EPT
also having an average of <400 unspend ressources seems to give you unproportional good results compared to the 400-2000 unspend ressources range even when considering logarithmic scaling
Do you macro like a pro? - Page 45
| Forum Index > SC2 General |
|
Cirqueenflex
499 Posts
also having an average of <400 unspend ressources seems to give you unproportional good results compared to the 400-2000 unspend ressources range even when considering logarithmic scaling | ||
|
Brandish
United States339 Posts
very interesting read. just did my last 15 games and my average SQ is 66, seems accurate | ||
|
DarthLeader
Canada123 Posts
 | ||
|
Jacobs Ladder
United States1705 Posts
| ||
|
Joey Wheeler
Korea (North)276 Posts
| ||
|
Revelatus
United States183 Posts
| ||
|
FlamingTurd
United States1059 Posts
| ||
|
KillerPlague
United States1386 Posts
| ||
|
SDream
Brazil896 Posts
On September 25 2011 12:16 KillerPlague wrote: seriously confused at the game with 150 workers... thats like 50 supply worth of army. also surprised terran make more workers on average than protoss. i would think with mules, they would require the fewest amount of workers. zerg obviously need the most, because they lose them to buildings. Everyone loses some drones to harass I guess, maybe some zerg made mass spine crawlers? It could be. | ||
|
SpaceYeti
United States723 Posts
1) The means in the OP are ROUGH estimates, taking data from single divisions and using small sample sizes. To truly be confident in the estimates of SQ for each league we would require THOUSANDS of subjects from each league. 2) The "ruler" metric in the OP does not show the standard error of those means, which given the small sample size is quite wide. If you are going to compare given the data in the OP, compare to the histograms and note that these are rough estimates constructed from 100 players in a single division of each league. 3) As noted in the OP and by several others, efficient spending is only one of many aspects of good game play, albeit a very important one. This metric does nothing to represent your micro, strategy, scouting, or decision making. These variables and others may significantly affect your overall skill as a player apart from efficient macro. That being said, this metric is excellent for gauging an individual's progress over time as they practice and refine their macro mechanics. If you are going to use this metric, however, I recommend taking averages from blocks of at least 10-20 games, computing means and errors, and using an appropriate statistical test to compared blocks of games. There is a good deal of variability between games from even a single player, as was also noted in the OP, and this will be especially true of players who are practicing their macro mechanics. | ||
|
figq
12519 Posts
(measuring Insane AI isn't worth it, because it doesn't use the same scale for gathering) | ||
|
Nurfie
Sweden24 Posts
Guess i need to work on those injects... | ||
|
Apom
France655 Posts
On September 25 2011 19:42 figq wrote: If I recall correctly, the very hard AI also gets a gathering bonus (6 per regular mineral, insane AI gets 7).This allows to measure the SQ of AI compared to SQ of real people (which doesn't mean the AI isn't terrible in general, only says about its spending efficiency). The SQ of Very Hard AI is around 70-80, which is diamond / low masters. The SQ of Hard AI however is below 40, so low Bronze. (measuring Insane AI isn't worth it, because it doesn't use the same scale for gathering) | ||
|
EtherealDeath
United States8366 Posts
 . Actually my micro is pretty damn abysmal.Great post btw. Edit - I notice my tendency is to have higher and higher IQ the longer the game goes. I'm usually low-mid 80s for things under 15mins, but as the game goes longer (assuming without long periods of positioning back and forth while maxed and thus having resources stockpile) my IQ will go up to the low-mid 90s.). Not sure if this is an intentional phenomenon or if my macro is simply better with longer games. | ||
|
whatthefat
United States918 Posts
On September 25 2011 23:16 EtherealDeath wrote: Hmm just looked at a random 22 minute game I played, U=849, I=1927, IQ =96.37. Makes sense to me, since I play protoss so macro is not that hard, and macro/tactics have always been my strong point. My micro and multitasking not so much . Actually my micro is pretty damn abysmal.Great post btw. Edit - I notice my tendency is to have higher and higher IQ the longer the game goes. I'm usually low-mid 80s for things under 15mins, but as the game goes longer (assuming without long periods of positioning back and forth while maxed and thus having resources stockpile) my IQ will go up to the low-mid 90s.). Not sure if this is an intentional phenomenon or if my macro is simply better with longer games. The SQ is essentially measuring how you spend relative to how other players spend at the same level of average income. What that means is that you are spending relatively better in longer games. In an absolute sense you may be spending worse (since everybody's spending tends to get worse in longer games with higher income), but you are performing at a higher percentile rank. | ||
|
whatthefat
United States918 Posts
On September 25 2011 19:42 figq wrote: This allows to measure the SQ of AI compared to SQ of real people (which doesn't mean the AI isn't terrible in general, only says about its spending efficiency). The SQ of Very Hard AI is around 70-80, which is diamond / low masters. The SQ of Hard AI however is below 40, so low Bronze. (measuring Insane AI isn't worth it, because it doesn't use the same scale for gathering) Fascinating, thank you! How many games did you base those figures on? | ||
|
Khai
Australia551 Posts
| ||
|
NExt
Australia1651 Posts
just wanted to say, appreciate the effort ♥ | ||
|
Zhul
Czech Republic430 Posts
| ||
|
figq
12519 Posts
On September 25 2011 23:07 Apom wrote: You might be thinking GTAI, or similar. (Ex.) Very Hard AI is supposedly the pride of Bob Fitch.. If I recall correctly, the very hard AI also gets a gathering bonus (6 per regular mineral, insane AI gets 7). On September 25 2011 23:42 whatthefat wrote: Using Jeppes spreadsheet, so 20. Results are very consistent, because it's AI. The only huge variation may come from not accepting when the AI gg's, and just keeping it barely alive while its SQ drops.Fascinating, thank you! How many games did you base those figures on? | ||
| ||
