|
France1810 Posts
The hot topic of the month is mineral gathering mechanics in general and comparison between double harvest, HotS harvest and LotV current econmy.
DH seems to be what the so called community wants. 2:1 harvesting is now the root of all evil, even if a lot of things are still not unadressed as suggested by many posters in the various thread about this topic we've had for several months now.
But I don't want to talk about the vespene/mineral balance, the possibly too huge mineral income boost for lower worker counts, mules/chronoboost/injects issues, nor about the expand rythm the game should have.
I'd like to talk about what bugs me since the begining : the data used and the way they're represented, which are basically everything we base our reasoning on.
First of all, the mining rates presented in the charts by Zeromus and others always seemed wierd to me and counter intuitive, exploring the data sheet didn't solve anyting.
Secondly, basics about worker pairing is that workers are paired and I was puzzled about the fact that nowhere we talk about this while we all know that depending on the map workers pairing can me more or less efficient.
So I thought it was good to try and test deeper the difference between the mining models by choosing specific maps and trying to get a better accuracy in the numbers.
Here are the results.
Method :
I simply choosed 3 key points : 8, 16 and 18 workers and had a base mining with these numbers for a period of 3 ingame minutes. Then I calculated the actual mining rate I was getting on this period. (NB: going for more than 3 points would have forced me to to several more games because of mineral field depletion, so I decided to go for these value as they usuall show the basics of the mining efficiency curve for each mode)
Then I choosed 2 maps :
- a very mining friendly map : Overgrowth and its 4 close patches
- an unfriendly one : Coda (cause leZael wants to watch the world burn  )
And run the test for :
1 - normal HotS eco
2 - the mod : "Double Harvesting (Team Liquid Model)" by oGsInca available on EU server
Results :
(all mining rates in mineral per ingame minute, rounded at 0 decimal)
1 - Overgrowth
HotS 8 workers : 370 (the 8 workers are mining the close patches only)
HotS 16 workers : 687
HotS 18 workers : 740
DH10 8 workers : 467
DH10 16 workers : 717
DH10 18 workers : 760
2 - Coda
HotS 8 workers : 313( 2 workers on the close patch rest wherever they want)
HotS 16 workers : 642
HotS 18 workers : 697
DH10 8 workers : 443
DH10 16 workers : 707
DH10 18 workers : 747
What can we conclude from this :
1 - as expected, DH10 doesn't care much about mineral patch proximity (a good thing for mapmakers, prolly )
2 - the data used until now is not accurate enough and needs a real reworking, workers pairing model is not analysed as well as it should be.
3 - For instance, the HotS 4 base vs 2 base 32 workers discussion needs to be rethought : if the 4 bases are like Coda's mains then its 1252 mins/min vs 1284(woah who would be fool enough to expand?!) but if they are like Overgrowth main its 1480 vs 1374, simply because the 8 workers income in Hots is 21% higher on Overgrowth than on Coda.
so this :
On April 22 2015 09:00 Plexa wrote:Three important observations - Same income for both when 32 workers are mining ...
is not true.
In the end, I'd really want the debate to be set on better data and better economics analysis in general. Too many people are jumping on conclusion in favor of one solution or another without going deep enough in the data.
I mean, what if this whole debate was totally secondary in the end, compared to simple map design and mineral placement?
Would'nt it be worth to test these models again before orgainizing tournaments and stuff?
edit : corrected the HotS 18 workers value 740, not 470 
|
|
|
France1810 Posts
Barrin :
Thx a lot for sharing your data, I totally missed your post in the flow of all answers (that why I created this thread in the first place, to have a place dedicated to the data discussion)
thx for pointing the bad typo, HotS 18 workers on overgrowth is 740, not 470 (stupid fingers, one of the many reasons I'm not a GM )
Big respect for the 40 mins test, I look like a scrub now (although I hope the 190+ trips I have with 8 workers during 3 minutes are enough to be accurate enough)
I also didn't know there was such a thing as a highly approved mineral line, learning every day.
Back on topic I see you have 323 at 8 workers for HotS, as opposed to my 370, Ill have to double check this, but if it's true it means that Overgrowth main mineral line is 14.6% which is a deal worth noticing, although I assume if you had paired the closest minerals before going on the other the result would be different. This strategy lead you to just apply the same figures to workers 9-16, while mine was based on trying to put the focus on closest patch and emphasis the difference between 1-8 and 9-16 workers.
For now I'm going to go through your data and see if it gives me more ideas 
|
France1810 Posts
Btw, before I start re-inventing hot water, has anyone got data for the time to patch depletion in HotS, DH10, DH8 yet?
|
United States4883 Posts
If you don't mind, I'd like to change the topic to a less controversial one and perhaps make it more of a collective knowledge thread in terms of pure data. Maybe. I'm not sure we're actually getting to anything new so far.
Gwavajuice: http://www.teamliquid.net/staff/ZeromuS/Economy/relativemineouttime.png
This was in Zeromus's original article which includes HotS and DH10.
|
yeah this should be renamed to "Data Resources about economic systems" or something. It can be useful.
e : and just as I write that... d:
|
France1810 Posts
On April 23 2015 05:16 SC2John wrote:If you don't mind, I'd like to change the topic to a less controversial one and perhaps make it more of a collective knowledge thread in terms of pure data. Maybe. I'm not sure we're actually getting to anything new so far. Gwavajuice: http://www.teamliquid.net/staff/ZeromuS/Economy/relativemineouttime.pngThis was in Zeromus's original article which includes HotS and DH10.
Thx, didn't want to be controversial at all, sorry if it sounded like this. I realize people are very emotional about the subject, I wouldn't want to hurt anybody's feeling. Data is everything though, the more accurate and understood they are, the wmarter we can be.
I knew the Zeromus figures, but I was more interested about the data behind them, because the results don't match what I experienced.
I did a rapid testing on that 9 o'clock echo base mineral line with HotS model. 16 workers fully deplete it in 18:30-18:35. Zeromus has 19:20 which is, from what I could collect, the time needed when you start at 6 workers and don't force pairing on closest mineral patches at all, on a base with 2 close patches. In this case, a Coda vs Overgrowth study would be nice imho.
Anyway the thing that is still not tested is the depletion time for DH8 model which has a very low 16 workers mining rate and thus a much longer depletion time in theroy.
Alas I will be away till next week and won't be able to look into this myself.
edit : you might have known that some charts in Plexa's answer to david kim were not accurate, I didn't, so this is already getting to something new so far, at least for me
|
United States4883 Posts
|
|
|
I think we'll have to ask some guys that have been using "DH8" for more than a year on their mod to get more info.
|
What's the point....
User was warned for this post
|
Aotearoa39261 Posts
Thanks for taking the time to do some research. I don't agree with your conclusion here
the data used until now is not accurate enough and needs a real reworking, workers pairing model is not analysed as well as it should be.
I admit there some assumptions in terms of mineral patch placement, but I don't think your example is evidence for a rework of the data. By your own numbers, in the best case scenario for HotS 2 base vs 4 base we have 1374 vs 1480 minerals/minute. That's two whole extra bases netting you 100 extra minerals/minute IF you are able to get your workers to pair on the close mineral patches. In our most conservative model of DH the difference is at least 300 minerals/minute (DH10 slightly higher) regardless of the mineral configuration (overgrowth will net you even more).
In the best case for HotS there is a marginal advantage to holding more bases with perfect play and at worst there is no advantage. In DH models there is always an advantage to holding additional bases and that advantage in some cases is quite significant (assuming expansions are taken at realistic points in the game).
|
LaLuSh, thank you for taking your time to bring original Starcraft BW into the equation.
It should be noted however, that "Double Harvest" in your graphs is (most likely?) DH 3x3 - that's the one I originally measured - and not the now more common DH 2x5 or DH 2x4.
|
On April 24 2015 14:05 BlackLilium wrote:
LaLuSh, thank you for taking your time to bring original Starcraft BW into the equation.
It should be noted however, that "Double Harvest" in your graphs is (most likely?) DH 3x3 - that's the one I originally measured - and not the now more common DH 2x5 or DH 2x4.
Yup, if that's what you measured then it's DH 3x3.
|
29 Apr 2015: Included Starbow Beta 1.145 for comparison
More data!
I made a new round of measurements, using Barrin's map (modified to include 32 bases). The testing period is 60 minutes.
I tested the following mineral modes
- The Standard
- The DH 2x5
- The old DH 3x3 of mine
- The newest modification of ZeromuS, which I titled TL DH 3x3 to distinguish from mine.
- StarbowBeta mod, v. 1.145
DH 2x4 is directly scaled down DH 2x5 and was not obtained through a separate measurement.
The difference between (TL DH 3x3) and (DH 3x3) is not with the method, but with the numbers. TL DH 3x3 has a bit higher harvest time and a bit lower wait-at-minerals time. This ratio between harvest-time vs wait-time is crucial, changing the shape of the curve.
Minerals per Minute
![[image loading]](http://i.imgur.com/sHVNVdn.png)
As you can see, the TL DH 3x3 drop is much more pronounced than the original DH 3x3. This is visible especially in the 16-32 worker range. Starbow starts together with TL DH 3x3, but then drops interestingly close to DH 2x4. As a result Starbow is most punishing at 12+ worker count.
Consequently, the benefit of having more bases is highest in Starbow, followed by the TL approach:
![[image loading]](http://i.imgur.com/SHLLO3u.png)
![[image loading]](http://i.imgur.com/muo6eRr.png)
![[image loading]](http://i.imgur.com/ImsHH9D.png)
![[image loading]](http://i.imgur.com/z08PBVT.png)
The New Worker Efficiency graph shows that TL DH 3x3 workers have much lower efficiency past the 12th worker. Starbow hits workers even earlier and stronger: workers 9-16 operate at only 40% efficiency!
The graph is very bumpy though. This is probably caused by the differences between sequential and interleaved mining - something that will require further investigation. Sequential vs Interleaved mining introduces a persistent error, which does not diminish when you run long tests.
![[image loading]](http://i.imgur.com/JLW1CoE.png)
|
Added the Starbow into the graphs for comparsion.
I want to give big thanks to Xiphias for helping me track some bugs which prevented me getting Starbow tests earlier.
I wouldn't have made it without you!
|
Those bugs were also present in Starbow, so that helped us fix some of our maps as well :D So I should thank you as well.
Very interesting graphs btw.
|
Guys, Starbow Econ is worth a test. I haven't been able to push a mod though, buggy workers happen all the time!
Would you ever release a semiofficial mod Xiphias?
|
On May 04 2015 22:33 JCoto wrote:Guys, Starbow Econ is worth a test. I haven't been able to push a mod though, buggy workers happen all the time! Would you ever release a semiofficial mod Xiphias?
I think Blizzard would be even less likely to implement Starbow Mining. Blizzard likes to have clean presentation, polish, and ease of use in their games. While it might make the game more dynamic, I think SBow Mining runs so contrary to those goals that they absolutely will not do it.
It also seems like DH gives the benefits of SBow mining without the stupidity, so that's SUPER cool.
However Lalush has convinced me that the overall speed of the game needs to be brought down before we can see the effects, so I do think that we should be using DH 2x4 and also bring in some Starbow-esque tweaks to the race specific macro mechanics.
|
On May 04 2015 22:33 JCoto wrote:Guys, Starbow Econ is worth a test. I haven't been able to push a mod though, buggy workers happen all the time! Would you ever release a semiofficial mod Xiphias?
You mean a more stable version of Starbow (less patches) or an extension mod with only eco?
|
|
|
|
|
|
EPT![[image loading]](/staff/Plexa/dh10/dh10_hots4base.png)
![[image loading]](http://i.imgur.com/n2EMpRT.png)
![[image loading]](http://i.imgur.com/3pix2o5.png)
![[image loading]](http://i.imgur.com/vXQxJVu.png)
![[image loading]](http://i.imgur.com/bp3OQFi.png)
![[image loading]](http://i.imgur.com/CA7CfcJ.png)
![[image loading]](http://i.imgur.com/vY3e3gp.png)
![[image loading]](http://i.imgur.com/1PqWRwM.png)
