This test was done with Terran.
Note that all times are measured using the replay timer, so they are in-game Fast times.
Test 1: 3 SCV per mineral patch:
+ Show Spoiler [Test 1] +
I examined seven of the eight mains/naturals, as well as the two gold mineral patches. I built 3 SCVs per mineral patch, then told them to start mining. I recorded how long it took for the first patch at each expansion to mine out, and how long for the last patch to mine out.
The results are very consistent for the first patch to mine out: 863 (plus/minus probably 2 from judgement calls) seconds across the board for 24 SCVs per 8 blue mineral patches before the first patch mines out.
To finish mining out, the results are less consistent. Top and right mains took 73 seconds longer to mine out the last patch than to mine out the first patch. Bottom main took only 18 seconds longer to mine out the last patch once the first one was gone. Left is somewhere between 82 seconds and 105 seconds (the computer started there, so the base had 1165 minerals per patch when I started. I'm not sure how starting at 1165 instead of 1500 affects this). The naturals were in the same range (I've linked to the replay at the end if you're really curious).
The most interesting finding is one that's actually already been hinted at by Aether: the patches that mine out first are the patches farthest away from the CC. In fact, the very close patches at top and right were the last patches remaining and are the reason those bases take longer to fully mine out.
In the pictures below the first spots to mine out are in green, the last in red:
Conclusions thus far:
3 SCVs per patch gives very consistent mining per base until the first minerals run out (I should look again at the replay sometime to see how many minerals are remaining at each base when the first patch mines out, but I've already spent a while on this and I'm tired of looking at it right now). So Blizz did a good job with making mining in all directions go at the same rate, at least for this saturation.
With 3 SCVs per patch, the farther patches mine out first, and the closer patches mine out last -> worker pathing keeps workers mining from more distant patches more often, and the longer travel time does not matter with this saturation (already seen in Aether's thread).
Replay: http://www.sc2rc.com/index.php/replay/show/2065
The results are very consistent for the first patch to mine out: 863 (plus/minus probably 2 from judgement calls) seconds across the board for 24 SCVs per 8 blue mineral patches before the first patch mines out.
To finish mining out, the results are less consistent. Top and right mains took 73 seconds longer to mine out the last patch than to mine out the first patch. Bottom main took only 18 seconds longer to mine out the last patch once the first one was gone. Left is somewhere between 82 seconds and 105 seconds (the computer started there, so the base had 1165 minerals per patch when I started. I'm not sure how starting at 1165 instead of 1500 affects this). The naturals were in the same range (I've linked to the replay at the end if you're really curious).
The most interesting finding is one that's actually already been hinted at by Aether: the patches that mine out first are the patches farthest away from the CC. In fact, the very close patches at top and right were the last patches remaining and are the reason those bases take longer to fully mine out.
In the pictures below the first spots to mine out are in green, the last in red:
Conclusions thus far:
3 SCVs per patch gives very consistent mining per base until the first minerals run out (I should look again at the replay sometime to see how many minerals are remaining at each base when the first patch mines out, but I've already spent a while on this and I'm tired of looking at it right now). So Blizz did a good job with making mining in all directions go at the same rate, at least for this saturation.
With 3 SCVs per patch, the farther patches mine out first, and the closer patches mine out last -> worker pathing keeps workers mining from more distant patches more often, and the longer travel time does not matter with this saturation (already seen in Aether's thread).
Replay: http://www.sc2rc.com/index.php/replay/show/2065
Test 2: 2 SCV per mineral patch:
+ Show Spoiler [Test 2] +
This test turned out to have more interesting results. I did the same things as in test 1 above, except I used 2 SCVs per mineral patch.
The key data I recorded are when the first patch at each main mined out, as well as the number of minerals remaining at that time. There is possibly something to be learned from studying the mining as the rest of the patches mine out, but I'm concerned with mining a full set of 8 patches here.
I noticed partway through my test that at the right and left mains, the SCVs did not split themselves up perfectly: in each case 1 patch ended up with 1 SCV, and 1 patch with 3 SCVs. I corrected this once I noticed it; it went on longer at the right than at the left. At the top and bottom as well as at all but the right natural the SCVs split to 2 per patch just fine.
The time until mining out the first patch varied by about 30 seconds among the postions that split properly. This is pretty small, and I didn't record enough about when the other patches mined out at each base to make any conclusions from this. The times were around 980 seconds, so having 3 SCVs per patch instead of 2 is a boost of probably about 10% in your mineral income. With 2 SCVs, the closest patches mine out first, so unless I go back and find the minerals per second from 3 SCVs I can't state anything conclusive yet.
The most important data I took was the minerals per second per expo, which I took as the total number of minerals mined until the first patch mined out, divided by the number of seconds until the first patch mined out.
The raw numbers are in the spoiler:
+ Show Spoiler +
The right main, right nat, and left main are probably slightly lower than they should be because of the SCV split problems discussed, but it does not seem likely this is major.
The difference between positions is up to 18 minerals per game-minute (ignoring the gold expos). Small but not perfect.
The gold expansions are about 12% better than the regular expansions. This seems to be because of their mineral placement--gold expansions bring in 42 minerals per full 6 patch trip, while regular bases bring in 40, and that's only a 5% difference. So the fact that the gold expansions have more close patches matters. SCVs not optimally separating seems to have little to no effect on mining efficiency until you mine out at least one mineral patch.
Conclusions: Close patches mine faster with 2 SCVs per patch. Mining direction doesn't immediately appear to correlate with mining speed. Lost Temple has some positional imbalances with respect to mining--hope you get the left side--but it's overall pretty well balanced in that regard.
Replay: http://www.sc2rc.com/index.php/replay/show/2128
The key data I recorded are when the first patch at each main mined out, as well as the number of minerals remaining at that time. There is possibly something to be learned from studying the mining as the rest of the patches mine out, but I'm concerned with mining a full set of 8 patches here.
I noticed partway through my test that at the right and left mains, the SCVs did not split themselves up perfectly: in each case 1 patch ended up with 1 SCV, and 1 patch with 3 SCVs. I corrected this once I noticed it; it went on longer at the right than at the left. At the top and bottom as well as at all but the right natural the SCVs split to 2 per patch just fine.
The time until mining out the first patch varied by about 30 seconds among the postions that split properly. This is pretty small, and I didn't record enough about when the other patches mined out at each base to make any conclusions from this. The times were around 980 seconds, so having 3 SCVs per patch instead of 2 is a boost of probably about 10% in your mineral income. With 2 SCVs, the closest patches mine out first, so unless I go back and find the minerals per second from 3 SCVs I can't state anything conclusive yet.
The most important data I took was the minerals per second per expo, which I took as the total number of minerals mined until the first patch mined out, divided by the number of seconds until the first patch mined out.
The raw numbers are in the spoiler:
+ Show Spoiler +
Min/sec | Position
10.88 | Top main
10.64 | Top nat
10.87 | Left main
10.81 | Left nat
10.77 | Right main
10.75 | Right nat
10.85 | Bottom main
10.58 | Bottom nat
12.15 | Top gold
12.11 | Bottom gold
The right main, right nat, and left main are probably slightly lower than they should be because of the SCV split problems discussed, but it does not seem likely this is major.
The difference between positions is up to 18 minerals per game-minute (ignoring the gold expos). Small but not perfect.
The gold expansions are about 12% better than the regular expansions. This seems to be because of their mineral placement--gold expansions bring in 42 minerals per full 6 patch trip, while regular bases bring in 40, and that's only a 5% difference. So the fact that the gold expansions have more close patches matters. SCVs not optimally separating seems to have little to no effect on mining efficiency until you mine out at least one mineral patch.
Conclusions: Close patches mine faster with 2 SCVs per patch. Mining direction doesn't immediately appear to correlate with mining speed. Lost Temple has some positional imbalances with respect to mining--hope you get the left side--but it's overall pretty well balanced in that regard.
Replay: http://www.sc2rc.com/index.php/replay/show/2128
Test 3: 2 probes per mineral patch:
+ Show Spoiler [Test 3] +
My hypothesis coming in is that probes mine at the same speed as SCVs in SC2. They have the same acceleration and speed and as far as I can tell a Nexus is the same size as an SCV. But good science doesn't take this for granted, so we must test. Same deal as test 2 except I'm Protoss and use probes.
This time the right main and the bottom gold expo had suboptimal probe spreads (3 on one patch, 1 on one, 2 on the rest). The fact that this happened again at the right main--and I know the "slow" patch was the same in both cases--is interesting in and of itself, suggesting that for whatever reason the worker AI does not like mining that one patch. I might get to uploading a pic showing which mineral patch to watch later today; it's the third from the bottom of the formation IIRC.
Data in the spoiler:
+ Show Spoiler +
The starred entries had suboptimal worker spreads for SCVs or probes or both, so the starred numbers are meaningless.
From comparing the data we see that probes mine at a different speed than SCVs at each base. Over the time period I'm considering here (~1000 seconds) I think these differences are meaningful, but I'd have to re-run the test to see what my uncertainty in each run is. My assumption is that it's very small, but the difference between probes and SCVs is also very small so I can't make any real conclusions about the relative mining rates.
As far as gameplay and mapmaking goes, these differences are very small: the largest is 0.046 minerals per second difference, which over the ~1000 seconds I'm measuring adds up to 46 minerals. That's less than a marine or a worker.
From the starred entries in the comparison data we see that worker spread almost certainly does have an effect, and it adds up to about 100 lost minerals or so over ~900 seconds if you have 3 workers on one patch and 1 worker on one patch. You also go from 8->7 patches sooner and take much longer to mine out the final patch. However in the long run this too is a small effect.
Conclusion: Probes and SCVs mine at nearly the same rate. To say any more would require data to see how consistent SCV mining rate is across multiple trials. In practical terms probes and SCVs mine at essentially the same rate.
Replay: http://www.sc2rc.com/index.php/replay/show/2427
This time the right main and the bottom gold expo had suboptimal probe spreads (3 on one patch, 1 on one, 2 on the rest). The fact that this happened again at the right main--and I know the "slow" patch was the same in both cases--is interesting in and of itself, suggesting that for whatever reason the worker AI does not like mining that one patch. I might get to uploading a pic showing which mineral patch to watch later today; it's the third from the bottom of the formation IIRC.
Data in the spoiler:
+ Show Spoiler +
Min/sec | Position | Difference from SCVs
10.87 | Top main | -0.006
10.56 | Top nat | -0.044
10.99 | Left main | 0.121*
10.86 | Left nat | 0.046
10.69 | Right main | -0.086*
10.87 | Right nat | 0.122*
10.84 | Bottom main | -0.011
10.63 | Bottom nat | 0.043
12.12 | Top gold | -0.030
11.62 | Bottom gold | -0.486*
The starred entries had suboptimal worker spreads for SCVs or probes or both, so the starred numbers are meaningless.
From comparing the data we see that probes mine at a different speed than SCVs at each base. Over the time period I'm considering here (~1000 seconds) I think these differences are meaningful, but I'd have to re-run the test to see what my uncertainty in each run is. My assumption is that it's very small, but the difference between probes and SCVs is also very small so I can't make any real conclusions about the relative mining rates.
As far as gameplay and mapmaking goes, these differences are very small: the largest is 0.046 minerals per second difference, which over the ~1000 seconds I'm measuring adds up to 46 minerals. That's less than a marine or a worker.
From the starred entries in the comparison data we see that worker spread almost certainly does have an effect, and it adds up to about 100 lost minerals or so over ~900 seconds if you have 3 workers on one patch and 1 worker on one patch. You also go from 8->7 patches sooner and take much longer to mine out the final patch. However in the long run this too is a small effect.
Conclusion: Probes and SCVs mine at nearly the same rate. To say any more would require data to see how consistent SCV mining rate is across multiple trials. In practical terms probes and SCVs mine at essentially the same rate.
Replay: http://www.sc2rc.com/index.php/replay/show/2427
Conclusions:
-Probes and SCVs mine at essentially the same rate for gameplay and mapmaking purposes.
-Some mineral formations can have the AI split 3 workers to one patch, 1 worker to one patch, and 2 workers to all others. This has an almost unnoticeable effect until the first patch at that base mines out.
-Using 3 workers per patch mines out patches one space farther away than the minimum faster than patches as close as possible.
-Using 3 workers per patch increases your total mineral income per base by about 10-15% (more accurate numbers may come if I return to my 3 SCV per patch data).
-Lost Temple's mineral formations are pretty close to positionally balanced (~300 minerals over 1000 seconds difference at most) but not perfect. The difference between positions is larger than the difference between races.
-Bases will mine out their first mineral patch after approximately 1000 game-seconds (16 minutes and 40 seconds) assuming 2 workers per mineral patch. With 3 per patch this drops to ~890 seconds (14 minutes and 50 seconds).
Things to do:
Test drones? Test 1 worker per patch?