|
On July 05 2010 07:20 LaLuSh wrote:
Skrag, why do you build OL at 11?
I always do the single extractor trick AFTER I've built the overlord, and I imagine most top players do it that way.
good catch, i didn't even notice that :p definitely OL on 10 and extractor trick after it.
also, the results are not entirely conclusive because people often send a drone to scout early on, and people do not have the micro of a computer. i would presume a human starts the 9OL later than the computer by a few seconds, but a human is also worse at the extractor trick.
in conclusion i think every player has to find out for themselves what works best, but the 2 methods are in the end more similar than different.
|
On July 05 2010 07:20 LaLuSh wrote:
Skrag, why do you build OL at 11?
I always do the single extractor trick AFTER I've built the overlord, and I imagine most top players do it that way.
Not sure if the finishing drone timings match exactly, but you get the hang of it. Single extractor trick after OL on 10.
I thought about trying this, but building the overlord after the 11th drone doesn't actually give you an overall economic advantage, instead only giving you a very small window of advantage from drones 11-14, and that advantage all comes from getting drones 10 and 11 faster than 9OL. If you delay the 11th drone, that advantage will shrink, likely making 10-OL-trick inferior to 10-trick-OL in every situation.
For the record though, I'm fairly sure that from the replays I've seen of top players (and I watch a LOT of replays), 10-trick-OL is the standard way of doing the trick.
Of course, the nice thing is that I already have a setup that makes it easy to try, so rather than theorizing, I can just modify the build order and see what comes out.
|
Ok, report from doing the extractor trick after the overlord rather than before:
sd07 0 0
sd08 15.9375 0
fd07 17 10
sd09 26.5625 5
FD08 32.9375 30
sd10 38.9375 0
FD09 43.5625 30
sovl 55 5
FD10 55.9375 10
EXTR 65.6875 45
sd11 66.1875 0
CANC 66.1875 0
FOVL 80 64
sd12 80 64
sd13 80.0625 14
FD11 83.1875 29
sd14 91.1875 34
FD12 97 89
FD13 97.0625 89
sd15 106.313 99
FD14 108.188 114
FD15 123.313 254
369 at 2:15
Over a number of runs, the final mineral result bounced pretty consistently from 369-374, which gives an overall cost of 0-5 minerals, which is better than 10-trick-OL by 5 minerals. If we compare this build to 10-T-OL, we can see the following:
Drone 11 comes out 15.875 seconds later, putting us behind by that many worker-seconds at 83.1874 seconds. We start to gain this back on drones 12 and 13, which are each faster by 9.375 seconds, and that both finish at 97 seconds. So at 104.9375 seconds, this build has caught back up to 10-T-OL, and at 106.375, we're ahead by 2.875 worker-seconds. We gain another .312 seconds on drone 14, and .25 seconds on drone 15, giving us a grand total of 3.4375 worker seconds advantage over 10-T-OL.
BUT....
We already know that both of these builds are behind, or at best equal to, 9OL, which is far easier to execute. The only reason to doing the extractor trick at all is to get an advantage at some point in the 10-15 worker range in order to get some building out faster. 10-OL-T has a smaller advantage over 9OL than 10-T-OL in the 10-13 range, and has a very negligible advantage over 10-T-OL at around the time we hit 15 supply.
So if we want to build in the 12-14 range, 10-T-OL clearly has an advantage in that time range, and if we're waiting til 15 supply (the only time 10-OL-T has an advantage over 10-T-OL), then we should just 9OL.
So, I realize you're a way better player than me, but this all suggests to me that you, and anybody else who is doing the trick after the overlord, should stop, and either 9OL or trick before the overlord, depending on whether you want to build a pool before 14 supply.
|
By the way, just for a frame of reference when we're talking about small numbers of worker-seconds, the conversion rate from worker-seconds to minerals ranges from .67 min/sec to .75 min/sec depending on how close the specific patch being worked is.
So 3.5 worker-seconds is a massive advantage of between 2 and 3 minerals, at best.
This is why, even though extractor tricks seem to gain an advantage in workerseconds if you just compare the drone timings, they actually end up behind in minerals over the entire time range.
Because 10 minerals (what we effectively have to make up in order to be even) is worth at least 13 workerseconds, and the drone building the extractor loses another 3 seconds to the extractor build/cancel, we'd have to be 16 workerseconds ahead at the end of the sequence just to break even, and 10-OL-Trick is only 11 seconds ahead after 15 workers, and 10-Trick-OL is only 8.5 seconds ahead of 9OL, so neither one gains enough worker-seconds to break-even on the cost of 10 minerals for canceling the extractor.
|
Also, just for shits and giggles, I threw in an 8OL build (nobody is doing this, but I was curious).
It is behind 9OL at every point, with a 20-second disadvantage on worker 9, never catches up from there, and ends up being 25 minerals short at the 2:15 mark.
|
Updated the original post with timings and conclusions on doing the extractor trick after the overlord, as well as the ever-so-unpopular 8OL.
|
I agree with LaLuSh. In the OP you compare 10ol ET to both 10 et 11ol and the 9ol. But if your comparing it with just the 10et 11ol, you will find that the overlord before 11th drone is more economic and doesn't waste larvae spawn time. It may be inferior to 9ol but it should be noted as being better than overlord on 11.
|
Thanks for running the tests Skrag. I think I'm going to 9OL in the future.
I always intuitively thought the differences between 9OL and 10-OL-T were negligible, so I didn't really care. But it's true that the execution of 10-OL-T is harder, and there's alot more elements where you can screw up. Especially with the timing of the extractor trick.
Probably settled this discussion once and for all.
|
The one thing you forgot in your tests was a scout (unless you didn't and just haven't mentioned it?), which is sort of crucial. The reason I do 10-OL-ET is because I can use the drone I use for the trick to scout (smaller maps), and that drone will be replaced within a matter of seconds. I have seen several players mention that they do the trick for the same reason.
It's all well and good to know that 9-OL is the most economic when you DON'T scout... But on smaller maps you will always scout before 15 (Steppes for example).
But it is useful information for larger maps where you may not send a scout till 14+, 9-OL being the best in these cases.
|
On July 05 2010 13:46 Swede wrote:
The one thing you forgot in your tests was a scout (unless you didn't and just haven't mentioned it?), which is sort of crucial. The reason I do 10-OL-ET is because I can use the drone I use for the trick to scout (smaller maps), and that drone will be replaced within a matter of seconds. I have seen several players mention that they do the trick for the same reason.
It's all well and good to know that 9-OL is the most economic when you DON'T scout... But on smaller maps you will always scout before 15 (Steppes for example).
But it is useful information for larger maps where you may not send a scout till 14+, 9-OL being the best in these cases.
Well, the point is that unless you're doing *something* with a worker in the small range where tricking gives you a timing advantage, then you shouldn't do it. I didn't mention scouting, but scouting is definitely "doing something", it's just good to be aware of the cost.
In this case, it is probably better to trick after the overlord rather than before, which you've said you're already doing. However, an super-early scout isn't really that necessary on smaller maps IMO. If the info you can get from a scout doesn't change what you're doing, then you've sent it too early, and a super early scout often isn't going to tell you anything actionable, so it might be worth figuring out if you can delay a little for a better economy, because tricking for a scout will increase the overall cost of the scout by quite a bit.
|
On July 05 2010 13:13 P00RKID wrote:
I agree with LaLuSh. In the OP you compare 10ol ET to both 10 et 11ol and the 9ol. But if your comparing it with just the 10et 11ol, you will find that the overlord before 11th drone is more economic and doesn't waste larvae spawn time. It may be inferior to 9ol but it should be noted as being better than overlord on 11.
I don't believe any of these builds waste any larvae. And yes, the discussion comparing the 2 extractor trick methods does say that tricking after the overlord is more economic than tricking before the overlord.
But the point is that it's still less economic than 9OL, and is inferior to tricking before the overlord during the range that the extractor trick actually gives an advantage. So if you want to act during that period of advantage (before 14 workers), then tricking before the overlord is superior, and if you don't, 9OL is superior. Tricking after the overlord is a mix between the two, but it doesn't seem to be good enough in either situation to actually be worthwhile.
|
Really nice way to do the measurements, I like how you tried to get rid of tester bias.
I guess that pros rarely fail the timing of the ET, but for me the fact that 9OL is so much more consistent, is reason enough to use it. Added bonus for being the most economic
On July 05 2010 13:46 Swede wrote:
The one thing you forgot in your tests was a scout (unless you didn't and just haven't mentioned it?), which is sort of crucial.
Then again, you also get your overlord faster with 9OL
I prefer going for an earlier pool, like 12 or 13 pool. In that case I very rarely scout before placing my pool.
From that point on, I feel your economy becomes dominated more by how hard you drone/save and how few lings you get rather than your exact opening.
|
I like to get fast ling speed so I go 10 extractor WITHOUT canceling then overlord and pull 1 drone to gas and then pool at 13 and pull 2 drones to gas when pool is almost done to make up for the 100 gas and research ling speed then pull drones back to mineral. Can anybody tell me if there's another way to get faster ling speed with a higher econ advantage build? I'd like to try. Just too lazy to test myself haha.
|
On July 05 2010 16:02 eScaper-tsunami wrote:
I like to get fast ling speed so I go 10 extractor WITHOUT canceling then overlord and pull 1 drone to gas and then pool at 13 and pull 2 drones to gas when pool is almost done to make up for the 100 gas and research ling speed then pull drones back to mineral. Can anybody tell me if there's another way to get faster ling speed with a higher econ advantage build? I'd like to try. Just too lazy to test myself haha.
Building your extractor just before your spawning pool, and putting 3 drones on gas as soon as its complete, will get you 100 gas just when the pool is finished.
|
tl;dr version:
I tested timings of 10pool up through 14pool using the various openings, and obviously, the fastest pool is going to come when you build the pool before your overlord. So, if you're going to 12 pool and you want it as fast as possible, double-extractor. If you want to pool on 11 asap, you should single extractor trick before the overlord. Duh. Genius, huh? Well, the timings below should give you an idea of how much time you're gaining and how much it'll cost.
But for 13pool and later, where you're going to have to build at least one overlord, 9OL is a pretty clear winner, meaning that the very small timing windows of advantage for extractor tricks simply aren't big enough to be worthwhile after the 13th drone.
------------------------------------------------
The "oh my god how much free time does Skrag REALLY have for fuck's sake" version:
Another update. The timings given in the OP didn't *exactly* show what was best when, so in a major fit of boredom, I manually tested a ton of different pool timings with different methods. I didn't try to automate this one, because that would have given a disadvantage to the double-extractor trick, but here's the testing methodology I used:
Playing on normal speed, which is slow enough to minimize the chance of me missing timings, and restarting every time I felt like I screwed up even the slightest amount, I tested pool timings from 10 pool all the way to 14 pool. For every test, I started a drone, and split the first 6 workers in exactly the same way, which looked identical to the computer's "perfect" split on the 6oclock position on lost temple, which is 3 on each close patch, where the extras peel off to the sides extremely nicely. I then built drones up to the specified number, rallying exactly the same way every time, cherry picking up to 10 drones, to make sure two got on the close patches, and then just leaving the rally point on the center. I tried to have the drone in position just when minerals hit 200 in all cases, and once the pool was started, continued building drones with every larva that popped.
I recorded the finish time of the key drone (drone 10 for 10pool, 11 for 11pool, etc), the start and finish times of the spawning pool, and the supply and mineral count at each point.
Results:
SPAWNING POOL TIMINGS
10Pool
Drone 10 57.25 10/10 110
72.3125 9/10 15
137.313 14/18 75
13 drones + 1 drone 10% done
Overpool (10 drones, overlord, then pool)
Drone 10 57.25 10/10 10
86.0625 9/18 5
151.0625 15/18 165
14 drones + 1 drone 50%
11Pool extractor trick before overlord
Drone 11 69.9375 11/10 129
80.125 10/10 4
145.125 14/18 154
13 drones + 1 drone 50%
11Pool 9OL
Drone 11 90.375 11/18 170
95.5 10/18 20
160.5 16/18 215
15 drones + 1 drone 75%
12Pool w/double extractor
Drone 12 85.3125 12/10 168
85.3125 11/10 8
154.375 15/18 198
14 drones + 1 drone 75%
12Pool w/double extractor OL first
Drone 12 85.375 12/10 68
102.125 11/18 3
167.125 17/18 253
16 drones + 1 drone 75%
12Pool 9OL
Drone 12 90.3125 12/18 120
101.813 11/18 10
166.813 17/18 270
16 drones + 1 drone 75%
13Pool 9OL
Drone 13 96.875 13/18 120
107.5 12/18 5
172.5 18/18 315
17 drones + 1 drone 25%
13Pool double extractor
Drone 13 117.25 13/18 23
108.688 12/18 18
173.688 18/18 273
16 drones + 1 drone 75% + 1 drone 25%
13Pool single extractor before overlord
Drone 13 107.188 13/18 4
107.25 12/18 4
172.25 18/18 284
17 drones + 1 drone 25%
13Pool single extractor after overlord
Drone 13 98.5625 13/18 134
109.063 12/18 19
174.063 18/18 299
17 drones + 1 drone 25%
14Pool 9OL
Drone 14 110.625 14/18 180
112.813 13/18 10
177.813 18/18 360
17 drones + 1 drone 50%
14Pool double extractor trick
115.063 13/18 13
Drone 14 117.125 13/18 28
180.063 18/18 353
17 drones + 1 drone 50%
14Pool single extractor before overlord
Drone 14 110.125 14/18 179
112.625 13/18 4
177.625 18/18 354
17 drones + 1 drone 50%
14Pool single extractor after overlord
Drone 14 110.063 14/18 179
113.438 13/18 4
178.438 18/18 359
17 drones + 1 drone 50%
Obviously, builds that got the requisite number of drones first, before an overlord (single extractor trick for 11 pool, and double for 12 pool), get the pool up quite a bit faster. Some comparisons:
10 POOL
Never never never ever ever overpool. Extractor trick 11-pool gets the pool out 6 seconds *faster*, and wastes almost exactly the same amount of larvae.
11 POOL
Unsurprisingly, a single extractor trick gets the pool out 15 seconds faster than 9OL to 11 drones. (which is actually kinda hard to remember to do right, and probably not smart, because 9OL has you starting drones 11 and 12 at the exact same time normally, when the overlord pops) If you're 11 pooling, speed is probably all that matters, and the cost doesn't appear to be too severe.
12 POOL
12 pooling with a double extractor trick beats getting to 12 drones via 9OL by just over 12 seconds. But it does waste larvae, and therefore your drone count will be lower, and your overall minerals lower by a fair amount. If getting a queen or zerglings 12 seconds faster is important, and worth the cost in delayed economy, double-extractor then pool then overlord is the way to go here.
I also tested a 12pool build I've seen TheLittleOne do a fair amount, which is double-extractor, overlord, then pool. I'm not sure why he does this, because this build is inferior to 12pooling via 9OL in every way. The pool comes about the same time (half a second slower actually in my tests, but as I already mentioned, just the way the drones move around the hatchery can make at least that much of a difference), but is about 20 minerals short. Zero speed benefit, for a decent cost.
13 POOL
I tested a *lot* of ways with a 13 pool, because this is the area where it's somewhat unclear which is best, because this is the time range where extractor tricks are starting to lose their advantage over 9OL.
9OL was clearly the winner economically, ending up with the best drone and mineral count, as well as being faster than everything except a single extractor trick before overlord, which came in only *slightly* faster (half a second), and had a clear cost associated with it.
14 POOL
Again, the single extractor trick before overlord came in *slightly* faster than 9OL (a fraction of a second again), but with a cost of 10-15 minerals, more than I personally would be willing to pay for a fraction of a second.
So again, 9OL is a pretty clear winner.
Just for reference sake, I measured 6pool through 9pool using the same methods, except that I only cared about the pool finish time. If you're pooling this early, you're doing it to rush, and you already know you're various levels of screwed economically, so the finish time is the only thing that matters.
6POOL: 106.75
7POOL: 114.125
8POOL: 122.438
9POOL: 130.188
|
Always done 9 OL because I heard it was good lol, but thanks for proving it
|
Thanks for doing such comprehensive testing on this. I just got the beta two days ago and was completely lost what opening was the best. 9OL from now on for me
|
This is impressive. Skrag I don't see where you account for larvae production and availability over time. Would there be a quick way to run these tests again and find when the larvae is produced and exhausted in the sdXX FDXX OLxx method? Of course we can infer the larvae used from your data, but it wouldn't account for the larvae production or availability at any given times.
If you want to distribute the work load you could share the map with TL too as someone else suggested earlier.
|
So,
From all the builds i like:
12Pool w/double extractor
13Pool 9OL
14Pool 9OL
I think 13 pool might catch up economically but i am not sure about 12Pool
Can you please test this builds for the same time and compare? or maybe until you build your expansion. like 12Pool Queen Hatch
let`s say we want as strong build so after pool you will star queen and build drones
I wonder ho much does 12Pool costs you maybe we can always do 12Pool and if we scout no rush push drones to keep up or lings to kill him
|
On July 14 2010 03:58 projectdp wrote:
This is impressive. Skrag I don't see where you account for larvae production and availability over time. Would there be a quick way to run these tests again and find when the larvae is produced and exhausted in the sdXX FDXX OLxx method? Of course we can infer the larvae used from your data, but it wouldn't account for the larvae production or availability at any given times.
If you want to distribute the work load you could share the map with TL too as someone else suggested earlier.
For the pool timings, where I 13pooled all the various ways for example, no build was ever left with any available larvae at the recorded end time, although many of them would have had one pop out shortly after the pool finished. 9OL never wastes any larvae, no matter where you build the pool (unless for some reason you wanted to 9OL 9Pool, but that would be silly), but any build that puts pool before overlord is going to waste some.
None of the builds ever wasted larvae after the overlord finished, so it probably wouldn't be that big of a deal to go through 10-12pools and see what larvae get wasted where.
If you're looking at the original list, the only build that wasted any larvae was 10OL, and you could just look at the drone start times to see when larvae pop, since after the first few, every larva was used as soon as it popped out. Of course, I didn't build a pool at any point in the original list either, since the whole point was to try to figure out what periods of time the various starts had advantages over each other.
As for distributing the map, it's just lost temple with like 2 triggers to print out text messages when things finish.
|
|
|
|
|
|
EPT
