I made 2 zerglings each attacking a different tank at the same time and then saved the game. When I repeatedly load the game it seems random which tank dies first. When I tested this with hydralisks I had the same inconsistent results. I made a video demonstrating my test.
I only show several attempts in the video, but I've tried it many times to make sure I wasn't just getting extremely lucky and seeing something rare. I didn't fish for these examples. I recorded once and saved the results.
I just remembered the ranged miss chance could be a factor for the hydras. I should have tested another melee unit.
EDIT: while trying to google for an answer I stumbled on the liquipedia page for BWcalc which states,
"It is generally accepted that there exists some randomness in attack cooldowns in Brood War. While the exact specifics are not known, it is believed that attack cooldowns can vary by approximately +/- 1-4 logical steps per attack"
Watching the video in slo-mo, you can see the time between zergling attacks is not constant. I'm not sure how the game calculates the cooldown between attacks, but clearly it worked out differently for these two zerglings. Two possibilities are a randomization factor or a cyclic pattern where the two zerglings each started at different points in the (asymmetric) cycle.
Isn't there a random 1 frame micro-delay, probably initially added so units like marines shoot more "naturally" and not at exactly the same frame like synchronized robots? With units like lings where their attack cooldown is very short - 8 frames without glands, 5 with glands - even 1 frame random delay can accumulate to noticeable difference.
On December 11 2024 18:40 ZeroByte13 wrote: Isn't there a random 1 frame micro-delay, probably initially added so units like marines shoot more "naturally" and not at exactly the same frame like synchronized robots? With units like lings where their attack cooldown is very short - 8 frames without glands, 5 with glands - even 1 frame random delay can accumulate to noticeable difference.
On December 11 2024 18:40 ZeroByte13 wrote: Isn't there a random 1 frame micro-delay, probably initially added so units like marines shoot more "naturally" and not at exactly the same frame like synchronized robots? With units like lings where their attack cooldown is very short - 8 frames without glands, 5 with glands - even 1 frame random delay can accumulate to noticeable difference.
random 1 frame micro delay for every attack? does this mean a unit could get unlucky and attack several frames slower than a unit that did not roll these delays?
I repeated the test for units that attack slower like zealots and DTs and saw the same results. It was a pain to get them to attack at the exact same time, but on the DT test I had them attack a tank which only takes 4 hits to kill. Even if a DT initially attacked 1-2 frames later than the other DT, sometimes it would kill the tank sooner than the DT that attacked first.
Not sure about zerglings, but if I'm not mistaken there are a bunch of random factors in attacking. Damage can sometimes vary, cooldowns can vary. Over enough attacks they all average out, but looking at individual zerglings doing damage might not be the best. Now, I'm not sure this holds for zerglings, as quite a lot of the RNG in things was removed. But even if all RNG is removed you'll still have to question how things are implemented. If something attacks every 0.497 seconds, as zerglings do, you might just have some CPU constraints that with enough attacks, they'll desync due to floating point operations, the way things are parallelized (or serialized), etc. I'm guessing that in the design of the game it was less important to have two zerglings kill tanks in *exactly* the same time every time, and more important that a swarm of zerglings responds rapidly and agilely to your commands without syncing all units all the time everywhere.
On December 15 2024 22:53 Acrofales wrote: Not sure about zerglings, but if I'm not mistaken there are a bunch of random factors in attacking. Damage can sometimes vary, cooldowns can vary. Over enough attacks they all average out, but looking at individual zerglings doing damage might not be the best. Now, I'm not sure this holds for zerglings, as quite a lot of the RNG in things was removed. But even if all RNG is removed you'll still have to question how things are implemented. If something attacks every 0.497 seconds, as zerglings do, you might just have some CPU constraints that with enough attacks, they'll desync due to floating point operations, the way things are parallelized (or serialized), etc. I'm guessing that in the design of the game it was less important to have two zerglings kill tanks in *exactly* the same time every time, and more important that a swarm of zerglings responds rapidly and agilely to your commands without syncing all units all the time everywhere.
I tested a bunch of other units too including DTs on tanks which only take 4 attacks to kill. In a later post here I linked a video of a DT attacking later than the other DT yet it kills the tank sooner. The conclusion seems to be there is indeed randomness in attack cooldowns. The liquipedia page for BWcalc describes it as "plus or minus 1-4 logical steps per attack" which at fastest game speed means attacks can vary from 42-168 milliseconds per attack.
While this doesn't have a big effect, I imagine it has impacted game results at times when a high damage unit was killed milliseconds before an attack came out like maybe with a big reaver or lurker shot. It's not like anything can be done about it, that's just how the game was coded, but I think it's interesting to know. Many people think DPS is constant and I thought it was as well.
This reminds me of one time I saw a KR stream where some pros seemed convinced that a zergling could 1vs1 an SCV in a direct fight if the zergling was perfectly microed. Like, someone at the level of Shuttle was spending 20-30 minutes clicking the ling, and trying to subtly influence the attack speed of the ling by declaring attacks really precisely. They never succeeded. The SCV always survived and the ling always died. Perhaps it is possible to influence the random factor by supplying some sort of really specifically timed manual input. Definitely not sure about that, though.
On December 18 2024 02:19 ThunderJunk wrote: This reminds me of one time I saw a KR stream where some pros seemed convinced that a zergling could 1vs1 an SCV in a direct fight if the zergling was perfectly microed. Like, someone at the level of Shuttle was spending 20-30 minutes clicking the ling, and trying to subtly influence the attack speed of the ling by declaring attacks really precisely. They never succeeded. The SCV always survived and the ling always died. Perhaps it is possible to influence the random factor by supplying some sort of really specifically timed manual input. Definitely not sure about that, though.
do you mean the other way around? a zergling always kills an scv in a 1on1 fight. they usually end up attacking sooner when engaging, but even when I created a save state where they both initially attack each other on the same frame the zergling still always wins and it's not close enough that the RNG can overcome it. the zergling always survives with either 2 or 1 hits remaining.
EPT
