|
|
United States47024 Posts
On October 10 2012 01:40 gaymon wrote:
you're right, changed my posts
edit: the exact number of heroes doesnt really matter because we have the formula which is right for every number of heroes
Right but my point is that the difference between 4950 and 5050 is the difference between 100 heroes and 101 heroes. Which you don't care to be picky about because 100 heroes is already an approximation of there actually being 110 heroes in the game.
|
I didn't mean like literally ALL of them, i meant like all the heroes with possibility of contention. There were only about 10 heroes that matched up against each other since it was pretty obvious which heroes are top tier (from numbers alone) in hte big "a move at lv1" meta
It was immediately obvious that no ranged hero was in contention after a few quick tests showing that even the best ranged hero will only get 1 and a half hits off, which is in no case enough to challenge the top tier "fat" heroes.
|
Care to share how you did your simulations? You could probably plug it in Matlab or something similar but the data input would be overwhelming.
It would be really cool if you could factor in range, passives and if crazy enough, spells!
Nevermore sucks lol.
|
nevermore UP buff nevermore.
|
On October 10 2012 01:45 Heh_ wrote:
Care to share how you did your simulations? You could probably plug it in Matlab or something similar but the data input would be overwhelming.
It would be really cool if you could factor in range, passives and if crazy enough, spells!
Nevermore sucks lol.
I didn't do that, some guy basically spent the time (a couple of hours) amoving the heroes that were obviously top tier in the "amove 1v1 at lv1" meta
|
On October 10 2012 01:47 Kupon3ss wrote:Show nested quote +On October 10 2012 01:45 Heh_ wrote:
Care to share how you did your simulations? You could probably plug it in Matlab or something similar but the data input would be overwhelming.
It would be really cool if you could factor in range, passives and if crazy enough, spells!
Nevermore sucks lol. I didn't do that, some guy basically spent the time (a couple of hours) amoving the heroes that were obviously top tier in the "amove 1v1 at lv1" meta
Well Yoshi said "according to my simulations". Even if he didn't do it himself, a source would have been good. I didn't wanna quote his post because the quote was pretty long. I was referring to him, not you =/
And doing 5995 simulations for 110 heroes is gonna take much more than "a few hours", even if you didn't need to load the map after ever single simulation (or even every 45 simulations).
|
FREEAGLELAND26782 Posts
So if we're about pure tankiness I think it's fair to look at this from a relatively boxed-off perspective. Can't really talk about skills or movespeed. Instead, I'd like to think of level 1 tankiness as simply how much physical damage a hero can take. While this is unrealistic, given that there are too many skills in the game to account for (straight up magic nukes, physical nukes, magic reduction, physical reduction, etc.) I think it's a good measure of assessing basic, stats-granted tankiness. So in my case, we're just looking at base STR (which gives base HP) and base armor (which gives base physical resistance).
# Hero Base STR Base HP Base Armor Physical EHP
1 Spiritbreaker 29 701 5.38 928
2 Ogre Magi 23 587 5.96 797
3 Night Stalker 23 587 5.52 781
4 Ursa Warrior 23 587 5.52 781
5 Geomancer 23 587 5.22 771
6 Pit Lord 25 625 3.68 763
7 Dark Seer 22 568 5.68 762
8 Beastmaster 23 587 4.52 746
9 Naga Siren 21 549 5.94 745
10 Legion Commander 26 644 2.52 741
...
101 Keeper of the Light 16 454 1.10 484
102 Demon Witch 16 454 1.10 484
103 Necrolyte 16 454 1.10 484
104 Shadow Fiend 15 435 1.80 482
105 Witch Doctor 16 454 0.82 476
106 Guardian Wisp 17 473 0.00 473
107 Phoenix 17 473 0.00 473
108 Windrunner 15 435 1.38 471
109 Gorgon 14 416 1.80 461
110 Nerubian Weaver 15 435 0.96 460
|
On October 10 2012 01:45 Heh_ wrote:
Care to share how you did your simulations? You could probably plug it in Matlab or something similar but the data input would be overwhelming.
It would be really cool if you could factor in range, passives and if crazy enough, spells!
Nevermore sucks lol.
heroa['dmg'] = damage of hero A
heroa['dmgred'] = Damage reduction from hero A
heroa['hp'] = HP from hero A
heroa['spa'] = seconds hero a needs to do one attack
attacksa = Attacks Hero A needs to kill Hero b, ceils turns the number into the next biggest integer
secondsa = Seconds hero A needs to kill hero B
dmga = heroa['dmg'] * herob['dmgred'];
dmgb = herob['dmg'] * heroa['dmgred'];
attacksa = ceil(herob['hp']/dmga);
attacksb = ceil(heroa['hp']/dmgb);
secondsa = attacksa * heroa['spa'];
secondsb = attacksb * heroa['spa'];
And then compare if secondsa is bigger than secondsb.
And that for all possible combinations
I also forgot to check if there are draws, I did it now and updated my previous post.
It will only count as a win if the hero actually kills the enemy
|
FREEAGLELAND26782 Posts
On October 10 2012 01:47 Kupon3ss wrote:Show nested quote +On October 10 2012 01:45 Heh_ wrote:
Care to share how you did your simulations? You could probably plug it in Matlab or something similar but the data input would be overwhelming.
It would be really cool if you could factor in range, passives and if crazy enough, spells!
Nevermore sucks lol. I didn't do that, some guy basically spent the time (a couple of hours) amoving the heroes that were obviously top tier in the "amove 1v1 at lv1" meta
Could do this easily. I just calculated for EHP. It's simple to look at effective level 1 DPS as well. Actually I might do that snice I'm bored and have an hour to kill.
Skills are too tough. Theoretically, you just need to choose the level 1 skill that grants either the most offensive burst. But then you have to factor in stun time, movespeeds, the fact that people can kind of kite (Poisonous Gale)... and then it starts becoming DotA and therefore becomes too hard.
|
Oh, doing it in matlab or a similar program, which can easily be done if have a formatted dota hero spreadsheet, doesn't yield accurate results. It was shown in practice (by that 1v1 guy) that it turns out stuff like attack animations and psudeorandom damage spreads, hp regen rates, and other hidden random factors make a difference among the top tier heroes in this, where victory is often by a single autoattack.
I'd dig up the thread but sifting through a year's worth of threads on a poorly organized dota forum with a average of over a thousand new threads a day is.... might as well have somebody go do either own.
Open up dota 2, use console to make heroes, amove them. If my memory serves correctly, top tier "1v1 amove meta" heroes were (in no particular order)
Axe (this u have to test because the hp regen adds significant amounts of ehp in certain cases)
Tree NS DS SB CK Orge (CK has won against every hero due to RNG of attack spread but after tests it was shown that on average it loses out to many of the top tier ones)
In this version, given that's changed and the absence of Pit Lord, SB should be #1 followed by NS
|
United States47024 Posts
If we're talking level 1 tankiness, melee heroes also have an inherent advantage due to how much of a significant effect Stout Shield has at that level.
|
Yoshi- has awakened the inner nerd within me... I feel a desire to attempt to write out a code to do these calculations myself (to integrate as much variables as I can), but alas, I no longer have Matlab installed on my computer =/
|
Katowice25012 Posts
I'm so in love with the idea of running the simulations in matlab, regardless of how stupid the results would be.
|
On October 10 2012 01:53 Kupon3ss wrote:
Oh, doing it in matlab or a similar program, which can easily be done if have a formatted dota hero spreadsheet, doesn't yield accurate results. It was shown in practice (by that 1v1 guy) that it turns out stuff like attack animations and psudeorandom damage spreads, hp regen rates, and other hidden random factors make a difference among the top tier heroes in this, where victory is often by a single autoattack.
Yea, i also have 897 draws which is just too much.
|
Okay i had a stupid mistake, now with only 8 draws
+ Show Spoiler +
Name Wins
1. Balanar 109
2. Azgalor 107
3. Ish\'kafel 107
4. Rikimaru 105
5. Aggron Stonebreaker 105
6. Tresdin 103
7. Strygwyr 102
8. Harbinger 102
9. N\'aix 101
10. Rexxar 101
11. Darkterror 99
12. Barathrum 97
13. Abaddon 97
14. Slardar 97
15. Terrorblade 95
16. Mangix 95
17. Ymir 95
18. King Leoric 93
19. Mogul Khan 92
20. Nessaj 91
21. Rooftrellen 91
22. Atropos 87
23. Rigwarl 83
24. Rattletrap 83
25. Sven 83
26. Ulfsaar 83
27. Bradwarden 82
28. Magina 81
29. Cairne Bloodhoof 81
30. Slithice 79
31. Lucifer 77
32. Magnus 75
33. Pudge 75
34. Lanaya 75
35. Meepo 75
36. Azwraith 74
37. Jakiro 74
38. Gondar 73
39. Kunkka 72
40. Dirge 71
41. Tiny 70
42. Purist Thunderwrath 67
43. Razzil Darkbrew 67
44. Yurnero 67
45. Mortred 66
46. Raijin Thunderkeg 64
47. Leviathan 63
48. Jin\'zakk 60
49. Banehallow 59
50. Eredar 58
51. Aurel Vlaicu 57
52. Traxex 56
53. Shendelzare Silkwood 56
54. Syllabear 55
55. Mercurial 54
56. Furion 54
57. Boush 54
58. Raigor Stonehoof 53
59. Knight Davion 50
60. Xin 50
61. Lesale Deathbringer 50
62. Razor 49
63. Auroth 49
64. Anub\'arak 49
65. Crixalis 47
66. Lina Inverse 46
67. Leshrac the Malicious 45
68. Kaldr 44
69. Demnok Lannik 44
70. Visage 42
71. Luna Moonfang 42
72. Alleria 41
73. Akasha 39
74. Chen 38
75. Jah\'rakal 37
76. Dazzle 37
77. Rhasta 36
78. Rizzrak 32
79. Nortrom 32
80. Darchrow 31
81. Krobelus 30
82. Black Arachnia 26
83. Vol\'Jin 26
84. Morphling 25
85. Rubick 24
86. Puck 23
87. Huskar 22
88. Thrall 21
89. Kael 20
90. Dragonus 19
91. Zeus 18
92. Slark 18
93. Zet 14
94. Pugna 14
95. Kel\'thuzad 14
96. Rotund\'jere 14
97. Viper 14
98. Mirana Nightshade 12
99. Squee Spleen and Spoon 10
100. Aiushtha 10
101. Medusa 8
102. Lion 6
103. Kardel Sharpeye 4
104. Ezalor 4
105. Clinkz 4
106. Rylai Crestfall 4
107. Icarus 2
108. Anub\'seran 2
109. Io 2
110. Nevermore 0
|
On October 10 2012 01:59 Yoshi- wrote:Show nested quote +On October 10 2012 01:53 Kupon3ss wrote:
Oh, doing it in matlab or a similar program, which can easily be done if have a formatted dota hero spreadsheet, doesn't yield accurate results. It was shown in practice (by that 1v1 guy) that it turns out stuff like attack animations and psudeorandom damage spreads, hp regen rates, and other hidden random factors make a difference among the top tier heroes in this, where victory is often by a single autoattack. Yea, i also have 897 draws which is just too much.
mind sharing the code? lol @ wisp beating nevermore
|
On October 09 2012 23:37 Kipsate wrote:
Undying due to his skill
if not its possibly Dark Seer or Night Stalker, due to having the highest base armors in the game.
Since when has anybody more armor lvl 1 than Rikimaru?
|
On October 10 2012 01:58 heyoka wrote:
I'm so in love with the idea of running the simulations in matlab, regardless of how stupid the results would be.
You like matlab? I know all the guys here hate it and prefer scilab for their stuff.
|
On October 10 2012 02:04 Kupon3ss wrote:Show nested quote +On October 10 2012 01:59 Yoshi- wrote:On October 10 2012 01:53 Kupon3ss wrote:
Oh, doing it in matlab or a similar program, which can easily be done if have a formatted dota hero spreadsheet, doesn't yield accurate results. It was shown in practice (by that 1v1 guy) that it turns out stuff like attack animations and psudeorandom damage spreads, hp regen rates, and other hidden random factors make a difference among the top tier heroes in this, where victory is often by a single autoattack. Yea, i also have 897 draws which is just too much. mind sharing the code? lol @ wisp beating nevermore
Well, I quickly wrote it in PHP as that was the only thing i had available,
Until line 65 it is just parsing the playdota hero list.
http://pastebin.com/syWMtNwJ
|
FREEAGLELAND26782 Posts
On October 10 2012 02:06 Mataza wrote:Show nested quote +On October 09 2012 23:37 Kipsate wrote:
Undying due to his skill
if not its possibly Dark Seer or Night Stalker, due to having the highest base armors in the game. Since when has anybody more armor lvl 1 than Rikimaru?
Rikimaru has 5.76 base.
Sol Keeper, Goblin Techies, Ogre Magi, and Naga Siren all have higher base armor.
|
|
|
|
|
|
EPT
