EPTScrapped as outdated
+ Show Spoiler +
What I have figured so far.
1. It is NOT maximum APM. It DOES sometimes get lower at the start of the game. But as the game goes on for at least 10 minutes it basicly never decreases (with one huge exception).
2. In long team games (not sure about how long but 10 minutes is not long enough but 20 is) after at least one player leaves the game ALL players take a huge APM drop. The most common thing to notice would be a drop from 250s to 150s. After another player leaves another drop occurs. On average from 150s to 100-110s. Also there is no drop when the last player from the team leaves the game (since the game ends, naturally) so there is no drop in 1v1s.
Update: ok not all players take a drop, only the ones who stay. And if you were the first to leave the game there will be probably no drop in game stats for you at the score screen (not confirmed). Also the drop is a bit disproportional.
3. APM sometimes rises even after some low-APM periods (current APM is lower than average APM) which doesn't make sense.
Some ideas as to what has changed:
1) edit: proved wrong by Befree.
Normal time instead of Blizzard's time which alone would increase APM by 37(?)%
2) edit: proved wrong by TheRabidDeer.
Camera movements are now taken into account.
3) APM now counts not all actions in the game but only in the most intensive periods of the game.
4) edit: proved wrong by ALPINA. Has been around for a long time.
Holding down production keys, especially for zerg, now counts as different actions.
1. It is NOT maximum APM. It DOES sometimes get lower at the start of the game. But as the game goes on for at least 10 minutes it basicly never decreases (with one huge exception).
2. In long team games (not sure about how long but 10 minutes is not long enough but 20 is) after at least one player leaves the game ALL players take a huge APM drop. The most common thing to notice would be a drop from 250s to 150s. After another player leaves another drop occurs. On average from 150s to 100-110s. Also there is no drop when the last player from the team leaves the game (since the game ends, naturally) so there is no drop in 1v1s.
Update: ok not all players take a drop, only the ones who stay. And if you were the first to leave the game there will be probably no drop in game stats for you at the score screen (not confirmed). Also the drop is a bit disproportional.
3. APM sometimes rises even after some low-APM periods (current APM is lower than average APM) which doesn't make sense.
Some ideas as to what has changed:
1) edit: proved wrong by Befree.
Normal time instead of Blizzard's time which alone would increase APM by 37(?)%
2) edit: proved wrong by TheRabidDeer.
Camera movements are now taken into account.
3) APM now counts not all actions in the game but only in the most intensive periods of the game.
4) edit: proved wrong by ALPINA. Has been around for a long time.
Holding down production keys, especially for zerg, now counts as different actions.
Important UPDATE:
The latest idea was inspired by lolphind and developed by grigorin. There is some kind of the average APM inflation which occurs at specific moments in time. Effectively at those moments a multiplier to average APM is added with cumulative effect.The most prominent ones are at 4:15 (255 seconds), 8:31 (511),17:03 (1023), and probably 34:07 (2047). Usually at that time average APM is stable enough to see big swings and I also suspect the % value of APM inflation is higher (15-20%) than during earlier ones. Earlier moments of APM inflation are at 2:07 (127 seconds), 1:03 (63), 0:31, and maybe 0:15 and 0:07 but APM is too volatile at those points and the % of APM increase is probably less than at later periods. So basicly if the the game lasted 20 minutes the average APM will be equal to real average APM multiplied by all the multipliers at 0:31, 1:03, 2:07, 4:15, 8:31, 17:03 and maybe by earlier multipliers which are not yet confirmed. So while we don't know the values of multipliers, if they are fixed or not, it's not that important. The important part is that we finally understand the mechanism of APM inflation.
For those who haven't noticed I'd like to point out the almost binary sequence of the timer count of APM multipliers:
31=32-1
63=64-1
127=128-1
255=256-1
511=512-1
1023=1024-1
An important post by grigorin
+ Show Spoiler +
On February 25 2013 15:06 grigorin wrote:
ok i give up...
test conditions: 50 or 100 commands in first or second minute of game
what i found out is:
- increase in avrg apm (despite 0 actual apm for a long time) at timer count: 63 (1:03) ,127 (2:07),255 (4:15),511(8:31) maybe 1023 (17:03) looks like some kind of overflow error (if not purposely apm buff implemented by blizzard)
- avrg apm update almost always (99%) on odd timer count during periods of 0 apm
- I have no clue how blizz calculates avrg apm ^_^
here some of my testdata
+ Show Spoiler +
sc2gears: 1 select drone + 99 move commands in first minute (last command 0:42)
timestamp; shown avrg apm; Limit – 1; Limit – 0.5; Limit 0; Limit + 0.5; Limit +1;
51 119 100.3 100.7 101.2 101.6 102.0
53 118 103.4 103.8 104.2 104.7 105.1
54 113 100.8 101.3 101.7 102.2 102.6
57 111 104.5 105.0 105.5 105.9 106.4
62 107 109.5 110.1 110.6 111.1 111.6
63 113 117.6 118.1 118.7 119.2 119.7
65 101 108.3 108.9 109.4 110.0 110.5
69 95 108.1 108.7 109.3 109.8 110.4
75 93 115.0 115.6 116.3 116.9 117.5
79 81 105.3 106.0 106.7 107.3 108.0
85 77 107.7 108.4 109.1 109.8 110.5
95 73 114.0 114.8 115.6 116.4 117.2
103 62 104.7 105.6 106.4 107.3 108.2
109 57 101.7 102.6 103.6 104.5 105.4
127 59 122.8 123.8 124.9 125.9 127.0
131 54 115.7 116.8 117.9 119.0 120.1
139 50 113.5 114.7 115.8 117.0 118.2
151 49 120.8 122.1 123.3 124.6 125.8
159 44 114.0 115.3 116.6 117.9 119.3
163 43 114.1 115.5 116.8 118.2 119.5
171 40 111.2 112.6 114.0 115.4 116.9
191 38 117.8 119.4 121.0 122.6 124.2
207 33 110.4 112.1 113.9 115.6 117.3
219 30 105.9 107.7 109.5 111.3 113.2
255 33 136.0 138.1 140.3 142.4 144.5
263 31 131.5 133.7 135.9 138.1 140.3
279 27 120.9 123.2 125.6 127.9 130.2
319 26 132.9 135.6 138.2 140.9 143.6
limit means upper and lower limit of total apm so that the shown avrg apm can be calculated (avrg apm = total apm / timestamp * 60)
3 different limit-pairs (-1,0 or -0.5,+0.5 or 0,+1) depending on how the result is rounded or cut off
ok i give up...
test conditions: 50 or 100 commands in first or second minute of game
what i found out is:
- increase in avrg apm (despite 0 actual apm for a long time) at timer count: 63 (1:03) ,127 (2:07),255 (4:15),511(8:31) maybe 1023 (17:03) looks like some kind of overflow error (if not purposely apm buff implemented by blizzard)
- avrg apm update almost always (99%) on odd timer count during periods of 0 apm
- I have no clue how blizz calculates avrg apm ^_^
here some of my testdata
+ Show Spoiler +
sc2gears: 1 select drone + 99 move commands in first minute (last command 0:42)
timestamp; shown avrg apm; Limit – 1; Limit – 0.5; Limit 0; Limit + 0.5; Limit +1;
51 119 100.3 100.7 101.2 101.6 102.0
53 118 103.4 103.8 104.2 104.7 105.1
54 113 100.8 101.3 101.7 102.2 102.6
57 111 104.5 105.0 105.5 105.9 106.4
62 107 109.5 110.1 110.6 111.1 111.6
63 113 117.6 118.1 118.7 119.2 119.7
65 101 108.3 108.9 109.4 110.0 110.5
69 95 108.1 108.7 109.3 109.8 110.4
75 93 115.0 115.6 116.3 116.9 117.5
79 81 105.3 106.0 106.7 107.3 108.0
85 77 107.7 108.4 109.1 109.8 110.5
95 73 114.0 114.8 115.6 116.4 117.2
103 62 104.7 105.6 106.4 107.3 108.2
109 57 101.7 102.6 103.6 104.5 105.4
127 59 122.8 123.8 124.9 125.9 127.0
131 54 115.7 116.8 117.9 119.0 120.1
139 50 113.5 114.7 115.8 117.0 118.2
151 49 120.8 122.1 123.3 124.6 125.8
159 44 114.0 115.3 116.6 117.9 119.3
163 43 114.1 115.5 116.8 118.2 119.5
171 40 111.2 112.6 114.0 115.4 116.9
191 38 117.8 119.4 121.0 122.6 124.2
207 33 110.4 112.1 113.9 115.6 117.3
219 30 105.9 107.7 109.5 111.3 113.2
255 33 136.0 138.1 140.3 142.4 144.5
263 31 131.5 133.7 135.9 138.1 140.3
279 27 120.9 123.2 125.6 127.9 130.2
319 26 132.9 135.6 138.2 140.9 143.6
limit means upper and lower limit of total apm so that the shown avrg apm can be calculated (avrg apm = total apm / timestamp * 60)
3 different limit-pairs (-1,0 or -0.5,+0.5 or 0,+1) depending on how the result is rounded or cut off
