EPT[G] TheCore - Advanced Keyboard Layout - Page 329
| Forum Index > StarCraft 2 Strategy |
|
JaKaTaKSc2
United States2787 Posts
| ||
|
Ninjury_J
Canada408 Posts
| ||
|
somesuchnonsense
United States19 Posts
So from what I can tell, there are a few things wrong with how you have this setup. So, for each total, you are adding the same key keyscore multiple times. How it is now, it looks like you are doing this: KS3weight * (key1 + key2 + key3) + KS4weight * (key1 + key2 + key3 + key4) + KS5weight *(key1 + key2 + key3 + key4+ key5) ... etc= total what you want is: KS3weight * (key1 + key2 + key3) + KS4weight * (key4) + KS5weight * (key5).. etc = total The purpose of using averages to compare is to allow you to compare set containing different number of values. In this case you are comparing sets with the same number of values, so the average does nothing for you. It doesn't hurt, it's just kinda pointless. Also, if you want more of an actual basis for your weights, you could use that data on thecore archive on frequency of ability use in games. Add up each of the abilities that would fall on ability keys 1, 2, and 3 for the "Ability KS3" weight. Then add up all the abilities on AK4 and divide by "Ability KS3" weight. Add all the abilities on AK5 and divide by ""Ability KS3" weight and so on. The CG KS weights could work the same way. This method isn't perfect, but at least there is less of a subjective basis for it. At least that is my first impression. I can't look any further into this tonight, but I could get on tomorrow and help out if you still need it. | ||
|
Ninjury_J
Canada408 Posts
I agree about the weights ideally having a little more empirical basis. Unfortunately the old data is pretty bad: first its pretty meta-dependent (hydralisks were rarely used back then, for example) and second, its taken from games where people were not using anything like theCore which means they didn't have access to the potential tools we have. For example, the tap with theCore (for T and P) does not require changing CGs for military units. In standard, many players will sift between control groups to check production. I would definitely enjoy looking through this with you tomorrow. I should be on at about 10:30am EST. | ||
|
Beedebdoo
130 Posts
On May 15 2014 14:03 somesuchnonsense wrote: Indeed Also, if you want more of an actual basis for your weights, you could use that data on thecore archive on frequency of ability use in games. Add up each of the abilities that would fall on ability keys 1, 2, and 3 for the "Ability KS3" weight. Then add up all the abilities on AK4 and divide by "Ability KS3" weight. Add all the abilities on AK5 and divide by ""Ability KS3" weight and so on. The CG KS weights could work the same way. This method isn't perfect, but at least there is less of a subjective basis for it.  We actually did this to begin with, to create a starting point to create our weights from. The problem with purely using this data is that it's both very old, and is gives data on pro play, not optimal play.Also, by using widely different weights, we can get a more nuanced picture of each layouts weaknesses and strengths. For example, we've tried with no weights, and dividing by 5 from weight to weight. | ||
|
YoTcA
Germany151 Posts
I created a sheet where the suggested changes (weighting of individual keyscores) is added. Also I added a second variant that uses the other finger as first finger. Maybe this could save you some time. The link to the google docs file: https://docs.google.com/spreadsheets/d/1vYkpjABcba24iZCinLeg2QyqwEnHySXubN2uWt7djPE/edit?usp=sharing PM me if you need full access. Greetings | ||
|
Ninjury_J
Canada408 Posts
I'm not sure what I'm looking at here, but I would love to talk about it with you. | ||
|
YoTcA
Germany151 Posts
sure, what programm would you propose to talk about it? edt: I have ICQ, IRC, TeamSpeak installed and hangouts should also be viable. | ||
|
Ninjury_J
Canada408 Posts
| ||
|
JaKaTaKSc2
United States2787 Posts
On May 15 2014 14:03 somesuchnonsense wrote: Here is my brief and tired take on this. I'll look at it further in the morning. So from what I can tell, there are a few things wrong with how you have this setup. So, for each total, you are adding the same key keyscore multiple times. How it is now, it looks like you are doing this: KS3weight * (key1 + key2 + key3) + KS4weight * (key1 + key2 + key3 + key4) + KS5weight *(key1 + key2 + key3 + key4+ key5) ... etc= total what you want is: KS3weight * (key1 + key2 + key3) + KS4weight * (key4) + KS5weight * (key5).. etc = total The purpose of using averages to compare is to allow you to compare set containing different number of values. In this case you are comparing sets with the same number of values, so the average does nothing for you. It doesn't hurt, it's just kinda pointless. Also, if you want more of an actual basis for your weights, you could use that data on thecore archive on frequency of ability use in games. Add up each of the abilities that would fall on ability keys 1, 2, and 3 for the "Ability KS3" weight. Then add up all the abilities on AK4 and divide by "Ability KS3" weight. Add all the abilities on AK5 and divide by ""Ability KS3" weight and so on. The CG KS weights could work the same way. This method isn't perfect, but at least there is less of a subjective basis for it. At least that is my first impression. I can't look any further into this tonight, but I could get on tomorrow and help out if you still need it. Concerning averages we are comparing sets containing different numbers of values. Control groups have 6 data points and abilities have 4. Am I misunderstanding what you mean? | ||
|
Ninjury_J
Canada408 Posts
| ||
|
Antylamon
United States1981 Posts
For example, this is blue's full calculation for total ability KS: KS3weight * (key1 + key 2 + key 4) + KS4weight * (key1 + key 2 + key 4 + key 6) + KS5weight * (key1 + key 2 + key 3 + key 4 + key 6) + KS6weight * (key1 + key 2 + key 3 + key 4 + key 5 + key 6) (KS3weight and KS5weight don't even include KS3 and KS5 respectively) And this is red's calculation: KS3weight * (key1 + key 2 + key 6) + KS4weight * (key1 + key 2 + key 6 + key 7) + KS5weight * (key1 + key 2 + key 3 + key 6 + key 7) + KS6weight * (key1 + key 2 + key 3 + key 4 + key 6 + key 7) (KS7 shouldn't even be included and only KS6weight includes its own KS) | ||
|
PZ31k0nauT
13 Posts
i think you´re going Crazy ^^. At First I Want to say that I love TheCore 2.0. I think it´s superior to TheCore 1.0. I got used to it pretty fast. The split of control groups and abilities is a Great idea. I don´t know if I understand it correctly. The problem you´re trying to solve now is to identify the most efficient split? This leads to my second question. Are key combinations included in the keyscores? I think HerrPfotig got a point there. Right now I think nearly everything is pretty fluent. For me the key h (RRM) is not that Bad if a key on the index finger has to be pressed afterwards. For me this are the most important key combos (RRM): Tapping Macro: CG: 0->J Moving Army around (attack cmd): 8(melee)-> P -> O (main army)-> P -> L(caster)-> P -> H((N,M)harrass)-> P (I like that one ^^) Base Cams: shift(hold down)+J->9->0->H->L Free Cams: ctrl(hold down)+I->P->K Build buildings: (!key combos with 3 keys) Sorry for not helping out with this post, I wish I could, but I wanted to ask my questions and to give feedback. Keep up the good work. ^^ | ||
|
Morik
65 Posts
I'm learning the core for the first time right now, but also haven't starcrafted for like 2 years. If the new version is available soon, I may want to wait for that? Or if I learn & teach myself muscle memory with 1.0, will 2.0 not be that hard to swap to when it is ready? | ||
|
somesuchnonsense
United States19 Posts
On May 16 2014 01:14 JaKaTaK wrote: Concerning averages we are comparing sets containing different numbers of values. Control groups have 6 data points and abilities have 4. Am I misunderstanding what you mean? Yes, but your aren't comparing control groups vs. abilities, unless I am mistaken. You are comparing abililities vs. abilities, CGs vs. CGs, or abilities plus CGs vs abilities plus CGs. Again, it should really hurt any think if red > purple > blue then avgRed > avgPurple >avgBlue. It just isn't necessary. For the keyscore weight thing, perhaps using it just for the abilities and not the CGs might still be close enough to decent? I don't know, probably not the most important thing to worry about at this point. | ||
|
Ninjury_J
Canada408 Posts
Without the ability to properly weight key use, it seems to me that we must admit a high amount of uncertainty in our choice. Because no split seems exceptionally better than the rest in all dimensions, I feel uncomfortable advocating any right now. What is needed is an objective way of measuring key use, and I think previous posters are correct that taking common combinations into account is likely beneficial. In my opinion, the core of TheCore is what we ought to consider most heavily: as previously stated, moving armies, selecting the macro control groups, making units, using base cams, building things (and spreading creep for Zerg) are, intuitively the most important functions. What we don't know is how they relate to each other. Here is what I did last using the current weighting: if we look at the graphs for abilities, 5 and 6 follow the same curve. 4 is slightly different. The same is true for CGs for 7 and 8 (6 follows the same curve as well, but it has different relative values and also I wanted to make 2 comparisons for each parameter). Taking 4A and 6CG, 4A and 7CG, 5A and 6CG, and 5A and 7 CG, I learned that the layouts are indifferent in terms of the ability axis (whether we choose to care about 4 or 5 we obtain the same results). If we care about 6 CG, green wins out slightly over blue, purple and orange. If we care about 7 CG, blue pushes slightly over green. My concerns: 1) obviously the weighing is arbitrary 2) even if the weighing is decent, the results are quite similar (the difference between green and the runner ups in the 6CG mode are very close) This leaves me quite unconfident in terms of which is best. | ||
|
YoTcA
Germany151 Posts
Example: Teal has a high KS 4, because of a bad key4. This bad key4 is also part of KS 5,6,7,8. So all these scores get pretty high, but mostly because of key4. And then you add up all the KS X. So for the 'total' you basicly get: KS1=key1 + KS2=key1+key2 + KS3=KS2+key3=key1+key2+key3 + KS4=KS3+key4=key1+key2+key3+key4 + … In the end its: 8*key1+7*key2+6*key3+5*key4+4*key5+3*key6+2*key7+1*key8 with keyX being the individual keyscore of each key, and the weight factor not taken into account. edt: thanks somesuchnonsense, got that wrong within my text. corrected it. | ||
|
somesuchnonsense
United States19 Posts
| ||
|
HerrPfotig
Germany39 Posts
2. Units have the most abilities, these are the ones that protect your buildings and enable you to make the other player lose all his buildings 3. Buildings allow you to produce units because of the "abilities" of a building 4. CGs should help making 2 and 3 easier This brings me to the conclusion, that Abilities should be more important(because of 2 and 3) than CGs and that the most important CGs should be for Macro(because of 3). TEAL has the best ability rating BLUE the best CGs rating PURPLE has the best total of both BLUE to TEAL: CGs Average: 61,2 Abilities Average: 92,2 Data is taken from the E and I column in the FingerSplitOptions sheet Now I will pick the layouts that are below the average ORANGE, TEAL, PURPLE plus a slight emphasis on abilities TEAL, PURPLE Which one is easier for the anatomy of the hand? (could be PURPLE) edit: typo | ||
|
HerrPfotig
Germany39 Posts
| ||
| ||
