Home | About | Tournament Winning Decklists | Forums

OCTGN data: computing Elo of factions


#1

I computed Elo on the OCTGN data dump, but considering each faction to be a player.

Code: http://pastebin.com/LmprXDhw

Here are the results:

(‘Weyland Consortium | Because We Built It’, 1368.3485186142764)
(‘Haas-Bioroid | Refining the Process’, 1479.0657999366792)
(‘Criminal | Laramy Fisk’, 1483.7526125195084)
(‘Haas-Bioroid | Stronger Together’, 1485.927959453509)
(‘Haas-Bioroid | Selective Mind Mapping’, 1503.7100252112718)
(‘Shaper | Rielle “Kit” Peddler’, 1511.281699955101)
(‘Haas-Bioroid | Guarding the Net’, 1514.5975180819796)
(‘Haas-Bioroid | Engineered for Success’, 1520.4860193341483)
(‘Shaper | The Collective’, 1524.292461126406)
(‘Shaper | The Professor’, 1539.3227273794946)
(‘Criminal | Retired Spook’, 1548.0477264972374)
(‘Jinteki | Biomedical Pioneer’, 1551.2815645302014)
(‘Jinteki | Selective Mind Mapping’, 1557.5227456353664)
(‘Jinteki | The Secrets Within’, 1561.3018329892777)
(‘Weyland Consortium | Building a Better World’, 1571.3088059952445)
(‘Criminal | Stealth Operative’, 1580.8865788021224)
(‘NBN | Making News’, 1586.3687131606418)
(‘Shaper | Nasir Meidan’, 1591.0280082134434)
(‘Anarch | Noise’, 1591.063476916651)
(‘Shaper | Exile’, 1597.9262533957647)
(‘Haas-Bioroid | Infinite Frontiers’, 1603.845118267196)
(‘Haas-Bioroid | Engineering the Future’, 1613.2147507856753)
(‘Shaper | Chaos Theory’, 1624.7119071559243)
(‘Anarch | Reina Roja’, 1633.6682318327064)
(‘Anarch | Quetzal’, 1643.1153523424541)
(‘NBN | Broadcast Center’, 1644.1525994367516)
(‘Anarch | Whizzard’, 1654.4175524706973)
(‘Weyland Consortium | Powering the Future’, 1655.7274761833348)
(‘Criminal | Trained Pragmatist’, 1655.7889068523182)
(‘Criminal | Disappeared Clone’, 1661.9775503230178)
(‘NBN | The World Is Yours’, 1672.5712704021605)
(‘Weyland Consortium | Power Unleashed’, 1672.833430116559)
(‘Jinteki | The Next Generation’, 1674.741623219883)
(‘Shaper | Kate McCaffrey’, 1682.2246859173947)
(‘Jinteki | Replicating Perfection’, 1703.621879003304)
(‘Criminal | Gabriel Santiago’, 1765.5888785433103)
(‘Jinteki | Personal Evolution’, 1777.9911670762524)
(‘Criminal | Andromeda’, 1792.286572322753)

Note that the data only goes up to The Source, so there are no Maxx matches.

The results look pretty plausible to me. Initially I expected random data since there is such a wide range of skill levels of the actual players involved. My guess is that the sample size is large enough for it to average out - for every noob Andy vs pro BWBI, there is a pro Andy vs noob BWBI. Of course there can be bias - maybe no pro would ever touch BWBI ;). A second source of bias is that the card pool has changed over time, so perhaps Gabe’s dominance above is a relic of Core through Spin Cycle, and now that few Gabe matches are being played, he hasn’t plummeted yet.

As a stab at accounting for both biases, I sliced the data a few different ways (you can see in the code where I commented / uncommented sections out). To account for skill level I did two things: 1) compute ELO of players, and do the identity-ELO calculation only over the top 10% of players; 2) run the calculation over the SHL2 dataset instead.

Considering top-10% players, I get:

(‘Haas-Bioroid | Stronger Together’, 1435.6807308327834)
(‘Haas-Bioroid | Engineered for Success’, 1454.0955777224915)
(‘Weyland Consortium | Because We Built It’, 1482.2958981153831)
(‘Haas-Bioroid | Refining the Process’, 1490.398935486028)
(‘Haas-Bioroid | Selective Mind Mapping’, 1495.4036191958742)
(‘Shaper | The Professor’, 1515.7306580029415)
(‘Criminal | Stealth Operative’, 1524.783858431168)
(‘Jinteki | Selective Mind Mapping’, 1524.9313863098457)
(‘Jinteki | The Secrets Within’, 1529.4623210264606)
(‘Jinteki | Biomedical Pioneer’, 1530.554698131379)
(‘Anarch | Reina Roja’, 1532.6558723190738)
(‘Haas-Bioroid | Guarding the Net’, 1539.753030257342)
(‘Weyland Consortium | Building a Better World’, 1572.2571125160264)
(‘Criminal | Laramy Fisk’, 1573.4337395314053)
(‘Haas-Bioroid | Infinite Frontiers’, 1579.1230291334546)
(‘Shaper | Chaos Theory’, 1586.5227048124532)
(‘Shaper | Nasir Meidan’, 1590.1281360791256)
(‘Haas-Bioroid | Engineering the Future’, 1591.192915753684)
(‘Criminal | Retired Spook’, 1592.201129110226)
(‘Criminal | Disappeared Clone’, 1599.7876917444123)
(‘NBN | The World Is Yours’, 1608.3331282651982)
(‘Shaper | Exile’, 1617.2535923823787)
(‘Shaper | Rielle “Kit” Peddler’, 1618.000275491522)
(‘NBN | Making News’, 1619.151012650262)
(‘Anarch | Whizzard’, 1636.9443941366053)
(‘Criminal | Trained Pragmatist’, 1638.1910308586114)
(‘Jinteki | The Next Generation’, 1644.3118342011578)
(‘Anarch | Noise’, 1647.8383569014582)
(‘Weyland Consortium | Powering the Future’, 1655.5098530471373)
(‘Anarch | Quetzal’, 1671.1379637573427)
(‘Shaper | Kate McCaffrey’, 1688.8191136299076)
(‘Weyland Consortium | Power Unleashed’, 1688.9743928990856)
(‘NBN | Broadcast Center’, 1691.563960585645)
(‘Criminal | Gabriel Santiago’, 1708.8078354277936)
(‘Shaper | The Collective’, 1708.9901980474826)
(‘Jinteki | Replicating Perfection’, 1715.4662362243957)
(‘Criminal | Andromeda’, 1744.4574636625189)
(‘Jinteki | Personal Evolution’, 1755.856313319935)

The tier-1 IDs are still there, and NEH is the top NBN ID. This looks a little more reasonable. Let’s take a look at the SHL2 version (and using all players, not just top 10%).

(‘Criminal | Stealth Operative’, 1477.8329209923165)
(‘Haas-Bioroid | Stronger Together’, 1493.557192782788)
(‘Shaper | Chaos Theory’, 1507.0945483408025)
(‘Haas-Bioroid | Engineered for Success’, 1509.708209917151)
(‘Haas-Bioroid | Refining the Process’, 1522.585564666345)
(‘Weyland Consortium | Because We Built It’, 1537.014865577953)
(‘Haas-Bioroid | Guarding the Net’, 1546.415180981656)
(‘Jinteki | Biomedical Pioneer’, 1550.4830214613098)
(‘Weyland Consortium | Building a Better World’, 1555.965297356903)
(‘Jinteki | The Next Generation’, 1556.4754716432826)
(‘Shaper | Exile’, 1558.6881176276445)
(‘Jinteki | The Secrets Within’, 1565.682215785295)
(‘Shaper | Nasir Meidan’, 1566.4742224676836)
(‘Weyland Consortium | Power Unleashed’, 1569.3587887295146)
(‘Criminal | Retired Spook’, 1571.064703277277)
(‘NBN | Making News’, 1582.1013138194446)
(‘Haas-Bioroid | Infinite Frontiers’, 1594.318642949138)
(‘Anarch | Reina Roja’, 1596.4415507353729)
(‘Haas-Bioroid | Engineering the Future’, 1602.520336468047)
(‘Shaper | Rielle “Kit” Peddler’, 1608.9225034312087)
(‘Anarch | Quetzal’, 1621.6568512241536)
(‘Shaper | The Professor’, 1631.829380174841)
(‘Jinteki | Personal Evolution’, 1632.1442008156573)
(‘Jinteki | Replicating Perfection’, 1639.6736567354176)
(‘Anarch | Whizzard’, 1640.9265383325555)
(‘Criminal | Disappeared Clone’, 1645.4622902079468)
(‘Criminal | Gabriel Santiago’, 1649.4979874929768)
(‘NBN | The World Is Yours’, 1653.5520889077463)
(‘Criminal | Trained Pragmatist’, 1670.4241931406636)
(‘Weyland Consortium | Powering the Future’, 1670.9477971635936)
(‘Criminal | Andromeda’, 1679.3900655519844)
(‘Anarch | Noise’, 1687.3855035994952)
(‘Shaper | Kate McCaffrey’, 1720.0852266973548)
(‘NBN | Broadcast Center’, 1784.319550944479)

This is the first result I would consider actually representative of the competitive metagame. For one more slice, let’s consider the whole OCTGN dataset, restricting to top 10% players, and restricting to matches played after version 3.16 which I believe is The Source. (The top-10% player calculation is over all OCTGN versions.)

(‘Haas-Bioroid | Refining the Process’, 1479.740579753991)
(‘Haas-Bioroid | Stronger Together’, 1481.7832721782543)
(‘Haas-Bioroid | Engineered for Success’, 1499.8654588891604)
(‘Shaper | The Professor’, 1512.7055533239914)
(‘Jinteki | The Secrets Within’, 1519.0969538183372)
(‘Anarch | Reina Roja’, 1523.289121369414)
(‘Haas-Bioroid | Guarding the Net’, 1531.6353798397424)
(‘Jinteki | Biomedical Pioneer’, 1532.7297692490722)
(‘Criminal | Stealth Operative’, 1541.54697152877)
(‘Weyland Consortium | Building a Better World’, 1555.5031559014196)
(‘Haas-Bioroid | Infinite Frontiers’, 1569.0885556682892)
(‘Shaper | Chaos Theory’, 1576.3700602923748)
(‘Shaper | Nasir Meidan’, 1579.2476167781674)
(‘Haas-Bioroid | Engineering the Future’, 1582.0088876516095)
(‘Criminal | Retired Spook’, 1582.4193297494908)
(‘Weyland Consortium | Because We Built It’, 1585.4592905336506)
(‘Haas-Bioroid | Selective Mind Mapping’, 1586.9657719330426)
(‘Criminal | Disappeared Clone’, 1588.7055404237888)
(‘NBN | The World Is Yours’, 1596.701907006088)
(‘Shaper | Rielle “Kit” Peddler’, 1608.52374510324)
(‘NBN | Making News’, 1609.4692143004352)
(‘Jinteki | The Next Generation’, 1619.2710231032659)
(‘Shaper | Exile’, 1624.6228033042223)
(‘Anarch | Whizzard’, 1625.8277299609001)
(‘Criminal | Trained Pragmatist’, 1628.9652759751725)
(‘Anarch | Noise’, 1638.4109334875516)
(‘Weyland Consortium | Powering the Future’, 1646.3115910135004)
(‘Anarch | Quetzal’, 1661.1603635974996)
(‘Weyland Consortium | Power Unleashed’, 1662.5098194529937)
(‘Shaper | Kate McCaffrey’, 1679.2835133106032)
(‘NBN | Broadcast Center’, 1682.343351819434)
(‘Criminal | Gabriel Santiago’, 1700.1929361987432)
(‘Jinteki | Replicating Perfection’, 1706.3591622802512)
(‘Criminal | Andromeda’, 1735.681068732948)
(‘Jinteki | Personal Evolution’, 1746.2042924705854)

Here I stopped, having run out of improvements.


#2

Interesting to really see the fall of Criminal there.

Keep up the good work, this is a great post.


#3

you mean identity, not faction
(why yes, i am a pedantic asshole)

interesting, and thanks for sharing


#4

fascinating data. the ELO margin between NEH and next Corp is huge. interesting to see if that changes in a few months. Kate also significantly higher than the next Runner, which is Noise surprisingly. I haven’t seen much Noise on OCTGN in some time, guess I just haven’t been getting paired with them


#5

I am really weirded out by GRNDL’s (strangely impressive) performance across all but one of the different metrics (the whole of SHL2). I mean, I’m aware that a bunch of people try wacky homebrews with Blue Sun, but damn, GRNDL’s ahead in every instance but one (admittedly the one that seems most representative of the competitive metagame, but even so, The Professor is way ahead of Chaos Theory in said data set, so don’t assume too much).


#6

People who play GRNDL tend to have got good at it, where as everyone plays Blue Sun.

For similar reasons, you see BWBI slowly improve through the lists.


#7

it has to be @PeekaySK boosting prof numbers (:


#8

If memory serves, in recent times I’ve played about 10 games with The Dude, and won 8 or 9 of them… so, possibly :stuck_out_tongue:


#9

I think i’ve played like 150+ games with GRNDL last year when I prepared for the french national. Almost an entire month when the only deck I played on OCTGN was GRNDL supermodernism.