Smash 64 ELO Rating

clubbadubba

Smash Master
Joined
Apr 27, 2011
Messages
4,086
UPDATE: Shears will now be handling all Elo related activities via the website he created. The NA console Elo can be found here:


http://elorankings.com/rankings.aspx?id=4

http://elorankings.com has other rankings present and anyone can create their own rankings as well. If you are interested check it out!


I have created a Java program that will calculate player ratings based on tournament results. The rating system is based on the ELO rating system, which was initially created to rate chess players, but can be used for any head to head competition format. For more information about ELO, you can start at the wikipedia page HERE.

The Logic:

ELO is a rating system for head to head competitions. Each new player starts with a default rating of 1500 (an arbitray number), which will increase as they win and decrease as they lose. The amount of points gained or lost from a game is dependent on the rating of the opponent played. For example, if a 2000 player beats a 1000 player, very little change will occur to either players rating. However, if a 2000 player loses to a 1000 player, both players ratings will change drastically. Players ratings will also change more drastically during the first games that are recorded for the ratings. This allows players to reach a stable equilibrium quickly.


Formulas:

Say we have player A and player B, who have ratings of Ra and Rb, respectively. The formulas are as follows:

Ra(new) = Ra(old) + K * ( W – Ea )
Rb(new) = Rb(old) + K * ( W – Eb )

where Ea and Eb are the expecated probability that player A and player B will win the matchup, respectively. W = 1 if the player won, and W = 0 if the player lost.

Ea and Eb can be calculated by:

Ea = 1/(1 + 10 ^ ((Rb-Ra) / 400) )
Eb = 1/(1 + 10 ^ ((Ra-Rb) / 400) )

K is a "K factor" which takes into account how many games a player has played. The fewer games a player has played, the greater the K factor will be. This allows players to reach a stable equilibrium much faster than if K were a non-changing value. In this rating, K for a particular player can be calculated by:

K = 800/(Ne)

where Ne is the total number of games played by a player in order to achieve this rating. The value of 800 is the recommended value for ELO. In order to assure that no player's K factor gets so low that their rating becomes stagnant, a floor value has been set to 64. This value was chosen to be higher than the standard ELO value of 16 or 32, because there are so few competitve games played. This number can be subject to change in the future. At the current value, a player's rating should not be considered "stable" until they have played a minimum of 13 games.


Scope of the project:

I have started this project out using only console tournaments. Soon, I will expand this project to online competitions as well. I plan to have 3 different ratings list: 1 for console, 1 for online, and 1 for console/online. International play is welcome to submit data as well.


Submitting Tournament Data:

In order to create these ratings I need the tournament data in text file format. I have so far created the text files for Genesis 2, Apex 2012, and Apex 2013, which are probably 3 of the larger tournaments. It would take a long time for me to do all of them on my own, so I am requesting help from anyone who has the time to create a text file representing tournament data. Thanks to Cobrevolution and Banze for submitting tournament data to me all ready for 3 NJ tournaments and 1 Brazilian tournament. I know there are a few more tournaments that have occurred in recent years, and I would appreciate all the help I can get in turning the data into the appropriately formatted text files.


Format:

1) List all of the tournament players names in a vertical column, with no extra spaces at the end of each name. At the end of the list, write "End Of Entrants List" without the quotes.

2) Immediately following the end of entrants line, proceed to list the games in the following fashion: Name of winning player, followed by a tab, followed by the name of losing player, and that is one line. On the next line you would do the same for the next game. When you are doing games, it is important that you list them in the proper order.

i) Skip all byes
ii) Add all of the winners bracket games all the way through the WF, and until that point be sure to add all of the previous rounds games before moving on to the next.
iii) After the winner's games are done, add the losers brackets games, again being sure to add all the games in each round before moving on to the next round. After adding the LF set, add the first set of GF, and then add the second set if there is one.
iv) If doing something more complicated than a double elimination tournament, just be aware that the purpose of the ordering is to make sure people get credit for the rating their opponent is at AT THE TIME THEY PLAYED. So, essentially each player's games need to be in chronological order.

3) Please simplify all entrant names so that they are all in lower case, and if anyone has numbers in their names in place of letters, please replace them with the proper lower case letters. So m3gav01t would be megavolt. And SuPeRbOoMfAn would be superboomfan. I need all names to be consistent across different tournaments, and this is the easiest way to do it.

Example Format:

kikoushi
isai
superboomfan
.
.
.
sandman
End Of Entrants List
bossman shaolinblunts
.
.
superboomfan bossman
.
.
.
kikoushi isai
.
.
.
isai superboomfan
isai kikoushi
kikoushi isai


And TO's, please email your text files to clubba22@gmail.com, and when you do please leave a post here telling me to check, because I do not check that email often.
 
Last edited:

clubbadubba

Smash Master
Joined
Apr 27, 2011
Messages
4,086
Console Ratings:

NORTH AMERICA
1. isai 2671.0 (22,4)
2. superboomfan 2604.0 (38,5)
*3. kikoushi 2597.0 (7,1)
*4. brax 2551.0 (5,1)
*5. moyashi 2535.0 (7,2)
*6. mariguas 2436.0 (6,2)
7. kefit 2401.0 (31,6)
8. ruoka dancho 2387.0 (10,4)
9. tacos 2382.0 (9,2)
10. jaimehr 2372.0 (17,6)
11. ybombb 2359.0 (19,5)
12. banze 2356.0 (14,5)
13. a$ 2346.0 (13,4)
14. kerokeroppi 2345.0 (61,24)
15. mew2king 2343.0 (25,9)
16. fireblaster 2330.0 (32,16)
*17. mr bushido 2310.0 (4,2)
*18. jam 2300.0 (7,2)
19. near 2297.0 (13,3)
*20. nangoku 2290.0 (4,3)
*21. parallax 2286.0 (7,2)
22. sheermadness 2269.0 (9,4)
*23. merihim 2256.0 (4,2)
*24. captain l 2246.0 (5,2)
25. revan 2240.0 (17,6)
26. zenyore 2220.0 (6,4)
27. battlecow 2213.0 (12,5)
28. star king (controller) 2201.0 (9,3)
*29. hero pie 2193.0 (4,2)
30. wizzrobe 2190.0 (14,6)
*31. b link 2180.0 (6,2)
*32. star king 2174.0 (6,2)
33. sensei 2173.0 (55,18)
*34. ballin4life 2172.0 (4,2)
*35. killer 2170.0 (5,2)
*36. dexter 2169.0 (4,2)
37. cc 2165.0 (6,4)
38. nintendude 2158.0 (31,16)
39. cobrevolution 2151.0 (54,33)
*40. king funk 2122.0 (4,2)
41. oxolane 2111.0 (6,4)
*42. meta 2089.0 (4,4)
43. firo 2081.0 (11,6)
44. ichiro 2080.0 (6,6)
*45. b-rex 2070.0 (4,2)
46. han solo 2064.0 (23,14)
*47. snorlax 2052.0 (4,2)
*48. dark link -x- 2048.0 (3,2)
49. clubbadubba 2034.0 (17,13)
50. greginator 2031.0 (15,8)
51. stranded 2030.0 (54,28)
*52. leon 2028.0 (3,2)
*53. moogle 2022.0 (3,2)
*54. gimpyfish 2015.0 (2,2)
*55. ownasaurus 2000.0 (4,2)
*56. castaway 1999.0 (2,2)
57. sextc 1989.0 (8,6)
*58. daddysthunder 1979.0 (5,2)
59. kyletree 1964.0 (6,4)
*60. chu dat 1959.0 (4,2)
*60. robert 1959.0 (3,2)
*62. ld 1933.0 (5,3)
*63. shademoneh 1923.0 (2,2)
64. ceasar 1922.0 (18,16)
65. *** vwls 1921.0 (8,6)
*66. blink 1919.0 (4,4)
*67. jimmyjoe 1918.0 (5,4)
68. megavolt 1915.0 (22,14)
69. jimmy joe 1904.0 (25,19)
*69. bowl 1904.0 (6,2)
*71. rocky 1900.0 (3,2)
*72. phyziqx 1897.0 (2,2)
73. weedwack 1896.0 (27,22)
74. shade 1895.0 (9,7)
75. combo blaze 1894.0 (6,6)
*76. eggmode 1889.0 (5,4)
*77. derek 1880.0 (4,4)
*78. atothez 1878.0 (3,2)
*79. amida 1872.0 (5,2)
80. bladewise 1870.0 (18,13)
*80. madrush21 1870.0 (2,2)
82. jnatty 1862.0 (22,16)
*83. the scrub 1860.0 (2,2)
*84. smashtista 1857.0 (3,2)
85. jel 1855.0 (17,13)
*86. krazy kirby 1848.0 (4,2)
*87. nyoouber 1846.0 (3,1)
*88. roland 1845.0 (4,2)
*89. jonnis 1837.0 (2,2)
*89. facecheekwall 1837.0 (3,2)
91. shears 1832.0 (16,12)
92. rumbly 1831.0 (7,6)
*93. black snow 1830.0 (2,2)
*94. domingo 1828.0 (1,1)
*94. hamburglar 1828.0 (2,2)
96. tom bombadil 1827.0 (6,4)
*97. chris fischer 1825.0 (4,1)
*98. lobo 1822.0 (4,2)
99. valoem 1820.0 (19,14)
*100. butch g 1819.0 (3,2)
101. preston 1817.0 (6,4)
102. madrush 1815.0 (8,8)
*102. jthumiliepauvretoi 1815.0 (3,2)
*104. killmaster 1812.0 (4,4)
*104. stricnyne 1812.0 (2,2)
*106. hectohertz 1811.0 (2,2)
*107. peruvian bone crusha 1809.0 (3,2)
*108. sleeh 1807.0 (2,2)
109. to joe 1806.0 (8,6)
*110. mandino 1804.0 (4,2)
*110. egman 1804.0 (3,2)
*112. pete 1803.0 (3,2)
*113. jordan 1802.0 (2,2)
*114. i love smash 1799.0 (5,2)
*115. down_and_b 1796.0 (3,2)
*116. unreal 1791.0 (5,2)
*117. nainsoo 1790.0 (2,2)
*118. thebawse 1782.0 (4,2)
*119. hipstur 1779.0 (2,2)
*120. roygbiv 1776.0 (3,2)
*121. bubba ray 1775.0 (2,2)
*121. hard body 1775.0 (3,2)
*121. espy 1775.0 (3,2)
*121. chocolatestorm 1775.0 (4,2)
*121. gs-djfliphop 1775.0 (3,2)
*126. smelly cat 1769.0 (4,2)
127. komotonoto 1762.0 (11,7)
*128. tank 1753.0 (2,2)
*128. mrmarbles 1753.0 (2,2)
130. karajan 1751.0 (17,16)
*131. godrage13 1747.0 (2,2)
*132. eggcelsior! 1744.0 (2,2)
*133. whats up tilt 1737.0 (2,2)
134. vex kasrani 1736.0 (6,4)
*135. vice 1735.0 (4,4)
*136. john b. 1734.0 (2,2)
136. justin 1734.0 (5,6)
*138. kamui8899 1722.0 (2,2)
139. skyfire 1716.0 (20,21)
*140. urineight 1715.0 (2,1)
*140. jack 1715.0 (2,2)
*142. runtmc 1714.0 (2,2)
*143. iroh 1713.0 (3,2)
*144. yoshiguru 1707.0 (4,2)
*144. serohk 1707.0 (4,2)
*146. rizner 1705.0 (1,2)
*147. asylum 1702.0 (3,2)
*148. olikus 1696.0 (1,2)
*149. danconia 1695.0 (2,2)
*149. fez 1695.0 (2,2)
151. sawa 1694.0 (7,4)
*152. tait 1689.0 (1,2)
*153. little spoon 1686.0 (1,2)
*154. sicko 1684.0 (3,2)
*155. wenbo 1683.0 (1,2)
*156. jared c 1682.0 (2,2)
*157. colin maresca 1679.0 (2,2)
*158. brendan 1675.0 (3,2)
*159. doctor_jazzy_jay 1670.0 (1,2)
*160. chibo 1667.0 (1,2)
*161. kratos g 1666.0 (2,2)
162. knitephox 1663.0 (13,13)
*163. mc godd 1661.0 (2,2)
*164. lilleswing 1659.0 (2,2)
164. tony 1659.0 (8,6)
*166. kdy 1655.0 (2,2)
167. korobeniki 1654.0 (10,12)
*167. daniel 1654.0 (3,2)
*167. what 1654.0 (2,2)
*170. yeroc 1651.0 (1,2)
*171. gary the snail 1650.0 (2,2)
*172. frantix 1646.0 (1,1)
*172. boxer 1646.0 (3,2)
*174. olis 1645.0 (2,2)
*175. yunque 1642.0 (5,4)
*176. mpintense 1641.0 (2,2)
*177. dimeq 1635.0 (2,2)
*178. jered 1633.0 (2,2)
*179. guitarhero23 1630.0 (2,2)
*180. puff daddy 1629.0 (2,4)
180. breakthrough 1629.0 (15,18)
*182. mohinder 1626.0 (2,2)
*182. dota_box 1626.0 (2,2)
*184. johnstermonzter 1624.0 (2,2)
*185. bossman 1622.0 (1,2)
*186. victor 1621.0 (3,2)
*186. gio 1621.0 (3,2)
*188. zedz 1617.0 (2,2)
189. maliki 1616.0 (19,26)
*190. love 1615.0 (1,2)
*190. callmedaddy 1615.0 (1,2)
*192. 0000000000 1614.0 (1,2)
*193. michael trewartha 1613.0 (2,2)
*194. randy 1612.0 (1,2)
*195. mcg 1611.0 (3,2)
*196. juice 1608.0 (1,1)
*197. gablixion 1607.0 (3,2)
*198. colin 1599.0 (1,2)
*199. chak 1596.0 (1,2)
200. chain ace 1595.0 (7,6)
*201. syrox 1592.0 (1,2)
*202. blitz 1590.0 (4,4)
*202. kmad64 1590.0 (3,2)
204. eliot 1589.0 (6,8)
*205. dvk 1588.0 (1,2)
*206. rockinrudy 1585.0 (1,2)
*206. kaboomachu 1585.0 (2,2)
*208. fatalak 1582.0 (2,2)
*209. cupa joe 1581.0 (1,2)
*210. eikelmann 1579.0 (2,2)
211. yobolight 1575.0 (8,8)
*211. joez 1575.0 (1,2)
*213. ref 1570.0 (3,4)
*214. pseudo 1569.0 (1,2)
*215. rami jackson 1563.0 (1,2)
*216. studstill 1559.0 (1,2)
*217. domingo1 1558.0 (2,2)
*218. tjim 1555.0 (3,2)
*219. mrschmucker 1554.0 (1,2)
*220. spencer wdb 1544.0 (2,2)
*221. nestor 1540.0 (2,2)
*222. chunk 1535.0 (2,2)
*223. phut 1531.0 (1,2)
*224. master morrison 1522.0 (1,2)
*225. hotdiggitydog 1519.0 (2,4)
*225. jomp17 1519.0 (2,2)
*227. dan 1517.0 (1,2)
*227. laser kirby 007 1517.0 (2,2)
*229. master keef 1513.0 (1,2)
*230. brown beard 1512.0 (2,2)
*231. najja 1507.0 (1,2)
*232. justin wong 1503.0 (1,2)
*233. mario marquez 1496.0 (2,2)
*233. a4gg 1496.0 (2,2)
*235. daffy 1495.0 (2,2)
*236. cazcom 1487.0 (2,2)
*237. thugz 1485.0 (2,4)
*238. alexmon27 1481.0 (1,2)
*239. doge 1478.0 (1,2)
*240. seel 1475.0 (1,2)
*241. kevin g 1474.0 (1,2)
*242. ganon 1467.0 (1,2)
*243. db2k 1464.0 (1,3)
*244. armen 1463.0 (1,2)
*245. apple 1461.0 (3,4)
*246. ryan 1458.0 (1,2)
*246. ray rocket 1458.0 (1,2)
*246. vanz 1458.0 (1,2)
*246. ~tac~ 1458.0 (1,2)
*246. ripple 1458.0 (1,2)
*246. duck 1458.0 (1,2)
*246. xaltis 1458.0 (1,2)
*246. spku 1458.0 (1,2)
*254. fssb64 1455.0 (2,2)
*255. foxy rump 1454.0 (1,2)
*256. vvv rapture 1453.0 (3,4)
*257. xjuiceman91x 1449.0 (2,2)
*257. corni 1449.0 (1,2)
*259. erick 1448.0 (1,2)
*260. vatgas 1442.0 (2,2)
*260. george 1442.0 (2,2)
*260. p90curl 1442.0 (2,2)
*263. gildo 1441.0 (2,2)
*264. allcaps 1440.0 (1,2)
*265. joe 1432.0 (1,2)
*266. sam 1431.0 (1,2)
*267. 420blazeit 1429.0 (1,2)
*268. d-von 1428.0 (1,2)
*269. charlie 1427.0 (2,2)
*270. astro 1426.0 (1,2)
*270. prisonchild 1426.0 (1,2)
*272. andrewajt62 1425.0 (1,2)
*273. minikilla13 1424.0 (1,2)
*274. kitten 1420.0 (1,2)
*275. drinkwildturkey 1419.0 (2,4)
*275. singularity 1419.0 (1,2)
*277. mx7 1418.0 (1,2)
*277. yoshifiredragon 1418.0 (1,2)
*279. jetx 1417.0 (1,2)
*280. tim rich 1403.0 (1,2)
*281. triptych 1398.0 (1,2)
*282. fox 1392.0 (1,2)
*283. zwarm 1390.0 (1,2)
*284. magica 1388.0 (1,2)
*285. kevin potter 1380.0 (2,2)
*286. kuraudo 1379.0 (1,2)
*287. hazard2kv 1376.0 (1,2)
*288. sandman 1375.0 (1,2)
*289. tweed 1373.0 (1,2)
*290. gunblade 1371.0 (2,4)
*290. thebuicklives 1371.0 (1,2)
*292. hammertime 1370.0 (2,4)
*293. the bees knees 1368.0 (1,2)
*293. paul lozbin 1368.0 (1,2)
*295. hazmat matt 1367.0 (1,2)
*296. kradi 1365.0 (2,2)
*297. connor walsh 1360.0 (1,2)
*298. doctor tom 1356.0 (2,2)
299. chain-ace 1352.0 (15,22)
*300. awesomesaucer 1346.0 (1,2)
*300. jess 1346.0 (1,2)
*302. leo 1342.0 (1,2)
*303. black cats31 1326.0 (3,4)
*304. mitch wdb 1321.0 (1,2)
*305. jackie evella 1319.0 (1,2)
*306. low-tier-or-no-tier 1311.0 (1,2)
*307. vudujin 1308.0 (1,4)

BRAZIL
1. banze 2356.0 (14,5)
*2. cabelo 2248.0 (3,2)
3. oxolane 2111.0 (6,4)
*4. aprendiz 1997.0 (2,2)
*5. danwear 1798.0 (2,2)
*6. hokage 1767.0 (2,2)
*7. dede 1611.0 (1,2)
*8. onizukasensei 1522.0 (2,2)
*9. joker 1496.0 (2,2)
*10. abuba 1373.0 (1,2)
* - Player has not played enough games to stabilize rating


Online Ratings (not up to date):


NORTH AMERICA
1. killer 2446.0 (15,1)
2. star king 2295.0 (24,4)
*3. dexter 2267.0 (4,2)
4. ld 2196.0 (11,2)
5. fireblaster 2190.0 (8,2)
6. sheermadness 2085.0 (8,2)
*7. near 2065.0 (2,1)
8. battlecow 1939.0 (21,14)
9. fays 1937.0 (6,4)
*10. sextc 1935.0 (3,1)
11. korobeniki 1934.0 (6,5)
12. a$ 1921.0 (7,5)
*13. chaosx522 1900.0 (1,0)
14. greginator 1895.0 (8,5)
*15. bane 1872.0 (1,1)
*16. loto 1850.0 (3,6)
17. reborn 1799.0 (6,4)
18. nova 1789.0 (8,6)
*19. stacked crooked 1650.0 (1,2)
20. kimimaru 1649.0 (4,10)
21. wilde 1606.0 (5,9)
*22. dc-nerd 1584.0 (4,3)
23. hipstur 1572.0 (4,10)
*24. twilight sparkle 1565.0 (1,0)
25. combo blaze 1486.0 (7,12)
*26. mahie 1484.0 (2,3)
*27. bumblebomb 1481.0 (2,3)
*28. mrmarbles 1437.0 (2,2)
*29. jesusv2 1422.0 (2,2)
*30. mint 1381.0 (1,1)
31. nyoouber 1356.0 (2,8)
*32. lulz 1283.0 (1,2)
*33. weedwack 1249.0 (1,3)
34. cobrevolution 1233.0 (2,11)
* - Player has not played enough games to stabilize rating


Online/Console Joint Ratings (not up to date):

NORTH AMERICA
1. superboomfan 2599.0 (26,3)
*2. kikoushi 2591.0 (7,1)
*3. brax 2558.0 (5,1)
4. isai 2523.0 (14,4)
4. killer 2523.0 (15,1)
6. kefit 2456.0 (22,4)
*7. parallax 2431.0 (4,0)
*8. ruoka danchon 2393.0 (5,2)
8. jaimehr 2393.0 (12,4)
10. star king 2375.0 (24,4)
11. sensei 2373.0 (37,10)
12. banze 2354.0 (10,3)
*13. dexter 2309.0 (4,2)
*14. keyboard king 2297.0 (1,0)
15. ybombb 2258.0 (9,3)
16. sheermadness 2248.0 (12,4)
*17. nangoku 2238.0 (4,3)
18. kerokeroppi 2191.0 (29,14)
19. revan 2153.0 (8,2)
*19. ballin4life 2153.0 (4,2)
21. nintendude 2147.0 (17,10)
22. firo 2135.0 (8,4)
23. ld 2132.0 (16,5)
*24. king funk 2122.0 (4,2)
25. a$ 2096.0 (14,7)
26. greginator 2087.0 (21,11)
27. battlecow 2079.0 (33,19)
28. oxolane 2077.0 (6,4)
*29. near 2069.0 (2,1)
*30. b-rex 2064.0 (4,2)
31. ichiro 2061.0 (6,6)
32. fireblaster 2058.0 (11,5)
33. megavolt 2050.0 (18,10)
34. stranded 2034.0 (27,14)
35. fays 2023.0 (6,4)
*36. moogle 2020.0 (3,2)
*37. gimpyfish 2012.0 (2,2)
38. jel 2002.0 (12,9)
*39. ownasaurus 2000.0 (4,2)
*40. sex 1977.0 (3,2)
*41. sextc 1974.0 (3,1)
*42. dark link -x- 1969.0 (3,2)
42. han solo 1969.0 (14,8)
44. ceasar 1965.0 (9,6)
*45. robert 1959.0 (3,2)
46. *** vwls 1941.0 (6,4)
*47. loto 1936.0 (3,6)
*48. blink 1916.0 (4,4)
49. jimmy joe 1914.0 (17,13)
*50. rocky 1900.0 (3,2)
*50. chaosx522 1900.0 (1,0)
*52. phyziqx 1897.0 (2,2)
53. nova 1891.0 (8,6)
*54. bane 1883.0 (1,1)
54. shade 1883.0 (9,7)
56. clubbadubba 1873.0 (7,5)
57. bladewise 1870.0 (13,9)
58. valoem 1864.0 (17,12)
59. wizzrobe 1859.0 (8,4)
*60. krazy kirby 1848.0 (4,2)
*61. atothez 1847.0 (3,2)
62. korobeniki 1839.0 (12,11)
*63. facecheekwall 1834.0 (3,2)
*64. hamburglar 1828.0 (2,2)
65. rumbly 1827.0 (7,6)
*66. stricnyne 1812.0 (2,2)
*67. peruvian bone crusha 1807.0 (3,2)
68. reborn 1802.0 (6,4)
*69. i love smash 1796.0 (5,2)
*70. unreal 1790.0 (5,2)
*71. gs-djfliphop 1781.0 (3,2)
*72. thebawse 1778.0 (4,2)
*73. butch g 1775.0 (3,2)
*73. hard body 1775.0 (3,2)
*73. espy 1775.0 (3,2)
*76. smelly cat 1769.0 (4,2)
*77. tank 1753.0 (2,2)
*78. eggcelsior! 1744.0 (2,2)
79. vex kasrani 1733.0 (6,4)
*80. blitz 1721.0 (2,2)
*81. whats up tilt 1720.0 (2,2)
82. knitephox 1715.0 (12,11)
*83. runtmc 1714.0 (2,2)
*84. yoshiguru 1706.0 (4,2)
85. jnatty 1703.0 (10,8)
86. wilde 1693.0 (5,9)
*87. tait 1688.0 (1,2)
*88. jared c 1682.0 (2,2)
88. hipstur 1682.0 (4,10)
90. kimimaru 1677.0 (4,10)
*91. preston 1675.0 (3,2)
*92. brendan 1673.0 (3,2)
*93. stacked crooked 1672.0 (1,2)
*94. danconia 1668.0 (2,2)
*94. lilleswing 1668.0 (2,2)
96. tony 1657.0 (8,6)
97. madrush 1650.0 (5,9)
*98. jered 1643.0 (2,2)
*99. johnstermonzter 1624.0 (2,2)
*100. bossman 1622.0 (1,2)
*101. victor 1621.0 (3,2)
*101. gio 1621.0 (3,2)
103. cobrevolution 1617.0 (17,24)
*104. randy 1612.0 (1,2)
*105. callmedaddy 1611.0 (1,2)
106. skyfire 1607.0 (11,13)
*107. vvv rapture 1596.0 (2,2)
*107. eikelmann 1596.0 (2,2)
*109. kmad64 1592.0 (3,2)
110. cmu6eh 1591.0 (10,8)
*110. dc-nerd 1591.0 (4,3)
*112. dvk 1580.0 (1,2)
113. yobolight 1571.0 (8,8)
114. combo blaze 1567.0 (10,16)
*115. mahie 1566.0 (2,3)
*116. twilight sparkle 1565.0 (1,0)
*117. rami jackson 1563.0 (1,2)
118. weedwack 1562.0 (10,13)
*119. to joe 1559.0 (4,4)
120. komotonoto 1555.0 (10,16)
*121. mrschmucker 1554.0 (1,2)
122. nyoouber 1547.0 (2,8)
*123. spencer wdb 1544.0 (2,2)
*124. hotdiggitydog 1515.0 (2,4)
*125. dan 1512.0 (1,2)
*126. brown beard 1511.0 (2,2)
*127. bumblebomb 1489.0 (2,3)
*128. thugz 1485.0 (2,4)
*129. ganon 1467.0 (1,2)
*130. ryan 1458.0 (1,2)
*130. ray rocket 1458.0 (1,2)
*130. vanz 1458.0 (1,2)
*130. ~tac~ 1458.0 (1,2)
*130. ripple 1458.0 (1,2)
*130. duck 1458.0 (1,2)
*130. xaltis 1458.0 (1,2)
137. chain-ace 1457.0 (12,14)
*138. foxy rump 1454.0 (1,2)
*139. corni 1449.0 (1,2)
*140. erick 1448.0 (1,2)
*141. gunblade 1446.0 (2,2)
*142. vatgas 1442.0 (2,2)
*142. george 1442.0 (2,2)
*142. p90curl 1442.0 (2,2)
*142. lulz 1442.0 (1,2)
*146. apple 1441.0 (2,2)
*147. mrmarbles 1437.0 (2,2)
*148. jesusv2 1434.0 (2,2)
*149. joe 1432.0 (1,2)
*150. charlie 1427.0 (2,2)
*151. astro 1426.0 (1,2)
*152. minikilla13 1424.0 (1,2)
*152. 420blazeit 1424.0 (1,2)
*154. kitten 1420.0 (1,2)
*155. triptych 1398.0 (1,2)
*155. tim rich 1398.0 (1,2)
*157. fox 1392.0 (1,2)
*158. zwarm 1390.0 (1,2)
*159. mint 1381.0 (1,1)
*160. kuraudo 1379.0 (1,2)
*160. kevin potter 1379.0 (2,2)
*162. sandman 1375.0 (1,2)
*163. vudujin 1372.0 (1,4)
*164. hammertime 1370.0 (2,4)
*165. the bees knees 1368.0 (1,2)
166. maliki 1361.0 (4,10)
*167. connor walsh 1352.0 (1,2)
*168. black cats31 1326.0 (3,4)
*169. mitch wdb 1321.0 (1,2)
*170. jackie evella 1319.0 (1,2)
*171. hazard2kv 1315.0 (1,2)
*172. eric wdb 1300.0 (1,2)
*173. jace 1294.0 (1,2)
*174. ryan lilly 1293.0 (1,2)
*175. will 1290.0 (1,2)
*175. myztek 1290.0 (1,2)
*175. mitch 1290.0 (1,2)
*178. shredz magoo 1288.0 (1,2)
*179. seagull 1276.0 (1,2)
*179. luigi sama 1276.0 (1,2)
*179. pwii 1276.0 (1,2)
*179. deltacod 1276.0 (1,2)
*179. air 1276.0 (1,2)
*179. xiivi (biblelove2007) 1276.0 (1,2)
*179. pikathunder 1276.0 (1,2)
*179. jwillii 1276.0 (1,2)
*187. low-tier-or-no-tier 1249.0 (1,2)
*188. thetimelimited 1218.0 (1,2)

BRAZIL
1. banze 2354.0 (10,3)
*2. cabelo 2237.0 (3,2)
3. oxolane 2077.0 (6,4)
*4. aprendiz 1984.0 (2,2)
*5. danwear 1788.0 (2,2)
*6. hokage 1764.0 (2,2)
*7. dede 1608.0 (1,2)
*8. onizukasensei 1521.0 (2,2)
*9. joker 1496.0 (2,2)
*10. abuba 1370.0 (1,2)
*-Player has not played enough games to achieve a stable rating yet


Current ratings include:

Console Events:

Apex 2012
FC Legacy
Hyper Combo Finish
GG Smash weekly 3
Apex 2013
Kero v Bcow money match
Red Bank Rumble
Sweet XII
XSmash
Broken Hit Box
Comboed On Spawn 2
Castle SSB Tournament
Zenith 2013
Impulse 2013
Brax Smashfest
Thanks For Playing 2
Smashacre Gluttony
Thanks For Playing 3
Smash Til You Crash 4
DDOS
LA Tourney (name?)
Rolling Thunder
Smashacre Frostbite
COA
Game Clucks
Thanks For Playing 5
Apex 2014
MTL Polybash
Smash Sundays EC 1
Toronto Tourney (name? early april 2014)
Smash Sundays EC 2
Smash of Ages LA
Smash Sundays EC 3


Online Events:

Galaxy Ladder
Double Dair
SSBL 2 (Through Week 7)
 
Last edited:

clubbadubba

Smash Master
Joined
Apr 27, 2011
Messages
4,086
First off, sorry for the rushed post last night. I'm going to be busy all weekend from here on out, but I wanted to get this out there to give people a chance to send me some text files that I could add to the rankings. Some time next week I will edit the first post to be more clear about things. For now all you need to know is that if I get text files of the proper format, I will add them to the ratings.

In terms of whether or not to iterate, I think it might be better to take into account Genesis 2 and Apex 2012, and THEN use Apex 2013, but only do each tournament once. The multiple iterations is kind of arbitrary and isn't consistent. Some players would fair better after 10 iterations, but then worse after 100 iterations. I don't feel like I should pick a number of iterations, because it effects people's ratings too much. Just getting some sort of baseline so not everyone starts out at 1500 could really help the accuracy.

I'm really hoping to get some feedback in the form of text files from tournaments. All those recent NJ tournies should definitely be used. I am also not opposed to using serious money matches, and if any 2 players want me to add a serious match between them that happened outside of a tournament, I am happy to do that as well.

I will probably have 2/3 ratings:

1) Console only
2) Online only
3) Mixed


edit: will probably change the way the ratings is effected by each game to reflect how many games were played by each player. In other words, less games played means each game might shift your score by hundreds. More games played means each game shifts your score by less. Should result in more distribution, instead of 5 way ties. Again, will do this after this weekend.
 

clubbadubba

Smash Master
Joined
Apr 27, 2011
Messages
4,086
-Updated the ratings to change more for the initial few games, which resulted in imo a much more accurate result. After just 1 iteration the results look pretty good.

-Updated 2nd post with new list, getting rid of 100 iters list. Now includes total tournament (W,L) of each player!
 

Froski

Boss of Bosses
Joined
Aug 15, 2010
Messages
381
Location
Zebes
Switch FC
SW-4108-7665-2021
The ratings are making me a little hysterical....


Played in my first FIDE tournament in half a year and lost to a 1900 guy (200 points below mine), 2000 player and a International Master...worst tournament Ive had in a while. Only picked up 2 draws and win.



Irrelevant...i know
 

The Star King

Smash Hero
Joined
Nov 6, 2007
Messages
9,681
I remember when I spent a few months trying to get good at Chess a while back

GOOD TIMES

I still suck
 

clubbadubba

Smash Master
Joined
Apr 27, 2011
Messages
4,086
@banze, please send me your tournament info!

Updates:

-Numbers added, so you can now tell what rank people are overall without counting

-Added a floor to the KFactor, which means after a certain number of games played the amount each game can change your rating will no longer decrease

-Program is now ready to use multiple tournaments


double fianchetto opening ftw. Works against all the noobs I play
 

clubbadubba

Smash Master
Joined
Apr 27, 2011
Messages
4,086
Updated again to include Apex 2012! Okay I've added 3 major tournaments. Time for other people to help me out and add some smaller tournaments.
 

Cobrevolution

Smash Master
Joined
Feb 8, 2012
Messages
3,178
Location
nj
i emailed you the three tourneys i was at whose entrants include myself, weedwack, sensei, megavolt, kerokeroppi, stranded, nintendude, etc. i figure dan is doing a lot of **** and probably doesn't have the time to write this up so I GOTCHU - i'm procrastinating anyway. if i ****ed it up let me know...don't think i did though.
 

clubbadubba

Smash Master
Joined
Apr 27, 2011
Messages
4,086
Thanks man! everything looks to be in the correct format. I'll have to edit some names so that the match the way I previously wrote the names down. Stupid alternating caps/ number for letters people are gonna be the death of me lol. But thanks that's really awesome. I'll get these up hopefully tomorrow.
 

clubbadubba

Smash Master
Joined
Apr 27, 2011
Messages
4,086
22 and 24 are both b rex.
Thanks, I just fixed it. If anyone sees any errors like that just let me know. If I don't get capitalization and spelling to match up between tourneys I get 2 different players. I can probably ignore case in the future, but different spellings I can't really fix.
 

KoRoBeNiKi

Smash Hero
Writing Team
Joined
Apr 26, 2007
Messages
5,950
Location
Northern NJ, 15 minutes from the GW or Manhattan
The ratings are making me a little hysterical....


Played in my first FIDE tournament in half a year and lost to a 1900 guy (200 points below mine), 2000 player and a International Master...worst tournament Ive had in a while. Only picked up 2 draws and win.
Irrelevant...i know
Chess is fun, personally am around 1750, around 1900 for blitz though.

Chess.com/CabassoG if anyone wishes to play a game.
 

Froski

Boss of Bosses
Joined
Aug 15, 2010
Messages
381
Location
Zebes
Switch FC
SW-4108-7665-2021
Chess is fun, personally am around 1750, around 1900 for blitz though.

Chess.com/CabassoG if anyone wishes to play a game.

1900 isnt bad at all might have to play you sometime.

Chess makes me pretty salty when i make foolish blunders....I ended up drawing a Russian in a queen vs 2 pawn endgame because i was under time pressure and had only 7 seconds left on my clock.....I miscalculated constantly and he kept marching those little ****ers down the board.... I screwed up once in every game that tournament, my rating suffered from it....Im now at 2173 (2294 originally)


Guess ill just have to make up some ground when i go to Chitown for the open tournament in may....those cash prizes are calling my name.
 

clubbadubba

Smash Master
Joined
Apr 27, 2011
Messages
4,086
Someone could do Sweet XVI, and there was also a few Canadian tournaments in the past few years if I recall. That one in Toronto?
 

thegreginator

Smash Journeyman
Joined
Oct 22, 2006
Messages
372
This list is awesome.

Kind of confused how it works though.
  1. This does take into account strength of opponent, right? But then how is this "strength" measured? Shouldn't it be measured using the full list (in the second post), which then creates a circularity in the calculation?
  2. Also I'm not following the upside/downside of doing more iterations.
  3. Does this take into account "close" series i.e. losing 2-3 is better than losing 0-3?
  4. Is there a potential to include money matches or other serious non-tournament games? You would think Isai should get some credit for consistently destroying Boom in 12 character battles, for example.
  5. Is "scale" taken into account here, or is it more about just ordering the list? You would think there would be way more separation and would see the top players not as closely grouped to players 10-15 or whatever (people like Isai, Boom and Kikoushi would be way higher than those a few places below them).
Cool stuff though. Just trying to understand how exactly it's calculated.
 

Mahie

Smash Lord
Joined
Aug 18, 2007
Messages
1,067
Location
Lille, France
This list is awesome.

Kind of confused how it works though.
  1. This does take into account strength of opponent, right? But then how is this "strength" measured? Shouldn't it be measured using the full list (in the second post), which then creates a circularity in the calculation?
  2. Also I'm not following the upside/downside of doing more iterations.
  3. Does this take into account "close" series i.e. losing 2-3 is better than losing 0-3?
  4. Is there a potential to include money matches or other serious non-tournament games? You would think Isai should get some credit for consistently destroying Boom in 12 character battles, for example.
  5. Is "scale" taken into account here, or is it more about just ordering the list? You would think there would be way more separation and would see the top players not as closely grouped to players 10-15 or whatever (people like Isai, Boom and Kikoushi would be way higher than those a few places below them).
Cool stuff though. Just trying to understand how exactly it's calculated.
1. The higher the rating of the opponent, the more point it will yield if you beat him. The opposite holds true, the lower he's rated, the more points you will lose, were you to lose.

2. As a direct result of that, multiple iterations of the same tournament increase the ratings of people, making wins over higher rated people worth more, losses against lower rated people also hurt more.
Clubba wanted to emulate the skill level of everyone, and how they should have been rated prior to Apex, by using multiple iterations of it in order to influence the initial ratings more. It results in something much more accurate than the 1st iteration.

3. It should not take into account a close series, it's a binary system, and in tournament nothing other than win or lose matters. (Unless pools etc... but still.)

4. That's not up to me to say.

5. It's a matter of top players not being rated high enough, because of what I explained in 1. : the higher you are rated, and the harder it is to improve your rating. It turns into a slow grind, since everyone else is too low for you, and that means it would take more tournaments with near complete invincibility, like Boom or Isai can have in America, to properly display their true ratings.

Also, not that it really matters, it's Elo, not ELO. And here's the wikipedia page for anyone who's interested in more : http://en.wikipedia.org/wiki/Elo_rating_system
 

clubbadubba

Smash Master
Joined
Apr 27, 2011
Messages
4,086
I updated the original post to hopefully answer some questions about the detail of how this works. I'll answer greg's questions here as well:

1. Yes it takes into account opponents strength. Strength is directly measured by their rating. Higher rating = stronger opponent. Initially, everyone starts out at the same rating (1500), because this is an unbiased list and everyone must earn there rating. The only way to avoid players like Isai starting out at 1500 would be for me to provisionally rate them, which creates bias. Next part is really about iterations (see #2)

2. Initially I thought like you, that I should create ratings, and then loop through them so that players strengths are taken into account more accurately for each loop. However, I found that this wasn't necessarily the case. Say player A had a rating of 1500, and after 1 iteration he had a rating of 1505. After 10 iterations, player A might have a rating of 1515, but then after 100 ratings that same player might have a rating of 1495. And, it appears the ratings don't necessarily converge. Some players were stuck in continuously negative or positive loops, as you will always get points when you win, and always lose points when you lose. So, it would be up to me to select the number of iterations that is "correct," which is arbitrary and then my selection would directly effect the rankings of players. This creates bias which I don't want to introduce.

3. It does not take into account close series, for a couple of reasons. First off, that data is not readily available, so it would be inconsistently accounted for. Secondly, too many players will "play for fun" if they are up in a series. So someone who doesn't deserve to might take a win over someone they would never beat in a real match. Its probably more accurate to use just sets.

4. Yes I definitely want to include non-tournament games that are in a serious setting such as money matches. However, only if these matches are done by a semi-standard tournament ruleset. I don't want to add money matches that are character locked (unless both players clearly wouldn't switch anyway and attest to that), use ridiculous stages, or are best of 1. 12 char battles are a definite no no, because the counterpick rules are completely different and the matchups rarely represent what players would use in a tournament. I could potentially do a separate rating for 12 char battles, but there aren't nearly enough of them in existence to this point. Be clear as well, that this rating is not a definitive list of who's best. It simply rates players based on what characters they have used in tournaments. We are all aware that Isai hasn't used high tiers in tournament recently, but I can't unbiasedly account for that.

5. There is no "scale" as in saying someone is "twice as good" if they have a rating twice as high. You can calculate expected probability of winning as stated in the OP if you want a direct comparison between players. Over time, good players will likely have their rating continue to increase. I will probably drop older tournaments as time goes on which will counterbalance that as well as keep ratings current (Genesis 2 will probably be dropped after Apex 2014 (its happening right?)). As far as expecting to see more separation, well simply not enough games have been played by most players. Until someone plays 13 games, their rating is still very dynamic. If everyone played enough games against a wide enough spread of talent, the ratings would eventually balance out.
 

thegreginator

Smash Journeyman
Joined
Oct 22, 2006
Messages
372
That definitely makes sense - thanks Mahie and clubba. I'm a big fan of the system, although obviously we need way more data points. It seems to me that interest in SSB64 regional tournaments is increasing, so hopefully we can get a lot more matches in there eventually.
 

clubbadubba

Smash Master
Joined
Apr 27, 2011
Messages
4,086
Updated to include Hyper Combo Finish, Red Bank Rumble, XSmash, and Broken Hit Box.

Some interesting stuff, but again, as long as people continue to compete against each other, the ratings should balance out in the end.
 

Cobrevolution

Smash Master
Joined
Feb 8, 2012
Messages
3,178
Location
nj
can you explain anything about whether or not the quantity is taken into account? like parallax being ranked higher than jaime, for example, with 4-0 vs 15-6.
 

thegreginator

Smash Journeyman
Joined
Oct 22, 2006
Messages
372
Very interesting. Does it make sense to set a minimum number of games needed to qualify for ranking? That prevents people from manipulating the rankings and you avoid stuff like Parallax at 4-0 being ranked #5. The minimum attempts criteria is used in just about every other sport/ranking system.

Also I think it would make sense to have a minimum "caliber" of tournament to qualify ("caliber" could be defined as the average Elo number of attending players). That prevents people from unfairly inflating their rankings with small tournaments where they can easily rack up wins against bad players.

Overall great stuff though.

EDIT: Cobr beat me to it.
 

Cobrevolution

Smash Master
Joined
Feb 8, 2012
Messages
3,178
Location
nj
Also I think it would make sense to have a minimum "caliber" of tournament to qualify ("caliber" could be defined as the average Elo number of attending players). That prevents people from unfairly inflating their rankings with small tournaments where they can easily rack up wins against bad players.
i'm not sure if i'm right but if people are inflating their rankings with small tournaments, it would take a hell of a long time, wouldn't it? like, everyone starts at 1500 and if a 1600 player beats five 1500s...he'd go up like, 20 points or something, maybe even less. at least that's how it was in age of empires.

i do agree that there should be a minimum number of games, though, and maybe have a break away point? like once you reach the part of the list with the lowest ranked player with the minimum games played, then the list starts anew with those who have yet to meet the minimum but are still ranked within that?

just tossing thoughts out.
 

clubbadubba

Smash Master
Joined
Apr 27, 2011
Messages
4,086
All good thoughts guys, keep 'em coming. This is definitely still in the development stage.

1. Is quantity taken into account?

Yes and no. In the OP, I describe a "K Factor" that is unique to each player. This K Factor will get smaller the more games you play, and the amount of points you gain is directly and linearly correlated with how many points you will gain from beating an opponent. This means that your first few games are worth a lot in terms of affecting your rating. The reason for this is if it didn't, then everyone would have to slowly crawl to where they should be. For example, without the K Factor, I think the highest rated player was around 1550, and the lowest was like 1450. If everyone played 10 games a day or something, this would not be necessary, but that is not the case for us. It is important to quickly establish a player's skill level, thus earlier games will drastically change your rating.

However, if you've only won 4 games, you will only receive points 4 times. If you've won 15 games, you get 15 games worth of points. You also get 6 games worth of losses though (for Jaime). Jaime's issue is that he lost 2 games in Genesis 2 early, whereas Parallax has won all 4 games which substantially bumps him up. Also, his particular tournament route was not "easy". Based on people's ratings at the start of Hyper Combo Finish he beat weedwack (1500), Valoem (1901), Megavolt (2085), and Megavolt (2085). Very impressive in terms of ratings.

2. Should there be a minimum # of games played?

Probably, but that would cut out a LOT of people. Tournaments don't happen often enough to o that imo. I would prefer to leave everyone up there and just put the record up there for people's reference. Could change in the future though.

3. Should we keep track of "caliber" of tournaments?

This is tricky, because a random small tournament would likely have an average of around 1500. Guess what else had an average of 1500? Genesis 2. I don't want to start being biased and saying that Genesis was good but this tournament was not. We've gotten ourselves into trouble with that before with places like Peru lol. I think player rating does a good enough job, no need to scale players further by the tournament average. Besides, if there are 30 players out there isolated somewhere, and one of them goes undefeated in a tournament between them, who's to say they aren't as good as someone who wins Apex (cough Gerson cough)? If two groups of players never play, I can't go saying that one group is definitively better than the other.



Oh, and another problem I'm seeing is the existence of forfeits. I know a lot of the matches that I'm using never happened. Particularly noticeable with Smoke2Joints going 0-2 last time he entered. I guess I have to count forfeits, but it definitely doesn't make for accurate ratings. Not to mention, forfeit data doesn't appear to be readily available.
 

clubbadubba

Smash Master
Joined
Apr 27, 2011
Messages
4,086
Absolutely it could, and I intend to do so. As I've said, any competitive set that is played under a ruleset that resembles the recommended ruleset will be used, so this could also include serious money matches.
 

Kefit

Smash Journeyman
Joined
Sep 2, 2003
Messages
357
Location
Bellevue, WA
Nifty. Actually this is really cool, though I think including smaller tournaments is going to dilute its worth. But there aren't exactly a lot of large tournaments so what can you do.

Question: does the data used include the tiebreakers for 5th and 7th place at Apex 2013? I could see those matches not being properly indicated on a standard double elimination bracket, so just making sure.
 
Top Bottom