know to NEVER fight probe vs probe despite knowing the opponent's probe had the slightly slower first attack...
I fail to understand why we need any random factor in a strategic game. Also one more thing, 'Random' is sometimes f*cked up. You may spawn as a race more than 3 times in a row. That's flawed imho although there's really small chance something like this happens. Like... 1/12 to be one race 4 times in a row?
I am starting to think the average tl posters have horrible math and logic. please brush up on non bayesian probability and law of large numbers. Thanks
Yeah but the main question is why does bc random delay differ from everything else in the game? Is that a mistake by blizz that got in to the game or is there a reason for it?
On July 17 2012 19:54 MasterReY wrote: Guys, i dont understand why some of you are saying something like "omg the bc is so bad now, that this is discovered". No. If the attack speed would be 0.2875 without a random delay, nobody would complain in years. But now that it is 0.2875 with the random delay and the tooltip says 0.225 its a big deal?
Its not like the BC got worse because this was discovered t-t
well, previous you could theory craft with bc dps and shit like that, making it seem more pop efficent unit than it really is(in theory craft), and always falling flat in real games.
Really stupid parameters that don't follow the rules are nothing new in this game, just look at the random armor-less units like Baneling, Ghost, Queen. Random attack delay is also pretty terribad for Hellion micro, I really wonder if BW had this much delay/delay variance.
On July 17 2012 19:54 MasterReY wrote: Guys, i dont understand why some of you are saying something like "omg the bc is so bad now, that this is discovered". No. If the attack speed would be 0.2875 without a random delay, nobody would complain in years. But now that it is 0.2875 with the random delay and the tooltip says 0.225 its a big deal?
Its not like the BC got worse because this was discovered t-t
well, previous you could theory craft with bc dps and shit like that, making it seem more pop efficent unit than it really is(in theory craft), and always falling flat in real games.
theory crafting is for build orders. no pro is sitting around with pencil and paper working out the math of a BC vs nexus and banking their winning strategies on it. so the tooltip is wrong.. maybe, depends on if you think random delay should be included or not. but really everyone who plays this game at a competent level plays more by feel and experience than actually running a dps meter in their head. It's not like people have been underproducing BCs because of the false sense of security of a tooltip telling them they're better than they actually are.
and another thing, people are buying into 1 off experiments way too easily. OP did 1 test, or at least only reported 1 test, with weird conditions like +3 shields and +2 armor, just unnecessary independent variables that didn't need to be included. and without anyone backing up his findings or anyone repeating his tests to be sure so many people immediately see the results and take it as truth immediately. it's the same witch hunt mentality that gets people going crazy whenever a new hacking accusation pops up.
now we know the ~22dps difference can be attributed to random attack delays unique to BC and reaper. so everyone can just chilllll
My first and only question.... did you take shield regen in to account? (do shield regen in combat? can't remember) 4-8 marauders kill a nexus so fast that regen rate doesn't really effect it much but BC's takin down a nexus in 58 seconds would allow for a fair bit of regen on the shields....
On July 17 2012 21:45 emythrel wrote: My first and only question.... did you take shield regen in to account? 4-8 marauders kill a nexus so fast that regen rate doesn't really effect it much but BC's takin down a nexus in 58 seconds would allow for a fair bit of regen on the shields....
protoss shields doesnt regen inside battle, after 10 sec without taking dmg they reg 2 hp per sek is how it works
some units, like reaper for example has random chance to gain a random cooldown (both very low so its barely noticeable). thats why results like this are unexpected when tested
On July 17 2012 19:01 Andr3 wrote: Wow I didn't know about the random delay attacks, this is quite huge...it can decide games.
Why would you have randomness in the game just so it can look "pretty", damnit Blizzard.
i believe sc1 had random cooldowns on a few units aswell that were barely noticable. an easy way to test it is just by have a bunch of units attack a building the same time and watch how they desync after a few seconds
this random delay has only been annoying for me when 5rax reaper was relevant. because optimal kiting would always be either include the random delay in your kiting timing making your kiting very reliable and stable but at the loss off potential dps while attacking discluding the random delay would result in 1-2 reapers not fire-ing at times because they were still on cooldown. but the d3 charge was abit more annoying cause that random felt bigger so when kiting queens while attacking buildings resulted in reapers not fireing at all at times because they had more delay than normal sometimes
On July 17 2012 21:45 emythrel wrote: My first and only question.... did you take shield regen in to account? 4-8 marauders kill a nexus so fast that regen rate doesn't really effect it much but BC's takin down a nexus in 58 seconds would allow for a fair bit of regen on the shields....
Are you serious? this is sc2 not BW(where shields would recharge no matter what) if you keep attacking a protoss unit or structure it won't recharge shield. Shields start recharging again 10 seconds after the last time the protoss unit/ structure was hit.
To those saying that there might be a problem with the testing the real rate of fire can by determined by theory knowing the average random delay and adding it to the nominal rate of fire(the tooltip one).
for BC the random delay it's between -0.0625 and 0.1875 average random delay=(0.1875-0.0625)/2=0.0625
real rate of fire = 0.225 + 0.0625 = 0.2875 which is almost identical to the OP's results
for the reaper the random delay it's between 0.1 and 0.5. Average delay it's 0.3
real rate of fire = 1.8 + 0.3 = 2.1 again almost identical the the OP's results
Also Random delay it's actually on all units but it's barely noticeable: for all the units except reaper and BC the average random delay it's 0.03125.(the delay it's between -0.0625 and 0.125)
On July 17 2012 21:45 emythrel wrote: My first and only question.... did you take shield regen in to account? 4-8 marauders kill a nexus so fast that regen rate doesn't really effect it much but BC's takin down a nexus in 58 seconds would allow for a fair bit of regen on the shields....
protoss shields doesnt regen inside battle, after 10 sec without taking dmg they reg 2 hp per sek is how it works
some units, like reaper for example has random chance to gain a random cooldown (both very low so its barely noticeable). thats why results like this are unexpected when tested
nice to know, personally the dps of units isn't important to me. I already knew not to make BCs or reapers (with a few exceptions) lol
So, the bug is actually the tooltip. Dont see a way for blizz to fix it, They are simple not gonna put a crazy math description on the unit and at most any value they put there will be average.
On July 17 2012 21:45 emythrel wrote: My first and only question.... did you take shield regen in to account? 4-8 marauders kill a nexus so fast that regen rate doesn't really effect it much but BC's takin down a nexus in 58 seconds would allow for a fair bit of regen on the shields....
Are you serious? this is sc2 not BW(where shields would recharge no matter what) if you keep attacking a protoss unit or structure it won't recharge shield. Shields start recharging again 10 seconds after the last time the protoss unit/ structure was hit.
To those saying that there might be a problem with the testing the real rate of fire can by determined by theory knowing the average random delay and adding it to the nominal rate of fire(the tooltip one).
for BC the random delay it's between -0.0625 and 0.1875 average random delay=(0.1875-0.0625)/2=0.0625
real rate of fire = 0.225 + 0.0625 = 0.2875 which is almost identical to the OP's results
for the reaper the random delay it's between 0.1 and 0.5. Average delay it's 0.3
real rate of fire = 1.8 + 0.3 = 2.1 again almost identical the the OP's results
Also Random delay it's actually on all units but it's barely noticeable: for all the units except reaper and BC the average random delay it's 0.03125.(the delay it's between -0.0625 and 0.125)
Then this explains it. However the displayed dps in the tooltip is still wrong, since it should account for that random delay.
For me, even though the random delay explains it, that's just poor game design. Why should these units be special? If we want to encourage amazing micro and decision making, shouldn't outcomes be relatively predictable, even in their randomness?
What I mean is, why have two units doing significantly less DPS than stated, while other units do roughly the same?
Why change the random delay on these units, while all other units are mostly equal? If it was a balance issue, then the attack speed should have been adjusted. If it was an aesthetics issues (with the BC) then the tooltip should reflect a more accurate expected DPS somehow.
Though, now that you point it out, it did always feel to me like the BC fired its weapons at odd intervals.
On July 17 2012 19:54 MasterReY wrote: Guys, i dont understand why some of you are saying something like "omg the bc is so bad now, that this is discovered". No. If the attack speed would be 0.2875 without a random delay, nobody would complain in years. But now that it is 0.2875 with the random delay and the tooltip says 0.225 its a big deal?
Its not like the BC got worse because this was discovered t-t
its exactly as I thought when I was reading through the thread and people stating that this was the reason why the bc was so bad. No one's going to analyse the damage to such a fine point, if it is bad it can be changed accordingly but as for now the units are how blizzard wants to leave them
On July 17 2012 21:50 nocrA wrote: for BC the random delay it's between -0.0625 and 0.1875 average random delay=(0.1875-0.0625)/2=0.0625
real rate of fire = 0.225 + 0.0625 = 0.2875 which is almost identical to the OP's results
for the reaper the random delay it's between 0.1 and 0.5. Average delay it's 0.3
real rate of fire = 1.8 + 0.3 = 2.1 again almost identical the the OP's results
I can confirm that the real attack speed is in that area.
I've build a short test map to do some attack time tests which are more accurate than looking at the ingame timer. :D
Have a screenshot of one test run:
- It records the time required between the first dealt damage and the killing blow and counts the attacks. -> time between the first damage and last damage. - attackSpeed = time / (attacks - 1) [first attack is subtracted as it starts the timer -> every attack has a cooldown in the time] - I've added 3 targets: one ground, one air, one ground structure with each 1k life, 0 armor, no health regeneration. - The attack speed is noted in the top left area of the screen.
That small map I made was made in the 1.5 arcade beta, but the BC had no balance changes regarding its rate of fire. So this still applies.
If people want me to make a usable test map on battle.net, just ask nicely and tell me features you would like to have like changeable unit type, repeatable tests, adjustable HP on the target test (or setting the amount of attacks that should be recorded), automated tests for all units, ...
I have also independently confirmed the findings for Battlecruiser AA and AT.
I have tested 1 BC firing against another, with +0 attack and +3 armor upgrades, repeated 9 times. Since BC AA damage is 6 and the max upgraded armor for the BC is 6, it means each attack was doing the minimum possible damage of 0.5 health. Since the BC has 550 health, it would take 1100 shots to destroy it. The mean time needed for the target to be destroyed was 317 seconds (varying from 315 to 320, most likely due to having +/- 1 second accuracy at best). Attacking 1100 times in 317 seconds gives a delay between attacks of 0.288s, in accordance with OP.
I also tested 1 BC against a +3 armored Thor. The Thor has 400 health and a max armor of 4, BC AT attack has 8 damage, so each shot deals 4 damage. Tested it 18 times and the mean time for target destruction was 29s, varying between 28 and 30. The BC needs 100 shots to kill the Thor, and it is done in 29 seconds, with means a delay between attacks of 0.29, again in accordance with the OP.
So yeah, the tooltip info is incorrect and they probably failed to take the random delay into account.
On July 17 2012 19:22 BronzeKnee wrote: I find it odd that you decided to add armor to the Nexus and Battle Cruiser when doing the damage test. My first reaction is you are trying to hide something, and my second after thinking about it is that it magnifies the results of your test. Because it takes longer (ie more shots) for the Battle Cruiser to do the stated damage you end up with more overall random delays, and thus you could find that the Battle Cruiser does less and less DPS over time.
So I ran two tests in the unit test map. (The first test against the Hydra begins at 43 seconds and ends at 45, the second against the Nexus begins at 1:45 and ends at 2:21)
First, the BC against a Hydra (low hp and no armor) with no upgrades yields the following: It takes the BC 11 shots to kill the Hydra and it takes the BC 3 seconds to kill the Hydra (and there is actually one more shot on the way toward the Hydra when it does).
So 3/11 = 0.2727 attack speed
So in that case the BC actually does damage faster than stated!
Second, lets look at the BC vs 1000 Nexus shields with no shield upgrades: It takes 125 shots to kill the Nexus shields and it takes the BC 36 seconds to do so (and again there is one more shot on the way to Nexus when 1000 damage is done).
36/125 = .288 attack
Again the BC actually does damage faster than stated!
So while in your scenario the BC does 22% less damage, in most real scenarios it will do much closer to it's stated DPS, and in some scenarios it even appears to do more damage, and thus I think you are misrepresenting the facts by picking unlikely scenarios.
However, your tests and mine do show problems with the damage system in SC2.
How is 0.2727 and 0.288 faster than 0.225 which is stated??!
On July 17 2012 19:22 BronzeKnee wrote: I find it odd that you decided to add armor to the Nexus and Battle Cruiser when doing the damage test. My first reaction is you are trying to hide something, and my second after thinking about it is that it magnifies the results of your test. Because it takes longer (ie more shots) for the Battle Cruiser to do the stated damage you end up with more overall random delays, and thus you could find that the Battle Cruiser does less and less DPS over time.
So I ran two tests in the unit test map. (The first test against the Hydra begins at 43 seconds and ends at 45, the second against the Nexus begins at 1:45 and ends at 2:21)
First, the BC against a Hydra (low hp and no armor) with no upgrades yields the following: It takes the BC 11 shots to kill the Hydra and it takes the BC 3 seconds to kill the Hydra (and there is actually one more shot on the way toward the Hydra when it does).
So 3/11 = 0.2727 attack speed
So in that case the BC actually does damage faster than stated!
Second, lets look at the BC vs 1000 Nexus shields with no shield upgrades: It takes 125 shots to kill the Nexus shields and it takes the BC 36 seconds to do so (and again there is one more shot on the way to Nexus when 1000 damage is done).
36/125 = .288 attack
Again the BC actually does damage faster than stated!
So while in your scenario the BC does 22% less damage, in most real scenarios it will do much closer to it's stated DPS, and in some scenarios it even appears to do more damage, and thus I think you are misrepresenting the facts by picking unlikely scenarios.
However, your tests and mine do show problems with the damage system in SC2.
In 'real scenarios' you could just as easily get a much longer delay due to randomness just as easily as the shorter delay you found. Testing on large HP buildings is a better test for the effect of the delay. It averages out the randomness, which more than likely is like WC3 crit chance and isn't 'true' random, and better ignores overkill effects like your hydra test. There's nothing 'odd' about his choice. Choosing a Nexus or rock is the correct choice.
In either case every test shows that the real dps is pretty off of the tooltip dps.
The random delay between attacks has been known about since early beta. There was a pretty famous thread where a guy's zealot swung first in a zealot vs zealot fight and ended up losing, thus discovering this random delay (this is before the map editor was released to the public).
For the vast majority of the units in the game, this random delay is much too small to matter. I know everyone wants SC to always be 100% consistent (fyi, BW had much more randomness on high ground damage calculations), but these random delays should never be enough to be a real deciding factor in a game, and they aren't. They are there primarily for cosmetic purposes.
Why they decided to give the reaper and BC a much bigger random window than the other units is beyond me, and it should probably be reduced.
On July 17 2012 15:40 kazie wrote: you're completely disregarding the random attack delay. BC and reaper D8 are the only 2 attacks that have a different random attack delay from the default. real attack speed of a BC is 0.1625 to 0.4125 which is 0.2875 average probably same with d8.
That doesn't change the problem that tooltip states incorrect values.
tooltips are correct in that blizzard chose to state base values, and that random delays are public knowledge. you can disagree with their decision i guess.
only thing i would say needs to be changed is liquipedia. it should at least have an attack speed page explaining the matter. if op went to another wiki like sc2pod, he would have known the cause of this.
Random delays doesn't seems to effect other terran units in any meaningful way, as their test results match theoretical values, I see no reason why BC and reapers should be different.
As was previously stated by another poster, most units' attack delays are +X/-X, where both values are the same. That means that the average delay is 0, and over a large number of attacks the units' DPS is unaffected.
BCs' delay is -.0625/+.1865, which means that the average delay (assuming all values are equally possible) is .0625. Given a base attack speed of 0.23, that means the actual average attack speed is 0.2925, a 21.4% increase. That roughly matches the results given by the OP.
Cool explaination, but that is a really odd way to have the BC attack speed done. Any guesses as to why they do it like that for the BC?
On July 18 2012 01:04 HardlyNever wrote: Why they decided to give the reaper and BC a much bigger random window than the other units is beyond me, and it should probably be reduced.
I think the main problem isn't that there's a large window, just that it doesn't average out to be very close to the actual number. All units are this way, but BCs moreso, due to their higher on average random delay and also very short attack period to begin with (it's a very significant difference at around %22). Reapers are just dumb because the fastest they can ever attack, even with the shortest random delay they still have a longer period than the tooltip states.
I think the OP's results are significant, and should be fixed by Blizzard. It's like having your marines stutter step, and not firing a shot once a while due to random delay.
Blizzard shouldn't indicate the rate of fire is 0.225 and have the actual rate of fire significantly higher at 0.28.
On July 17 2012 19:22 BronzeKnee wrote: I find it odd that you decided to add armor to the Nexus and Battle Cruiser when doing the damage test. My first reaction is you are trying to hide something, and my second after thinking about it is that it magnifies the results of your test. Because it takes longer (ie more shots) for the Battle Cruiser to do the stated damage you end up with more overall random delays, and thus you could find that the Battle Cruiser does less and less DPS over time.
So I ran two tests in the unit test map. (The first test against the Hydra begins at 43 seconds and ends at 45, the second against the Nexus begins at 1:45 and ends at 2:21)
First, the BC against a Hydra (low hp and no armor) with no upgrades yields the following: It takes the BC 11 shots to kill the Hydra and it takes the BC 3 seconds to kill the Hydra (and there is actually one more shot on the way toward the Hydra when it does).
So 3/11 = 0.2727 attack speed
So in that case the BC actually does damage faster than stated!
Second, lets look at the BC vs 1000 Nexus shields with no shield upgrades: It takes 125 shots to kill the Nexus shields and it takes the BC 36 seconds to do so (and again there is one more shot on the way to Nexus when 1000 damage is done).
36/125 = .288 attack
Again the BC actually does damage faster than stated!
So while in your scenario the BC does 22% less damage, in most real scenarios it will do much closer to it's stated DPS, and in some scenarios it even appears to do more damage, and thus I think you are misrepresenting the facts by picking unlikely scenarios.
However, your tests and mine do show problems with the damage system in SC2.
On July 17 2012 19:22 BronzeKnee wrote: I find it odd that you decided to add armor to the Nexus and Battle Cruiser when doing the damage test. My first reaction is you are trying to hide something, and my second after thinking about it is that it magnifies the results of your test. Because it takes longer (ie more shots) for the Battle Cruiser to do the stated damage you end up with more overall random delays, and thus you could find that the Battle Cruiser does less and less DPS over time.
So I ran two tests in the unit test map. (The first test against the Hydra begins at 43 seconds and ends at 45, the second against the Nexus begins at 1:45 and ends at 2:21)
First, the BC against a Hydra (low hp and no armor) with no upgrades yields the following: It takes the BC 11 shots to kill the Hydra and it takes the BC 3 seconds to kill the Hydra (and there is actually one more shot on the way toward the Hydra when it does).
So 3/11 = 0.2727 attack speed
So in that case the BC actually does damage faster than stated!
Second, lets look at the BC vs 1000 Nexus shields with no shield upgrades: It takes 125 shots to kill the Nexus shields and it takes the BC 36 seconds to do so (and again there is one more shot on the way to Nexus when 1000 damage is done).
36/125 = .288 attack
Again the BC actually does damage faster than stated!
So while in your scenario the BC does 22% less damage, in most real scenarios it will do much closer to it's stated DPS, and in some scenarios it even appears to do more damage, and thus I think you are misrepresenting the facts by picking unlikely scenarios.
However, your tests and mine do show problems with the damage system in SC2.
lol dat math
lol indeed... I was lost and had to reread it 3 times and like WTF??!
On July 17 2012 18:36 eScaper-tsunami wrote: Random delay actually explains a lot of things that's been happening to me... I go into probe vs probe battles getting a slightly faster attack off at the beginning but end up losing the fight..... thank you random delay, now I know to NEVER fight probe vs probe despite knowing the opponent's probe had the slightly slower first attack...
I fail to understand why we need any random factor in a strategic game. Also one more thing, 'Random' is sometimes f*cked up. You may spawn as a race more than 3 times in a row. That's flawed imho although there's really small chance something like this happens. Like... 1/12 to be one race 4 times in a row?
Actually, that's incorrect. The chance for spawn any race 4 times in a row is 1/27. + Show Spoiler +
The first roll doesn't matter, it just determines what race you are rolling for, and then the remaining is 1/3x1/3x1/3
If you want a specific race four times in a row that would be 1/81.
And yeah, I've personally rolled Protoss five times in a row in team games before as Random.
On July 17 2012 18:36 eScaper-tsunami wrote: Random delay actually explains a lot of things that's been happening to me... I go into probe vs probe battles getting a slightly faster attack off at the beginning but end up losing the fight..... thank you random delay, now I know to NEVER fight probe vs probe despite knowing the opponent's probe had the slightly slower first attack...
I fail to understand why we need any random factor in a strategic game. Also one more thing, 'Random' is sometimes f*cked up. You may spawn as a race more than 3 times in a row. That's flawed imho although there's really small chance something like this happens. Like... 1/12 to be one race 4 times in a row?
Actually, that's incorrect. The chance for spawn any race 4 times in a row is 1/27. + Show Spoiler +
The first roll doesn't matter, it just determines what race you are rolling for, and then the remaining is 1/3x1/3x1/3
If you want a specific race four times in a row that would be 1/81.
And yeah, I've personally rolled Protoss five times in a row in team games before as Random.
Edit: Fail Spoiler.
Their RNG is definitely not the best. I've rolled Terran 11 times in a row in team games, and same race 5-6 in a row is not uncommon. This is actually pretty common as depending on the seed chosen, RNG can have distinct patterns.
On July 17 2012 18:36 eScaper-tsunami wrote: Random delay actually explains a lot of things that's been happening to me... I go into probe vs probe battles getting a slightly faster attack off at the beginning but end up losing the fight..... thank you random delay, now I know to NEVER fight probe vs probe despite knowing the opponent's probe had the slightly slower first attack...
I fail to understand why we need any random factor in a strategic game. Also one more thing, 'Random' is sometimes f*cked up. You may spawn as a race more than 3 times in a row. That's flawed imho although there's really small chance something like this happens. Like... 1/12 to be one race 4 times in a row?
Actually, that's incorrect. The chance for spawn any race 4 times in a row is 1/27. + Show Spoiler +
The first roll doesn't matter, it just determines what race you are rolling for, and then the remaining is 1/3x1/3x1/3
If you want a specific race four times in a row that would be 1/81.
And yeah, I've personally rolled Protoss five times in a row in team games before as Random.
Edit: Fail Spoiler.
Their RNG is definitely not the best. I've rolled Terran 11 times in a row in team games, and same race 5-6 in a row is not uncommon. This is actually pretty common as depending on the seed chosen, RNG can have distinct patterns.
There's no proof behind this, and I definately don't have a large enough sample size, but my personal experience is that it tends to pick the race that I've played the least, as of recent.
On July 17 2012 18:36 eScaper-tsunami wrote: Random delay actually explains a lot of things that's been happening to me... I go into probe vs probe battles getting a slightly faster attack off at the beginning but end up losing the fight..... thank you random delay, now I know to NEVER fight probe vs probe despite knowing the opponent's probe had the slightly slower first attack...
I fail to understand why we need any random factor in a strategic game. Also one more thing, 'Random' is sometimes f*cked up. You may spawn as a race more than 3 times in a row. That's flawed imho although there's really small chance something like this happens. Like... 1/12 to be one race 4 times in a row?
Actually, that's incorrect. The chance for spawn any race 4 times in a row is 1/27. + Show Spoiler +
The first roll doesn't matter, it just determines what race you are rolling for, and then the remaining is 1/3x1/3x1/3
If you want a specific race four times in a row that would be 1/81.
And yeah, I've personally rolled Protoss five times in a row in team games before as Random.
Edit: Fail Spoiler.
Their RNG is definitely not the best. I've rolled Terran 11 times in a row in team games, and same race 5-6 in a row is not uncommon. This is actually pretty common as depending on the seed chosen, RNG can have distinct patterns.
There's no proof behind this, and I definately don't have a large enough sample size, but my personal experience is that it tends to pick the race that I've played the least, as of recent.
I wish they'd hire me (SQA mostly working with automation / database and some UI)... I'd think they'd have significantly fewer bugs. I QA apps that handles transaction up to millions daily and all of them are in infinitely better shape than their RMAH. Oh well.
On July 18 2012 02:13 Lefiathen wrote: .... the nexus had shields in the BC cruiser test, so the BC ovbiously is going to take more atacks to destroy the nexus...
Did you consider that the attack rate values are given for Normal speed setting. So that's why you end up with more time between attacks, because Faster game speed is faster than normal?
On July 18 2012 02:53 Yoshi Kirishima wrote: Did you consider that the attack rate values are given for Normal speed setting. So that's why you end up with more time between attacks, because Faster game speed is faster than normal.
On July 18 2012 02:53 Yoshi Kirishima wrote: Did you consider that the attack rate values are given for Normal speed setting. So that's why you end up with more time between attacks, because Faster game speed is faster than normal.
I'm pretty sure that the measurements are based on in-game time, so it shouldn't matter.
On July 18 2012 02:53 Yoshi Kirishima wrote: Did you consider that the attack rate values are given for Normal speed setting. So that's why you end up with more time between attacks, because Faster game speed is faster than normal.
it's correct for all other units.
Thanks.
Also oops, you counted in-game time, so ignore my post.
The real problem here is that the randomness is not symmetric. The fire rate of a BC is 0.225 + some random number between -0.0625 and 0.1875. As previously stated, that makes the average attack rate 0.2250 + (0.1875-0.0625)/2 = 0.2875 Blizzard should report it as 0.2875 +/- 0.1250. That would be far less confusing.
Are you accounting for the difference between real time and Blizzard time, were their any conversions messed up in those methods, because that would throw it off by a similar amount?
On July 18 2012 05:42 RedDragon571 wrote: Are you accounting for the difference between real time and Blizzard time, were their any conversions messed up in those methods, because that would throw it off by a similar amount?
He's using in-game clock so the time point is irrelevant.
On July 18 2012 05:42 RedDragon571 wrote: Are you accounting for the difference between real time and Blizzard time, were their any conversions messed up in those methods, because that would throw it off by a similar amount?
He's using in-game clock so the time point is irrelevant.
wouldn't this be something for the battlenet forums instead? the only way this will produce results if it is bugged is if blizzard sees it which is far from a guarantee on TL unless this gets to be 100+ pages or headline worthy
Sadly, I wrote my post and did my tests at 5 am this morning and wasn't thinking and thus apologize. No more posting before first shift. The original post is correct.
Yeah it's kind of like... so what? If Blizzard wanted BCs and reapers to be more powerful then they would be more powerful, regardless of whether or not the tooltip DPSes are correct.
On July 17 2012 18:36 eScaper-tsunami wrote: Random delay actually explains a lot of things that's been happening to me... I go into probe vs probe battles getting a slightly faster attack off at the beginning but end up losing the fight..... thank you random delay, now I know to NEVER fight probe vs probe despite knowing the opponent's probe had the slightly slower first attack...
I fail to understand why we need any random factor in a strategic game. Also one more thing, 'Random' is sometimes f*cked up. You may spawn as a race more than 3 times in a row. That's flawed imho although there's really small chance something like this happens. Like... 1/12 to be one race 4 times in a row?
Actually, that's incorrect. The chance for spawn any race 4 times in a row is 1/27. + Show Spoiler +
The first roll doesn't matter, it just determines what race you are rolling for, and then the remaining is 1/3x1/3x1/3
If you want a specific race four times in a row that would be 1/81.
And yeah, I've personally rolled Protoss five times in a row in team games before as Random.
Edit: Fail Spoiler.
Their RNG is definitely not the best. I've rolled Terran 11 times in a row in team games, and same race 5-6 in a row is not uncommon. This is actually pretty common as depending on the seed chosen, RNG can have distinct patterns.
There's no proof behind this, and I definately don't have a large enough sample size, but my personal experience is that it tends to pick the race that I've played the least, as of recent.
The distribution from SC2s RNG is very even. Getting Terran 11 times in a row will happen.
It's actually pretty interesting: whenever a pseudo-random number generator is involved, conspiracies about a bias based on an incredibly minute sample size will arise. Even when we understand that 11 is a small sample when we consider how many random numbers the game generates, we still like to create a reason behind the madness.
On July 18 2012 01:04 HardlyNever wrote: The random delay between attacks has been known about since early beta. There was a pretty famous thread where a guy's zealot swung first in a zealot vs zealot fight and ended up losing, thus discovering this random delay (this is before the map editor was released to the public).
For the vast majority of the units in the game, this random delay is much too small to matter. I know everyone wants SC to always be 100% consistent (fyi, BW had much more randomness on high ground damage calculations), but these random delays should never be enough to be a real deciding factor in a game, and they aren't. They are there primarily for cosmetic purposes.
Why they decided to give the reaper and BC a much bigger random window than the other units is beyond me, and it should probably be reduced.
Actually this makes a real big difference in mirror match up spawns in the early game on the custom game Desert Strike in expert mode. Also with Firebat vs Dark Zealot/Zealot mix spawns in spawn 1 and 2
Would be nice if you put a final word or tl;dr for people who don't want to waste 10 minutes of their lives parsing through meaningless numbers and percentages. But nice work, it's a good find.
This is not a bug, but a feature, that is clearly displayed in the map editor, so pretty much everyone that has played around with it, knowns about it. BW also had random attack delay.
All attacks except the PDD's anti-missile attack have a random delay. This is a property inherited by all weapons and if some weapon has different values that means it was changed specifically, so it's unlikely that different values are put there by mistake.
Most units have it at -0.0625 to 0.1250 or from -1 to +2 frames, which is the same as in BW, so on average the attack times for most units are 0.0625 longer. This is introduced to make units look less like robots that fire in perfect unison and to spread out the attack calculations that have to be done by the processor.
The BC has -0.0625 to 0.1875, likely to make the attacks look more like multiple independent batteries firing, instead of a constant stream.
Reaper mines have a random delay of 0.1 to 0.5 seconds.
All units have slightly lower DPS than the cooldown by itself would imply and all weapon tooltips display just the cooldown.
On July 17 2012 20:48 ciox wrote: Really stupid parameters that don't follow the rules are nothing new in this game, just look at the random armor-less units like Baneling, Ghost, Queen. Random attack delay is also pretty terribad for Hellion micro, I really wonder if BW had this much delay/delay variance.
helions do not have random attack delay as you mean them to have.
they turn ttheir turrets around before firing. thats why they have much smaller delay before they fire while pursiing then kiting.
On July 18 2012 05:52 chuDr3t4 wrote: Random attack delay is in BW too lol. Not a bug, just a feature.
Yeah it's actually funny how many people like to point to BW as the pinnacle of competitive games because of its lack of randomness, but BW had MANY random elements from high-ground miss chance to random firing delays. If anything, SC2 is LESS random than BW which is sort of an ironic conclusion for many people.
Ever wonder why it was so difficult in BW to kill off both your remaining SCVs when you were an observer? Cooldown variance... aka random attack delay.
Well...this combined with the weak attack of battlecruisers (high fire rate low damage) explains why they simly DONT kill things fast enough.
Thankfully they are still flying beasts that just don't die when supported/upgraded...
They should fix the tool tip or alter teh "true" DPS of the battlecruiser. Cruisers are really difficult to get out in sufficient numbers to matter, let alone have upgraded. No reason to allow them to be further gimped by in game mechanics when other, simpler units such as tha marauder are uneffected.
On July 17 2012 18:36 eScaper-tsunami wrote: Random delay actually explains a lot of things that's been happening to me... I go into probe vs probe battles getting a slightly faster attack off at the beginning but end up losing the fight..... thank you random delay, now I know to NEVER fight probe vs probe despite knowing the opponent's probe had the slightly slower first attack...
I fail to understand why we need any random factor in a strategic game. Also one more thing, 'Random' is sometimes f*cked up. You may spawn as a race more than 3 times in a row. That's flawed imho although there's really small chance something like this happens. Like... 1/12 to be one race 4 times in a row?
I've got news for you: spawning as one race 4 times in a row is just as likely as any other pattern of spawning (like zerg then protoss then terran then zerg). And no, the odds of spawning as one race 4 times in a row (or any particular racial spawning pattern is one in 3^4, or 1 in 81. (Actually it's 3/81, given that it doesn't matter which race you spawn as, but the odds of spawning as any given race 4 times in a row is 1/81).
On July 18 2012 05:52 chuDr3t4 wrote: Random attack delay is in BW too lol. Not a bug, just a feature.
Yeah it's actually funny how many people like to point to BW as the pinnacle of competitive games because of its lack of randomness, but BW had MANY random elements from high-ground miss chance to random firing delays. If anything, SC2 is LESS random than BW which is sort of an ironic conclusion for many people.
Ever wonder why it was so difficult in BW to kill off both your remaining SCVs when you were an observer? Cooldown variance... aka random attack delay.
I've known about this issue for a long time, and have personally wondered why the random modifiers on periodic attacks aren't always equalized on both sides (ex +0.1 and -0.1), or at least don't include the discrepancy into the in-game listed attack period.
I guess that a call to action about this issue is way overdue.
These anomalies exists on more units than just reaper and battle cruiser. I think I recall it existing with zerglings, and may also exist on units like infested terrans, broodlings, or other units that attack quite quickly.
These random attack period modifiers do have a significant effect, but I don't think that they are the only cause.
The rest of the cause is probably due to limited sections of time that the game engine can work with with regards to attacks, which is not a simple thing to hash out exactly, but could probably be approximated.
One thing I still dont get, how do replays work with the random delay? Does the replay need to include what the computer rolled? My understanding that the replay was just the player commands.
On July 23 2012 20:44 Archerofaiur wrote: One thing I still dont get, how do replays work with the random delay? Does the replay need to include what the computer rolled? My understanding that the replay was just the player commands.
Computer randomisation is psuedo random. I think if you start with an identical seed you can replicate the identical randomisations as long as they happen at the same time relative to the initial seeding if the seeding is based on time.
On July 23 2012 20:44 Archerofaiur wrote: One thing I still dont get, how do replays work with the random delay? Does the replay need to include what the computer rolled? My understanding that the replay was just the player commands.
It probably uses a seed from the start time, player ids or something like that.
That way the RNG will calculate the same numbers every time.
(Pseudo) Random Number Generators is provided a seed when initialized and is from there deterministic. Example: Say a really bad RNG looks like this: y = x^2 (mod 13) Then it will give the chain 2 -> 4 -> 3 -> 9 -> 3 -> 9 -> ... if initialized with the number 2 and 5 -> 12 -> 1 -> 1 -> ... if initialized with the number 5
Yes, if you need a lot of random numbers in your program, only the very first random number (the seed) is done by looking at stuff like the hardware clock and other hardware measurements. This would be too slow to do all the time, the rest of the random numbers are produced by doing some kind of simple calculation on the previous random number, so everything can be reproduced if the seed is saved in the replay.
EPT
Or have maybe the Maurauder do a dps test against a +3 shield Nexus, because for the maurauder you did it without shield upgrades, but for the others, you did![[image loading]](http://i.imgur.com/I42sI.jpg)
![[image loading]](http://img36.imageshack.us/img36/3338/screenshot2012071716004.jpg)
