Larva disappearing Glitch in 1.5 (not about 20th larva)

First of all, I need to list some basic statistics/mechanics about larvae production to make a common ground to start.
1. At 0min00sec, you start the game with 3 larvae.

2. Once larvae count becomes less than 3 at any point of the game, namely 0, 1, 2, larva(e), Hatchery starts spontaneous larva generation. It takes exactly 15sec for one larva to appear at south side of Hatchery. If total 3 larvae count is not reached, generation continues every 15sec.

3. Hatchery stops generating any more larva when 3 or more larvae exist at the particular Hatchery. Therefore, state with more than 3 larvae can be reached only by Spawn Larva ability from Queen that will be explained below. In terms of competitive game play, this no spontaneous larva generation means your larva production is wasted.

4. Queen has Spawn Larva ability. It costs 25 energy to “inject” Hatchery. 4 larvae spawn from the injected Hatchery at the same time after exactly 40 seconds later. You cannot stack injects. Injected Hatchery still generates larvae just like normal Hatchery, following the same rule above.

5. Queen has initial 25 energy ready to inject. Energy regeneration rate is 0.5625 energy/sec. Therefore, in order for the queen to regenerate next 25 energy, it takes 25/0.5625=44.4444 seconds. Yes, it takes around 45 seconds for next inject, NOT 40seconds contrary to some popular false belief. In perfect world, Queen always has to wait for about 5sec to inject again, not immediately after 40sec Spawn Larva at Hatchery ends.

6. 19 is the maximum number of larvae one Hatchery can hold. Any additional larvae are killed immediately. Therefore, Queen injects are nothing more than waste of 25 energy and your precious APM when Hatchery already has 19 larvae.

7. SC2 engine recognizes less than 1 sec time period. For example, 6min45.2sec is 0.5seconds earlier than 6min45.7sec. This does not directly come into play in this topic, but you need to be aware of this fact when watching replay and writing down the exact timings for similar research.

8. When 40sec Spawn Larva period finish, we don’t get those 4 larvae immediately. It takes about 1-1.5sec for the animation to finish and larvae land on the ground, at which point we can finally use them. 4 larvae don’t even land on the ground at the same time. If you look closely, 4 larvae land one after another with 0.2sec or so apart from each other. So, it is possible that you may only get 1-3 of those if you do famous SDDDDDD a bit too early.

9. For some reason, Unit tab in replay shows larvae count slightly earlier than actual larvae count at Hatchery command card at bottom right.

10. Your APM is limited, so spending 4 larvae from Queen Spawn Larvae takes some time. It is impossible to spend all with no time lag even in the hands of a pro, it was usually 2-5sec for me, so let me assume it takes 4 sec including the animation that I mentioned above to use those 4 larvae. Blue cells in Table A&B below signify this.

11. IMPORTANT: The progress of spontaneous larva generation doesn’t reset when 4 larvae from Spawn Larva pop up. Progress merely pauses. The progress is not obvious to players, yet SC2 engine definitely retains the progress. For example, assume Hatchery generated a larva 10 seconds ago. If 4 larvae spawn at that point, imaginary larva generation progress bar pauses at 10/15. After using those 4 larvae, you only need to wait for the rest of 5 sec to see the next larva generated by Hatchery, not another 15sec to start all over again.

All of the above can be easily tested by anyone who has SC2, so those Rules need no further proof.

Now, with all those Rules in mind, imagine some hypothetical cases in game.

5min59sec Hatchery has 0 larva
6min00sec Hatchery generates 1 larva
6min15sec Hatchery generates 1 larva
use 2 larvae
6min30sec Hatchery generates 1 larva
use 1 larva
6min45sec Hatchery generates 1 larva
“Calling this larva Larva X. This particular Larva X is the center of this topic.”
use 1 larva
7min00sec Hatchery generates 1 larva
Nothing interesting for this no Queen example so far.

With Queen,
5min59sec Hatchery has 0 larva
6min00sec Hatchery generates 1 larva
6min05sec Queen Injects AROUND this time
6min15sec Hatchery generates 1 larva
use 2 larvae
6min30sec Hatchery generates 1 larva
use 1 larva
6min45sec…

How do you think larvae count changes in this case after 6min45sec? With the Rules above, 4 larvae should spawn at 6min45sec, which is exactly 40sec later from the injection. Also, Larva X should appear from spontaneous generation at 6min45sec, which is exactly 15sec later from the last generation. Do we all agree that 5 larva appear at 6min45sec? Is it really what happens? To answer this question, that suspicious “AROUND” 6min05sec injection needs to be addressed. There are 3 different scenarios to this. I classify them as Case E for “E”arlier Larva X than 4 larvae, Case L for “L”ater Larva X than 4 larvae, and finally Case “N” for “N”o Larva X which I consider a glitch.

Case E
5min59sec Hatchery has 0 larva
6min00sec Hatchery generates 1 larva
6min06sec Queen injects
6min15sec Hatchery generates 1 larva
use 2 larvae
6min30sec Hatchery generates 1 larva
use 1 larva
6min45sec Hatchery generates 1 larva = Larva X
6min46sec 4 larvae spawn
use 5 larva
7min04sec Hatchery generates 1 larva
Total 9 larvae so far

In Case E, generation of Larva X is slightly earlier than 40sec Spawn Larva finish. So, it gets generated and 4 larvae subsequently spawn, making larvae count 5. Then, since larvae count is currently higher than 3 for 4 sec (check Rule 10), next larva generation gets delayed by 4 sec compared to no Queen case. Therefore, 7min04sec with 9 larvae is the result.

Case L
5min59sec Hatchery has 0 larva
6min00sec Hatchery generates 1 larva
6min03sec Queen injects
6min15sec Hatchery generates 1 larva
use 2 larvae
6min30sec Hatchery generates 1 larva
use 1 larva
6min43sec 4 larvae spawn
use 4 larvae
6min49sec Hatchery generates 1 larva = Larva X
use 1 larva
7min04sec Hatchery generates 1 larva
Total 9 larvae so far

In this variation, Spawn Larva finishes earlier than Larva X generation. According to Rule 11, progress of the generation pauses at 6min43sec – 6min30sec = 13sec, leaving 2 more seconds to go. 4 larvae get used after 4sec (Rule 10) so at 6min43sec +4sec = 6min47sec, the progress resumes. 6min47sec + 2sec =6min49sec is the time when Larva X appears. Next larva is exactly 15sec later, so 7min04.0sec. Doesn’t this 9 larvae at 7min04.0sec sound familiar? Yes, Case E and Case L results are virtually the same in the end. The tricky part lies in the next Case N.

Case N
5min59sec Hatchery has 0 larva
6min00sec Hatchery generates 1 larva
6min05sec Queen injects
6min15sec Hatchery generates 1 larva
use 2 larvae
6min30sec Hatchery generates 1 larva
use 1 larva
6min45sec 4 larvae spawn
use 4 larvae
No Larva X generated during this time
7min04.0sec Hatchery generates 1 larva
Total 8 larvae so far

Do you see the difference? For Case N, 4 larvae spawning timing is exactly the same as Larva X generation timing. However, unlike Case E and L, no larva gets generated before nor after the 4 larvae spawn. Since larva generation progress only pause, NOT RESET when there are 3 or more larvae (Rule 11), there is no reason why there shouldn’t be any larva until next 7min04sec generation. Case N only produces 8 larvae in total, which is 1 less than Case E and L. Inject timings are L<N<E in chronological order. When Case L and Case E have 9 larvae in total, Case N, whose inject timing is right in between them should also have 9 larvae, shouldn’t it?

My hypothesis is that SC2 engine does calculate and generate the Larva X internally and resets the generation progress bar, which we will never get to see, yet fails to reflect the calculation result of Larva X progress into actual game display. Maybe it is because of the animation it takes for those 4 larvae to land on ground and become available to players (Rule 8). Larva X probably gets mixed into those 4 larvae and lost in void forever. As a matter of fact, Larva X generation timing doesn’t necessarily exactly coincide with 4 larvae spawning timing. As long as Larva X generation timing is less than about 1.0-1.5sec after 4 larvae spawn, then Larva X never appear. Whenever the animation is on at the point when Larva X is supposed to come out, no Larva X comes out according to my observation. Or, as Rule 9 says, there are 2 different larvae count at a particular time during this very short yet sensitive period. Spontaneous larva generation and Spawn Larva ability could be using those 2 different ones to calculate. Whatever the case, this is not how it should work and how Blizzard intended this game to work.

Enough of hypothetical case with no actual replay or proof? Don’t worry, here it is me vs computer just to see larvae count at my hatchery.

http://drop.sc/141266

Experiment starts at 4min00sec and ends at 14min07sec.
In this experiment, I made drones only from spontaneously generated larvae and overlords only from larvae from Spawn Larvae. I kept both at left side of the base. Ignore the number of drones mining and overlords on the right side as they are the ones produced before 4min00sec. Here are the results in tables. Table A is what happened in this experiment and Table B is what I consider should have happened.
+ Show Spoiler +
Spawn Larvae finishing at 6min07sec, 6min55sec and 11min31sec results in Case N where Larva X gets lost (Green in table).
Spawn Larvae finishing at 5min23sec and 10min46sec results in Case E (Pink).
Spawn Larvae finishing at 7min41sec results in Case L (Orange).
Blue color in table shows 2-5 delay to spend 4 larvae to keep Hatchery larvae count lower than 3. (check Rule 10)
Yellow color in table shows what I call a glitch. More than 30seconds from last spontaneous larva generation means one larva, which I named Larva X, is skipped.
At the very end of the experiment, I ended up with 34 dornes and 48 overlords on the left. The fact that I didn’t get 37 drones as in Table B, which is what should have happened and how Blizzard needs to patch, confirms that 3 larvae are definitely lost forever because of the unfortunate larva generation timing. I was very unfortunate, or maybe fortunate in this case to prove my point, to lose 3 larvae in 12 injects. Considering 15sec generation interval and 1.5sec or so for possible glitch period, 1.5/15=10% would be my fair guess for the frequency. So lucky(?) to see 25% chance for the experiment intended to find this out.

Here is replay under patch 1.5.
http://drop.sc/234257

Done in similar way to the original experiment above. Experiment starts at 4min00sec and ends at 18min20sec.

Result in table
+ Show Spoiler +
Spawn Larvae finishing at 12min48sec and 16min31sec results in Case N where Larva X gets lost (green in table).
This proves that patch 1.5 DID NOT address this issue. The glitch still exists in game.

The following Chart E shows Larva X timings for all Cases explained above. Note that in previous sections, Spawn Larva timings varied, but in this Chart E, Spawn Larva timing is set and instead Larva X timings are variables so that it is clear what “early” and “late” Larva X means.

For those of you who don’t believe from a replay of me vs computer at off-line setting and think that it’s only my computers problem, I have examples in professional game replays. In order to add authenticity, I chose 3 important matches of this and last month with easy access to replays.
Game 5 of Stephano vs Polt on Korhal Compound at Assembly Winter 2012 Final
Game 6 of DRG vs MKP on Antiga Shipyard at MLG Winter Arena Championship
I chose both of those because those were the very last matches of the Final in respective tournaments. I also chose another short one.
Game 2 of DRG vs MKP on Daybreak at MLG Winter Arena Championship
You may download replays from websites below.
Assembly
http://tournaments.peliliiga.fi/winter12/tournaments/view/asus-rog-starcraft-ii-tournament
MLG Winter Arena
http://www.majorleaguegaming.com/news/vod-and-replays-from-winter-arena-now-available-to-all

Stephano vs Polt
Macro hatchery at main
12min24sec larva generation
12min38sec 4 larvae
12min58sec next larva generation
Larva X is missing at 12min43sec

DRG vs MKP on Antiga Shipyard
Main hatchery
9min08sec larva generation
9min22sec 4 larvae
9min42sec larva generation
Larva X is missing at 9min27sec

3rd Hatchery
11min50sec larva generation
12min05sec 4 larvae
12min27sec larva generation
Larva X is missing at 12min12sec

Also, Case E where Larva X barely managed to appear earlier happens at 5min35sec at main hatchery in this replay.
Case L where Larva X appears slightly after 4 larvae are used happens at 12min39sec at main, too.

DRG vs MKP on Daybreak
Natual Hatchery
8min34sec larva generation
8min48sec 4 larvae
9min07sec larva generation
Larva X is missing at 8min52sec.

These are the first 3 replays I analyzed and although not too many, all of these replays contained missing Larva X at least once. Did it matter for the game results? I would say absolutely not. Even one of the best players in the world like DRG and Stephano sits with more than 3 larvae at a hatchery quite often after mid game, so I highly doubt this missing 1 or 2 larvae in 10min+ games mattered for anything. I’m sure they don’t know the problem exists either. In fact, I could well be the only one who is aware of this problem in the entire world at the time or writing, or is it too bold of a statement.

Ironically, this problem would matter the most in ZvZ. In ZvZ, larvae count is super important as we all know. Imagine 2 players doing the exact same build. If player A was very unlucky and managed to lose 2 Larva X in a row from his first 2 injects while player B didn’t, with everything else being equal, 2 less larvae would be 2 less drones or 4 less lings, provided enough mineral is available. At that early stage of the game, 2 less drones means 5-10 % of total economy. With linear engagement without neither side having good concave, 14lings win vs 10lings any day of the week. However rare it may sound, I find this potentially game-breaking. Even with 400APM and stellar timing awareness, it is practically impossible to intentionally avoid this unfortunate timing. Starcraft is a great game because everything is deterministic so to say. A roach does 16 damage, not 14-18 where there is 20% chance to get 14,15,16,17 or 18 damage respectively depending on your luck as in some other games. I like SC2 high ground advantage over Broodwar one because there is no 50% chance gamble in SC2. However, with this problem in place, luck could come into play. The chance of losing Larva X is probably 10% or less, but we already know that 25% chance is possible with my unlucky me vs computer game on Scrap Station.

If patch hits, ZvP and ZvT balance would shift in favor of Z ever so slightly by something like 0.0000001% or so. With current trend of macro play, number of games where this glitch matters is very small in the first place. I would say this glitch could matter vs prolonged rather successful 2 rax and 4 gate because these happen around the time your first inject finishes and having extra 2 lings does make a difference in killing a bunker or getting last hit against stalker. Having 2 less lings could prevent you from breaking 2 rax repaired bunker contain. It would be painful to think your luck prevented you from killing the bunker by hair’s breadth and see red life bunker getting repaired back to full life, killing your natural. Vs 4 gate, a Stalker, with 1 life remaining due to 2 less lings, that comes back with full shield 1 min later cannot help Zerg either. Having listed those particular situations, these games are fairly rare and this glitch itself is rare enough to change the balance significantly, though it certainly does.

Zerg winrate in ZvZ is apparently 100%, but patch would matter the most in ZvZ. In ZvZ, unit battles happen earlier and first inject finish just in time for 14gas14pool ling speed upgrade research is done. I’m sure many of you have done or being done 15 drones no-more-drones mass ling attack right when speed finishes at least once in your career. (or every game if you are a ZvZ hater and go quick & easy coin flip with this build ^^) In this scenario, you usually have 20 “or so” lings running at enemy base. I haven’t done enough research on this one, but I’m sure having 2 less lings would matter a lot at less than 6 min into the game. Therefore, the patch would favor actual good player and make ZvZ less about luck. ZvZ is coinflippy enough without this glitch after all, so why keep it? :D

Interesting feedback on balance change

Delaying 1 roach for potentially 15sec sounds the biggest balance concern so far. I’m lazy to explore this for now and prefer to do other researches, but I might dig into this if patch doesn’t hit for a long time, in order to force Blizzard to realize potential big problem.

“What? Making Larva X appear as in Table B by reprogramming is the only way to patch this. There is no need to discuss how it should be patched.” I don’t blame you for thinking that way. I had thought so as well until I read the following feedback.
+ Show Spoiler +

This alternative solution caught my attention. Many might say it’s just a QQ against Zerg, but that last part is not what is interesting about this post. Interesting concept is that when there are 2 rules that contradicts each other and there is a need to fix one of them to have 1 universal rule that applies to all situations, this post suggests that the majority rule is the one that needs to be fixed and minority rule must be applied. Fixing 90% so that 10% rules all sounds so weird. Bill Clinton wins 50% popular vote, Ralph Nader wins 1% popular vote, then now we should have Ralph Nader as our President!!! That’s how it sounds like. However, if fixing the glitch was the sole purpose, this approach works, too. Inspired by this post, I recalculated everything in my me vs computer experiment on scrap station and simulated what WOULD it be like if larva generation timer reset instead of pause when 3+ is available at Hatchery. Take a look at Table C below.
+ Show Spoiler +
When larva generation progress reset every time 4 larvae pop from Spawn Larva, I would have ended up with 29 drones, which is 5 larvae shorter than current state. The key is that generating 3 larvae takes 15sec*3+(2~5sec)=47~50sec while minimum Spawn Larvae interval is 44.4sec. Due to the shorter Spawn Larvae interval, only 2 larvae gets generated between 2 injects and 3rd larva’s progress is almost always lost at 13/15. Now, then what if I just timed injects so that 3 larvae always appear between 2 injects? Table D shows this scenario. I would have ended up with 37 drones, which is 37-29=8 more larvae than when inject timing is not timed well as in Table C.

Coincidentally, this 37 larvae is exactly the same as in Table B, which is how I originally considered the only way to patch. Can this be alternative way to patch? For pure game logic reason, I would say yes. However, in terms of practical game play, I must say absolutely not. More natural patch as in Table B is so much more flexible regarding the inject timings. Making injects 1sec earlier or later has almost no impact overall. On the other hand, alternative patch as in Table C&D punishes 1sec earlier injects so severely that as much as 8 larvae can get lost in 10 min time. Terran equivalent of this would be dropping MULE 1 sec too “early,” not late, somehow gives you less resource despite better play by dropping mule on time. Protoss equivalent of this would be chronoboosting Nexus 1 sec too “early,” not late, somehow gives you less probes despite better play by doing it early. Had this rule implemented, keeping track of your inject timings and intentionally delaying it would become the #1 priority in Zerg play, and even DRG level player would have to stare at his hatcheries like some bronze players. Whoever thinks alternative patch is viable even after reading this section must be a true Zerg hater.

Those who hang around at Zerg Strategy Chat.Channel for some advice/opinion, especialy Pikacho for finding right terms.
VoirDire for post that inspired me for this work [D] Finding the optimal macro ZvP opening
MLG/Assembly Organizing for releasing awesome professional game replays
Blizzard for patching it. No matter when it might be, thank you in advance.
All of you who are reading this part. I assume you read it all. Thank you!! Spread the words if you think this needs to be patched. Feedback is much appreciated.