EPTI didn't see a thread on this, at least not with real data, so here it is:
Using QXC's build tester, I very carefully tested whether splitting actually makes a difference. I assumed it would make some small difference (as I think most people assume), but I was wrong. I tested 3 split methods. One method is to have no split, to simply send all 6 workers to mine one patch. One method is the half split, to send all 6 workers to one patch, then quickly select 3 and send them to a different patch. The final method tested is the one I use in real games, to quickly press f1 (select idle worker) and right click an individual mineral patch, to send each worker to its own patch. I also tested whether it matters if you build the first worker before or after sending the starting ones to gather.
I used Protoss in every trial, but the race should have no affect what so ever, because no supply buildings are produced.
The rally point was the same at all times for all trials. I made sure I restarted if I misclicked at all, so that each trial would be within an infinitesimally small margin of error. No chrono boost is used, because the way you split has no effect on your chrono boosts, and it would just add another margin of error.
The effectiveness of each split is measured by the number of minerals at the 1 minute and 2 minute marks.
+ Show Spoiler +
![[image loading]](http://img192.imageshack.us/img192/7595/workersplit.png)
If the tables look identical, they should, because they are. That is no exaggeration, it's real data I carefully recorded from in-game tests.
Yes, 1 and 2 minutes aren't very far into the game, so maybe the difference could pan out later, but really, considering how precisely equal each method is, I don't think the split would ever have any affect on any game. If you aren't seeing even a single extra 5 mineral return after 2 minutes into the game, I don't think you'll ever get an advantage large enough to account for even a single unit at any point through a game.
Thoughts?
edit: Kletus ran some tests of his own:
+ Show Spoiler +
On June 03 2010 07:34 Kletus wrote:
Okay, a couple of things to note:
i) Speed was slowest throughout the whole 1min timer; the timer I used was real-time supplied by the build order tester map VERSION 2.3.
ii) This was timed to the second. By that I mean I stopped looking at the minerals once I saw 1:00 so if an scv was JUST about to put in minerals, he gets cut.
iii) Race used was terran.
iv) For F1 split, I started the split from 6th patch from the bottom. Then just clicked down each patch. 7th scv was placed on the nearest non-occupied patch.
Also, I have noticed a pattern with the F1 selection AI and it can probably be done even more efficiently than just clicking down a line; will investigate this later.
v) For half split, I split the scvs to the patches 2nd from the top and 2nd from the bottom. 7th scv was placed on the nearest non-occupied patch.
vi) For no split, I simply took all my scvs and right-clicked the fifth patch from the bottom, which is the closest and most centered patch on the beta tester map. 7th scv was placed on the nearest non-occupied patch.
These are my findings:
Minerals @ 1min
![[image loading]](http://img.photobucket.com/albums/v621/Cletus/excelyo.jpg)
The difference between half and none is truly negligible.
The difference between F1 and the rest is maybe a couple of milliseconds, just enough for one scv to finish his trip.
Okay, a couple of things to note:
i) Speed was slowest throughout the whole 1min timer; the timer I used was real-time supplied by the build order tester map VERSION 2.3.
ii) This was timed to the second. By that I mean I stopped looking at the minerals once I saw 1:00 so if an scv was JUST about to put in minerals, he gets cut.
iii) Race used was terran.
iv) For F1 split, I started the split from 6th patch from the bottom. Then just clicked down each patch. 7th scv was placed on the nearest non-occupied patch.
Also, I have noticed a pattern with the F1 selection AI and it can probably be done even more efficiently than just clicking down a line; will investigate this later.
v) For half split, I split the scvs to the patches 2nd from the top and 2nd from the bottom. 7th scv was placed on the nearest non-occupied patch.
vi) For no split, I simply took all my scvs and right-clicked the fifth patch from the bottom, which is the closest and most centered patch on the beta tester map. 7th scv was placed on the nearest non-occupied patch.
These are my findings:
Minerals @ 1min
The difference between half and none is truly negligible.
The difference between F1 and the rest is maybe a couple of milliseconds, just enough for one scv to finish his trip.
So there is a difference 40 seconds in, but this difference does not increase over time. So essentially, you get those 5 minerals with a perfect f1 split milliseconds sooner than without, but there is no exponential increase, so you won't see that difference in the 1 min or 2 min results.
Also, some people have mentioned (and I've stated myself) that the map matters. Some maps have obviously split mineral patches that will benefit from worker splits, especially if patches are optimized on the edges and not the middle where you would click for a no-split.
Btw, this thread has nothing to do with BW, or whether or not this is a good thing, it's just about data. I don't think that "requiring" splitting was at all a bad thing for BW, and I also don't think it's at all necessary for SC2, it's something you'll only do once a game, and it just is what it is.
