EPTThe effect of larvae-mineral distance on Zerg. - Page 7
| Forum Index > SC2 General |
|
Randomaccount#77123
United States5003 Posts
| ||
|
selaas
Norway40 Posts
On December 02 2011 06:52 FreeTossCZComentary wrote: You really think that first move is OP? Having first move can be even problem actually! Imagine 2 guys runing into minefield... one who runs in sooner is in worse position than one who runs in faster... I guess this is off topic, but I guess I'd redirect you to this regarding the 'First-move advantage' in chess: + Show Spoiler + The first-move advantage in chess is the inherent advantage of the player (called White) who makes the first move in chess. Chess players and theorists generally agree that White begins the game with some advantage. Statistics compiled since 1851 support this view, showing that White consistently wins slightly more often than Black, usually scoring between 52 and 56 percent. White's winning percentage[1] is about the same for tournament games between humans and games between computers. However, White's advantage is less significant in rapid games and in games between weaker players. - Wikipedia | ||
|
Duban
United States548 Posts
On December 02 2011 06:22 1st_Panzer_Div. wrote: Nice OP; but you have a serious flaw. Part A you just took self-admitted 'rough' numbers; despite making calculations based on a quarter of an in-game second, hopefully you can see the flaws in that; especially with what is below: Part B Standard Deviation! Below are my calculations based on your numbers, broken down in a uniform measurement of seconds. Again these are your numbers! Top Deviation Calculation: X1 = 386 X2 = 389 X3 = 392 Mean = 389 (heh, easy one) Deviation = Root ( ((Mean – X1)^2 + (Mean – X2)^2 + (Mean – X3)^2)/ Number in Sample) = Root( ( (389-386)^2 + (389-389)^2 + (389-392)^2 ) / 3 ) = 6 Bottom Deviation Deviation = 6.2 So your claim is that there is a 6 second difference between mining times at the top and bottom. But the standard deviation for your top and bottom tests are at least as big. This means that if you ran only the top 66% of your data would be 389 seconds, give or take 6 seconds. And 33% of the time it would be MORE than 6 seconds difference. Or phrased this way, running only the top experiment, your data would be 380-392 seconds most of the time; with a large chunk of it outside of even those time ranges. And this is just what the data that you provided says will happen if you alone try experimenting only the top part. The idea behind your experiment is awesome, but you need either a much larger sample size, and likely more practice at it to keep your deviation to a minimum. Because right now I can claim with over 91% confidence (look up confidence intervals if you're curious) that based on your data the mining time between the top and the bottom are EXACTLY the same time. Which we know by observation isn't true. A plea to everyone reading this thread; if you have GM or above mechanics, please help the OP out, try the top/bottom mining experiment. Do it 10 times to warm up. Do it 20 times to record the time top. Then do the same thing for the bottom. Give the OP your 20 Top and 20 bottoms (and please be honest) and then he can come back with meaningful numbers. He's got a great idea and even a good start, it's just the numbers he has don't prove anything. (Seriously if like... Sheth did this, you could have pretty fantastic data with maybe even an 80% CI.) There is an equation for a sample and a control, since we have multiple tests, but I can't remember what it was. It would probebly be better but if anyone wants to help me do testing it'd be great. The build is 9 OL 13 scout 15 pool 16 @150 minerals extractor 15 hatchery 14 queen 16 OL 16 zergling 19 lair 22 queen 24 OL 24 creep tumor "first queen does inject ,tumor, inject" 30 hydralisk den 32 overlord 4 zerglings 36 overlord 3 hydralisks P.S. Also, someone else created a model based on my first observation that matched, almost exactly, my result. As for A, I also rounded down my answer to the minimum difference at 2 seconds and still found an 8 mineral difference. | ||
|
Brotatolol
United States1742 Posts
 Nice work testing it and everything, I never thought there was that much of a difference. | ||
|
Randomaccount#77123
United States5003 Posts
| ||
|
-stOpSKY-
Canada498 Posts
NesTea won a GSL with a flawless record. So clearly the "marked effect on a game between two players of equal skill" isnt very drastic and need not be attended to. And a human performing a "highly accurate build" to test such a hypothesis is kind of ironic? Dont you think? | ||
|
Artisian
United States115 Posts
Maybe on most maps you can make up for that eco disadvantage with a rush advantage? | ||
|
rd
United States2586 Posts
On December 02 2011 07:37 stOpSKY wrote: Im not disagreeing with the results you found but I will definitely disagree with this statement; "While the difference between a close larvae-mineral position and a distant one won’t cause a great player to lose to a terrible one it can still have a marked effect on a game between two players of equal skill." NesTea won a GSL with a flawless record. So clearly the "marked effect on a game between two players of equal skill" isnt very drastic and need not be attended to. And a human performing a "highly accurate build" to test such a hypothesis is kind of ironic? Dont you think? SC2 is a game of timings. It's difficult to point out games where this small positional imbalance could affect a game because it's overshadowed, -usually-, by much larger mistakes. When you see on the production one timing slower than the other leading to a huge window for instant victory, it's very easy to simply accept that one timing was faster and there was nothing behind it other than inferior mechanics. These seconds are so precious in defending/executing timing attacks. Although there's no Z, the most obvious example of how having ANY aspect of your build being a little late can get you killed is PvP. It's extremely obvious, warp gate is a huge timing, and delaying your gateway/core/warpgate all add up to delaying it, and the ramifications are immediate when his units are warping in before your warpgate is even finished. Hongun fought Deezer earlier, 4 gated him against a counter 4 gate with sentries and a complete wall-in at his front and Deezer still got rolled. With a late warp gate that hongun noticed- and im sure Deezer knew it too. I don't remember who Sase fought but in code A his WG was behind by over 10 seconds which is kind of dramatic, but still, he straight up died to a 4 gate. Point being, when the seconds add up, and as SC2 gets more and more mapped out, these miniscule imbalances become bigger and bigger as timings become etched into stone. | ||
|
Steel
Japan2283 Posts
| ||
|
Eschaton
United States1245 Posts
On November 27 2011 13:14 Goldfish wrote: IMO they should make it balance by allowing you to control where the larva spawns/patrol. Protoss + Terran can already do this technically (rally their workers, then they'll spawn in the direction you rally). A third rally point from the hatch for larvae would be quite nice. I've accidentally trapped a spawned ultralisk due building placement and wandering larva. | ||
|
chaopow
United States556 Posts
Also, they should allow terran buildings to choose which side they want to put the add-on on. Its just small and unnecessary disadvantages. | ||
|
Randomaccount#77123
United States5003 Posts
| ||
|
Scisyhp
United States200 Posts
| ||
|
Duban
United States548 Posts
Combined new & old data Close mineral-larvae position + Show Spoiler + 6:23 6:20 6:24 6:26 6:23 6:25 Best: 6:20 Far position + Show Spoiler + 6:26 6:29 6:32 6:25 6:27 6:28 6:30 6:31 Best: 6:25 | ||
|
Kharnage
Australia920 Posts
On November 26 2011 04:49 decemberscalm wrote: Someone is ALWAYS behind the other player by default of spawn positions. That is an unfair, undue advantage. Maybe the top spawning player should be thinking defensive. The slghtly stronger economy won't be enough to take away the defenders advantage so in ZvZ it's probbly smarter to aggressive only if you're sure you're better than your opponent or if you have the better spawn position, or if you're july zerg | ||
|
Staboteur
Canada1873 Posts
:D | ||
|
Tektos
Australia1321 Posts
| ||
|
1st_Panzer_Div.
United States621 Posts
On December 02 2011 08:54 Duban wrote: I've done more samples which means more data points. It would be too difficult to upload every replay I have but I do have them saved. I guess i could email the replays to people if they really wanted it and gave a good reason in PM. Combined new & old data Close mineral-larvae position + Show Spoiler + 6:23 6:20 6:24 6:26 6:23 6:25 Best: 6:20 Far position + Show Spoiler + 6:26 6:29 6:32 6:25 6:27 6:28 6:30 6:31 Best: 6:25 I'll gladly run these numbers again when I get home and post the results. Okay I only did the far position; and I actually did it by hand on the bus, but this is much better; Sigma aka Standard Deviation is 2.69 (give or take .02 seconds) If I assume that the close position is the same, at the very least it makes it more likely than not that far position does in fact take longer; I'm guessing 20 results indeed should give a pretty good interval of how many seconds you lose. It is very interesting that this is a much larger difference than I would have thought. | ||
|
XRaDiiX
Canada1730 Posts
| ||
|
Xapti
Canada2473 Posts
All that said, I hate how in brood war the imbalance was/is like 5x-10x as bad as it is in Starcraft 2. On December 02 2011 08:57 Tektos wrote: doesn't fix the problem though, and in fact causes more harm than good in my opinion (splitting larva up across both sides of the building). It's also rather difficult to build buildings where your larva are considering that there's usually at least 1 egg blocking the way.Build a spawning pool / evo chamber / roach warren / any building at the bottom of your hatch and it shifts the larvae around your hatch, closer to your mineral patches. On December 02 2011 07:54 Eschaton wrote: A third rally point from the hatch for larvae would be quite nice. Yeah, but I don't even think they need a third rally. I don't think anyone will at all care that their army units morph on the same side their drones do — it's the worker spawn point that matters. Just having the drone rally point work as the direction for the larva spawn & cluster/rally would work fine in my opinion. It's not like they'd have trouble moving around the drones — they can walk right through them. (it was a problem in BW though) On December 02 2011 06:22 1st_Panzer_Div. wrote: A plea to everyone reading this thread; if you have GM or above mechanics, please help the OP out, try the top/bottom mining experiment. Do it 10 times to warm up. Do it 20 times to record the time top. Then do the same thing for the bottom. Give the OP your 20 Top and 20 bottoms (and please be honest) and then he can come back with meaningful numbers. He's got a great idea and even a good start, it's just the numbers he has don't prove anything. (Seriously if like... Sheth did this, you could have pretty fantastic data with maybe even an 80% CI.) Better to have a computer to it to get the true difference, in my opinion. That said, pros' results could be useful too, since the discrepancy vs a computer and a player could be quite significant, but overall the consistency of a computer makes it an absolutely fantastic tool to measure things like this. | ||
| ||
