Jump to content
RichardNixon

WN9 candidate prototype

Recommended Posts

Everyone knows that if a rating makes ME look good, it is a great rating.  Anything that makes me look like a tunnel-visioned noob is clearly a bad, albeit accurate, rating.

Check THIS out.  Yes, I was a bad Leo1 driver, but I was also the UNLUCKIEST Leo1 driver to ever drive said pixel tank.  A sample that big with such a big deviation between WR and expected WR.

Leopard110MT49444.1%50.7%1101.91313.21492.71519.21519.20.940.980.900.850.90
Link to post
Share on other sites

as an arty player, I thought I'd throw myself in here.

NNGvgHE.png
CWXAdT5.png

 

sorted by battles played:

8YDeLra.png

top WN9c1

MuDxzF3.png

top WN9c2

nXd8ePb.png

top WN9c2 unmod

TlYxo3N.png

arties I've played mainly or entirely pre-8.6:

SU-26, S-51, M37, M7 priest, M41, M12, Bison, Grille, GW Tiger, the low tier french arties, Lorraine 155 50.

Ones that I have played a lot before and after

Hummel, GW Panther, GW E100, M40/43, AMX 13 F3 (The current Tier 6 french), Lorraine 155 51, Obj 261.

 

when 8.6 hit, I had the 261 (from S-51), GWE, M40/43, and Lorraine 155-51 as the high points of each line.  Everything else I have mainly (or entirely) played post-9.6

Link to post
Share on other sites

I actually look forward to the xvm camo I'll receive when WN9 goes through, pls nerf the T-54 and the Chaffee harder.

Those I think are my tanks that hit the hardest, the Chaffee because I was bad when I played it for ~450 games, the T-54 because reasons, and the T69 too for some reason. Probably because of the high pre-HEAT nerf stats of those tanks?

Link to post
Share on other sites

I actually look forward to the xvm camo I'll receive when WN9 goes through, pls nerf the T-54 and the Chaffee harder.

Those I think are my tanks that hit the hardest, the Chaffee because I was bad when I played it for ~450 games, the T-54 because reasons, and the T69 too for some reason. Probably because of the high pre-HEAT nerf stats of those tanks?

T69 was a HEAT nerf. I'm not sure if it was the HEAT nerfs or the turret armour nerf that hurt the T-54 most, but probably the former. Armour's mostly a winrate hit. The Chaffee used to be effectively a tier higher, although it's a better tier 5 scout than it was a tier 6 scout.

There's no reliable way to detect whether people played a specific tank before or after it was nerfed. For per-tank stats I'm inclined to provide both numbers and let players choose which applies to them. A minimal version would just have an indicator on tanks which were marked down.

Link to post
Share on other sites

Not sure how much my metrics are worth, but the AMX 13 57 for me has a Win% of 62.9% BUT an adjusted Win% of 93.9%. It is also one of my top tanks in WN8 and WN9 1 but not WN9 2, I'm guessing because I don't use it as a proper scout. My low tier and arties are affected by WN9 as others have said, but strangely my 112 spikes rather high on WN9 2

Link to post
Share on other sites

Interesting.  Overall seems pretty good.  I am confused about why I see a 104 dropoff between C1 and C2.  I went throught the majority of my high battle count tanks and almost all increased.  The only significant shift I see in any Lt is the 28.01 which went from 2287.2 to 1576.6 (!!!!) which is probably pretty ridiculous.  I played the 28.01 recently and it's a terrible tank for metrics because it was so overplayed before the nerf and because people try to put in a Taz adjustment.   My T69 has a similar problem where it was dropped nearly 300 points between C1 and C2.  Since that had the same problem of pre-HEAT nerf over-padding I'm assuming this is a problem again with adjusting for players who played pre-whatever release that was.

 

I also went through all of my tiers and none have more than a 40 point drop between c1 and c2 so I'm not seeing how the overall dropped that significantly.

 

Anyway would appreciate someone explaining the 104 point dropoff between c1 and c2 and someone bringing out a color scale to help faciiliate discussion between non-unicums would be nice.

 

Account list
 

leftarrow12.gifAccount stats


Winrate54.7%Plain old useless winrate.
Adj. Winrate54.2%Per-tank adjusted winrate, as if you'd played all your games in an average tank of around tier 7. Uses per-tank skill scaling but is still vulnerable to platoon padding and low game counts.
Top Gun medals139Number of top gun medals.
BIA medals8Number of Brothers in Arms medals. If you platoon with a similarly-skilled player, you'll get roughly equal numbers of BIAs and Top Guns.
CC medals0Number of Crucial Contribution medals. The ratio of CCs to Top Guns is highly dependent on average frags, with good triple platoons approaching a 1:1 CC to TG ratio.
Platoon ratio0.06Rough model for overall platoon padding, using frags, Top Guns, BIAs and CCs. 0 means no platooning. 1.0 suggests heavy platooning, while 2.0 suggests heavy triple-platooning, or platooning with better players.
TEFF1271.7Account TEFF, battle-weighted from the XVM per-tank formula. Uses v10 expected values plus missing tanks. 1000 means "average", 2000 means top-100 player in all tanks.
Personal Rating7537WG's personal rating, read directly from the API. 8800 is approximately equal to WN8 unicum.
WN81827.6Standard account WN8 with v20 expected values plus missing tanks.
WN9 c11658.4WN9 candidate based on similar principles to WN8. Uses recent expected values with grind adjustment, per-tank skill scaling and a pseudo-linear formula. Zero means AFK, 2000 means roughly twice the contribution of 1000. Scaled to put 1565 at the average recent for a tier 10 player. WN8 unicum point is around 2000.
WN9 c21554.8WN9 candidate based on a new method suggested by bjshnog. Formula acts on performance relative to tier averages, with a final per-tank offset & scale adjustment. Should work better than WN9 c1 for light tanks and arty. Scale simlar to WN9 c1.
rWin21.010Single-stat components using data from WN9c1. 1.0 is the level of an average tier 10 player. 0 should be AFK, but the scaling is unfinished.
rDmg21.070
rFrag21.010
rSpot20.978
rDef21.371

leftarrow12.gifTier stats


TierBattlesWinrateAdjWinTEFFWN8WN9c1WN9c2rWin2rDmg2rFrag2rSpot2rDef2
16645.5%42.0%968.41012.71020.41061.80.780.670.690.780.45
235555.2%49.1%938.81093.91085.31120.70.920.690.750.780.77
3108957.1%52.0%1192.51277.21219.31224.40.970.780.820.741.36
4107854.9%50.6%1147.71184.91169.41190.20.940.730.820.791.27
5283056.5%53.6%1341.41673.51570.51523.51.001.011.040.921.10
6266653.7%52.3%1296.61641.41559.21523.80.971.011.010.951.34
7254054.5%54.2%1284.11691.41594.41581.01.011.051.040.881.70
8330454.1%55.2%1260.61742.71740.51703.11.031.121.091.051.32
9246854.9%56.8%1330.91855.31825.01776.61.061.171.171.081.47
10176253.2%57.0%1245.51828.81750.91754.61.061.101.011.221.81

leftarrow12.gifTank stats:

Tank namedownarrow12.gifTierdownarrow12.gifClassdownarrow12.gifBattlesdownarrow12.gifWinratedownarrow12.gifAdjWindownarrow12.gifTEFFdownarrow12.gifWN8downarrow12.gifWN9c1downarrow12.gifWN9c2downarrow12.gifWN9c2 unmoddownarrow12.gifrWin2downarrow12.gifrDmg2downarrow12.gifrFrag2downarrow12.gifrSpot2downarrow12.gifrDef2downarrow12.gif
RhB_Waffentrager8TD30255.0%56.0%1360.21884.31999.01836.52020.21.041.271.291.221.44
T675TD28964.7%58.5%1480.51943.11753.41791.31791.31.091.121.131.190.71
JagdTiger_SdKfz_1858TD26953.5%56.0%1277.61619.41724.81754.11754.11.041.091.031.330.58
M18_Hellcat6TD17955.9%53.3%1340.21815.41799.81662.51828.80.991.131.151.291.09
Waffentrager_IV9TD17856.2%58.0%1470.82268.82169.21977.52175.31.081.361.271.481.77
FCM_36Pak403TD16159.6%54.4%1084.51231.11209.11265.81265.81.010.850.810.340.72
Hetzer4TD11751.3%46.8%1136.3895.91080.11011.21112.30.870.660.800.721.38
Nashorn6TD10952.3%53.0%1327.71567.21591.91615.91615.90.990.990.921.480.44
M8A14TD10064.0%57.9%1716.32559.01954.81966.11966.11.081.221.301.162.14
SU-1006TD9853.1%49.8%1442.71986.61674.11690.61690.60.931.031.151.012.48
Waffentrager_E10010TD9456.4%60.5%1337.52135.62036.01789.02057.41.131.251.381.451.37
Sturer_Emil7TD8146.9%49.0%1297.01828.91643.51632.91632.90.911.000.961.312.01
T25_27TD7565.3%65.7%1524.42522.51921.51946.91946.91.221.231.261.151.18
SU-1527TD7266.7%67.1%1551.72506.72032.02098.22098.21.251.361.310.921.36
StuG_40_AusfG5TD7254.2%50.8%1331.81667.61321.41376.41376.40.950.890.810.571.65
Pz_Sfl_IVc5TD6654.5%54.8%1487.91871.61599.01627.01627.01.021.031.101.34-0.03
G20_Marder_II3TD5858.6%53.3%1086.41156.71224.51127.41240.10.990.750.801.030.86
SU100M17TD5448.1%46.6%1318.51941.81613.11621.81621.80.870.990.941.132.45
E-257TD5455.6%56.2%1104.11648.31527.61533.51533.51.050.960.791.142.61
SU-855TD5459.3%55.9%1320.91643.71294.61350.41350.41.040.840.870.750.73
M36_Slagger6TD5356.6%53.1%1694.92588.11928.41934.31934.30.991.211.041.491.74
JagdPzIV6TD5149.0%47.9%1503.72058.11549.61582.71582.70.890.981.240.682.25
T25_AT7TD4860.4%58.6%1319.91819.91580.71620.91620.91.090.990.941.520.13
ISU-1528TD4643.5%43.5%1132.21547.81457.01501.01501.00.810.940.741.230.22
DickerMax6TD4647.8%47.5%1231.31552.51327.21387.01387.00.890.891.100.76-0.10
Marder_III4TD3652.8%47.5%1418.91786.61549.61584.81584.80.890.991.041.110.20
T823TD2853.6%48.7%1319.71668.21655.61644.81644.80.911.020.891.182.99
T404TD1758.8%52.7%1484.11918.21745.61742.91742.90.981.060.931.552.10
PanzerJager_I2TD1050.0%43.5%1.1360.3185.2292.1292.10.810.190.200.10-0.26
SU100Y6TD771.4%70.5%1315.71750.71806.31775.91775.91.310.971.292.75-0.07
JagdPanther7TD30.0%-0.3%227.4335.2522.2540.8540.8-0.010.440.300.06-0.04
G104_Stug_IV5TD2100.0%110.3%2645.82909.22650.62655.52655.52.061.502.271.922.53
A102_T28_concept7TD10.0%-6.1%1068.3867.81556.41518.31518.3-0.120.551.963.14-0.14
M53_559SPG54254.6%58.1%1520.21761.42132.82008.92209.81.081.401.490.982.23
GB77_FV3046SPG15551.6%50.7%1396.41470.31815.31815.61997.20.941.341.070.561.13
M40M438SPG15458.4%61.9%1425.01767.32273.51879.82443.71.151.591.560.861.12
Lorraine155_518SPG14147.5%48.6%1160.21375.91899.31485.71931.40.911.211.061.182.16
Lorraine155_507SPG14156.0%60.0%1225.01231.51734.41442.72164.11.121.421.420.273.33
AMX_13F3AM6SPG11450.9%51.3%968.0932.01371.61087.81468.50.960.921.140.640.59
M127SPG10952.3%53.4%1190.21236.11642.51367.41914.30.991.271.360.390.09
M415SPG10947.7%45.1%878.71021.41565.8981.11716.90.841.091.200.421.75
_M446SPG9952.5%51.9%1398.71644.81638.11616.91778.50.971.161.180.670.06
M374SPG5653.6%49.7%1250.81372.31156.91213.21213.20.930.740.810.790.51
M7_Priest3SPG5657.1%53.5%267.8398.7889.2516.5878.01.000.500.540.921.59
GB29_Crusader_5inch7SPG5545.5%45.3%1373.81902.21936.51800.91981.00.841.231.240.954.01
GB28_Bishop5SPG5350.9%46.1%1337.61610.51305.71361.21361.20.860.831.120.681.09
GB26_Birch_Gun4SPG2646.2%41.0%1824.11992.21454.11709.21709.20.761.171.26-0.214.01
GB27_Sexton3SPG2676.9%79.2%1962.02410.71776.21795.71795.71.481.111.341.200.61
GW_Mk_VIe2SPG1154.5%52.2%-97.1262.2466.3508.6508.60.970.320.110.830.06
T572SPG540.0%31.0%504.7626.6693.9580.0754.00.580.520.250.273.73
GB25_Loyd_Carrier2SPG475.0%74.3%1309.41556.91545.21440.21440.21.390.800.582.396.86
GB78_Sexton_I3SPG20.0%-19.9%2988.12484.41811.62118.52118.5-0.371.582.78-0.16-0.19
Bat_Chatillon25t10MT70052.4%56.0%1260.31817.11725.71749.91749.91.041.081.001.221.90
T-549MT34853.7%55.2%1229.31752.51689.31559.41715.41.031.061.071.111.57
Object_14010MT32051.6%55.5%1159.11596.01680.71692.21692.21.031.050.931.232.22
M4A3E8_Sherman6MT28059.6%56.1%1467.22055.91715.21716.91716.91.051.091.051.002.03
T54E19MT24356.8%57.4%1246.41770.51693.21731.41731.41.071.101.021.090.85
GB07_Matilda4MT22055.5%49.0%1235.81327.81295.51329.81329.80.910.830.820.841.03
VK3002M6MT21143.1%43.8%1250.91394.21415.11432.61432.60.820.890.930.782.14
Lorraine40t9MT20449.0%52.7%1275.71786.61680.91705.31705.30.981.091.001.021.29
M4_Sherman5MT17658.5%54.9%1299.11588.11796.21391.31808.71.021.171.110.981.38
Pershing8MT16861.9%62.2%1227.41782.91680.21704.61704.61.161.051.011.211.36
M46_Patton9MT15956.0%56.2%1327.22093.11690.61736.61736.61.051.061.181.170.77
T-448MT15748.4%48.1%1246.81677.91540.01571.21571.20.900.971.000.991.45
KV-137MT15154.3%52.5%1392.81446.61359.51396.01396.00.980.880.880.771.06
T698MT11945.4%46.1%1296.62044.41956.71698.21952.90.861.211.111.273.86
PzIII_IV5MT11152.3%51.2%1280.81301.41248.81325.11325.10.950.761.000.930.48
AMX_30_prototype9MT10459.6%61.6%1355.52776.82013.21837.31837.31.151.261.271.242.39
T-345MT9751.5%48.3%1287.31306.51273.61324.11324.10.900.791.030.760.91
PzV7MT9251.1%50.0%1332.01899.51567.71602.01602.00.931.011.020.881.33
T-437MT9154.9%52.9%1154.41324.41232.71271.81271.80.990.770.940.681.66
M48A110MT8647.7%52.9%1159.01840.81647.91636.41636.40.991.060.801.122.61
F68_AMX_Chasseur_de_char_468MT8653.5%56.4%1269.61947.31770.51763.11763.11.051.181.000.911.33
T207MT7656.6%54.4%1281.01932.41587.71598.21598.21.010.980.961.132.25
Panther_M107MT7553.3%53.5%1319.91863.91552.41581.51581.51.000.990.990.921.34
S35_captured3MT7258.3%50.6%1226.91664.71356.81385.61385.60.940.910.970.501.52
GB22_Comet7MT6863.2%62.4%1376.52050.81699.81684.31684.31.161.130.950.752.39
T-284MT6839.7%34.2%784.3370.1545.7589.4589.40.640.330.410.510.41
A447MT6264.5%63.8%1598.62672.01968.11985.01985.01.191.251.291.151.65
GB06_Vickers_Medium_Mk_III3MT5354.7%49.8%966.9808.1811.4853.3853.30.930.530.520.391.26
A436MT5156.9%55.2%1243.61554.11349.41371.01371.01.030.900.780.770.79
T62A10MT5058.0%61.1%1178.31827.81570.41596.81596.81.141.030.910.990.73
M3_Grant4MT5062.0%60.9%1303.31359.51181.21178.11178.11.130.740.720.603.63
GB21_Cromwell6MT4250.0%47.9%1432.22277.61672.71714.21714.20.891.131.171.020.11
M7_med5MT3855.3%52.1%1464.62052.21576.21574.31574.30.971.011.020.861.77
GB23_Centurion8MT3650.0%49.8%1104.81492.71335.21384.31384.30.930.881.050.870.08
VK2001DB4MT3650.0%44.3%516.1363.8640.5694.4694.40.820.390.390.920.10
Ch21_T345MT3259.4%57.9%1442.42091.61673.01653.51653.51.081.091.010.703.54
GB05_Vickers_Medium_Mk_II2MT2955.2%49.0%498.3546.6550.9609.8609.80.910.350.480.57-0.03
T-255MT2657.7%55.1%1421.31932.11527.91576.81576.81.031.031.210.880.00
Chi_Nu5MT2572.0%71.5%1587.72174.51576.41647.21647.21.331.021.300.920.23
Matilda_II_LL5MT2263.6%59.1%1160.41326.11209.31265.71265.71.100.750.950.970.23
Sherman_Jumbo6MT1442.9%41.0%1327.91763.91405.01320.91453.00.761.031.240.45-0.04
Chi_He4MT1471.4%69.8%1428.41641.81407.11416.61416.61.300.871.120.662.58
Chi_Nu_Kai5MT1080.0%84.0%1272.81800.61651.11666.51666.51.571.070.791.082.54
M2_med3MT742.9%35.1%1410.41634.91422.21293.31293.30.650.750.940.6514.91
GB01_Medium_Mark_I1MT650.0%44.2%513.6537.3791.1777.7777.70.820.420.302.050.02
GB85_Cromwell_Berlin6MT580.0%74.9%2392.95224.72912.42551.92551.91.402.092.060.970.11
Ram-II5MT560.0%55.8%1384.51722.21533.31565.91565.91.041.010.811.430.07
Chi_Ha3MT475.0%67.0%2309.42908.12073.02098.92098.91.251.291.741.680.03
D23MT4100.0%100.8%1721.31823.11415.81406.11406.11.880.881.010.673.23
Chi_Ni2MT425.0%19.7%552.9746.3793.3810.8810.80.370.560.171.010.12
T2_med2MT2100.0%86.4%1367.5979.51165.41238.51238.51.610.581.501.440.05
ELC_AMX5LT56552.4%51.8%1285.71453.01520.01565.01565.00.970.970.980.921.03
AMX_13_908LT52651.5%55.5%1228.31787.41749.91776.61776.61.031.121.101.081.31
M3_Stuart3LT31958.0%52.4%1445.51371.21291.81332.71332.70.980.810.860.891.14
M24_Chaffee5LT27254.0%52.5%1341.11871.41949.01625.41869.20.981.311.000.951.28
AMX_12t6LT21247.6%47.3%856.3575.4936.71015.21015.20.880.580.600.930.41
T717LT18952.9%55.5%1372.02069.81885.51820.11820.11.031.231.040.912.67
Tetrarch_LL2LT17659.1%52.5%1341.01566.91551.01567.11567.10.980.971.090.971.22
AMX_13_757LT15550.3%53.3%786.8583.31010.61122.91122.90.990.630.720.980.24
M41_Bulldog7LT10351.5%52.7%1268.01802.71858.41840.91840.90.981.191.101.032.50
LTP3LT9347.3%44.2%1186.21203.01311.11318.01318.00.820.810.890.792.10
M5_Stuart4LT8959.6%59.7%1104.3850.11338.71192.51431.01.110.811.090.761.54
T498LT8860.2%65.8%1324.11977.21847.21877.81877.81.231.251.051.210.21
T376LT7456.8%57.4%1525.72732.42175.82160.42160.41.071.341.561.063.44
T-504LT7055.7%53.2%345.6298.4554.7696.1696.10.990.290.440.960.30
Auf_Panther7LT6443.8%45.5%1269.61722.81631.51615.31615.30.851.050.881.002.51
M22_Locust3LT6355.6%50.6%1362.21816.61460.51466.11466.10.940.950.870.791.30
MT256LT6262.9%67.0%1534.52469.81791.91814.61814.61.251.131.310.841.76
R107_LTB7LT6155.7%56.7%1542.12408.41968.51996.61996.61.061.281.440.931.06
AMX404LT5846.6%42.1%1133.2962.0797.5845.0845.00.780.450.590.721.43
T216LT4463.6%66.0%1458.12087.61661.01673.61673.61.231.110.821.280.33
T1_Cunningham1LT4146.3%42.5%1192.21212.71180.01222.61222.60.790.780.890.580.71
PzII_Luchs4LT3956.4%52.3%342.2455.1655.5733.3733.30.970.430.470.600.15
T-463LT3847.4%42.0%845.0504.0549.4593.4593.40.780.320.400.730.13
VK16025LT3754.1%54.2%1039.1889.11073.81105.61105.61.010.690.560.911.39
M2_lt2LT3441.2%36.0%294.8455.1561.1621.5621.50.670.360.340.690.05
VK28016LT3253.1%54.3%1477.02748.92287.21576.62128.31.011.411.450.799.46
PzII2LT3240.6%36.0%923.9876.0745.0815.7815.70.670.510.500.500.05
GB20_Crusader5LT3161.3%59.1%1663.82548.41805.41860.41860.41.101.161.301.120.44
Ch09_M54LT2646.2%42.2%1033.2834.81025.41067.41067.40.790.670.690.590.69
A-204LT2458.3%56.7%1348.6910.31025.11116.51116.51.060.650.870.810.14
T-1273LT2369.6%62.6%1328.31940.41578.41585.91585.91.171.010.921.051.20
Pz_II_AusfG3LT2055.0%48.4%409.9578.5669.4735.2735.20.900.410.530.740.08
R109_T54S8LT1872.2%76.7%1355.92215.81746.71775.11775.11.431.171.250.970.18
AMX383LT1752.9%46.6%1002.8855.7743.5805.6805.60.870.490.610.450.43
PzI2LT1675.0%68.8%796.8690.5627.5711.7711.71.280.510.580.190.07
GB60_Covenanter4LT1560.0%56.1%1566.81853.41635.21592.01592.01.040.981.360.654.78
Ch08_Type97_Chi_Ha3LT1573.3%65.0%1128.41160.51117.91112.61112.61.210.630.990.633.34
PzI_ausf_C3LT1553.3%49.4%314.8608.9591.4647.9647.90.920.410.400.460.01
BT-73LT1145.5%41.8%1265.0969.3903.4976.9976.90.780.580.840.630.12
Ltraktor1LT812.5%14.7%70.9296.2343.7442.8442.80.270.250.170.38-0.01
_Hotchkiss_H352LT728.6%24.3%347.8587.7560.8607.9607.90.450.370.310.64-0.15
T-262LT757.1%48.8%499.2459.8527.6582.1582.10.910.300.380.491.49
Ch07_Vickers_MkE_Type_BT262LT771.4%61.8%710.9837.4822.8859.7859.71.150.540.470.83-0.08
Ch06_Renault_NC311LT540.0%35.6%1089.71103.11065.91103.61103.60.660.690.690.93-0.02
GB59_Cruiser_Mk_IV3LT450.0%43.2%1439.71256.51678.51398.51398.50.810.701.850.6823.96
D12LT3100.0%88.8%1452.81379.61156.61236.11236.11.660.890.720.29-0.08
BT-22LT3100.0%94.4%803.6611.1649.3708.4708.41.760.450.460.440.05
GB58_Cruiser_Mk_III2LT333.3%32.2%769.0735.9779.7837.0837.00.600.520.550.590.00
RenaultFT1LT366.7%64.7%1003.6889.11018.01039.91039.91.210.660.451.150.12
Ch24_Type646LT250.0%51.4%1421.51660.61649.81865.31865.30.960.971.901.310.22
T7_Combat_Car2LT20.0%-6.7%-578.90.038.882.682.6-0.130.04-0.000.65-0.02
NC271LT2100.0%92.2%1623.81541.11486.61475.81475.81.720.960.551.920.11
Ch15_59_166LT1100.0%111.5%978.91686.21381.31053.31053.32.081.110.261.090.21
MS-11LT1100.0%91.4%-324.2261.0115.7151.6151.61.700.10-0.120.72-0.07
KV15HT54958.7%54.0%1411.21864.71569.61597.41597.41.010.981.140.961.18
IS-68HT48557.1%55.7%1188.61519.81563.01593.21593.21.041.030.960.850.91
R106_KV856HT48354.7%51.9%1253.61485.11558.11458.41604.20.971.001.001.040.75
T297HT29959.5%55.8%1399.41903.51734.11745.61745.61.041.111.140.882.03
IS-38HT28451.4%50.4%1124.61382.41374.01407.11407.10.940.890.820.821.28
IS7HT27553.1%50.1%1140.91211.91197.91238.41238.40.930.760.760.870.74
IS-710HT25155.8%59.0%1354.52013.71895.31905.91905.91.101.221.081.161.68
IS89HT21352.6%53.4%1145.51452.61496.21521.91521.91.000.950.930.931.60
AMX_50_1008HT21054.3%55.0%1290.41839.31768.51784.71784.71.031.131.110.883.00
M1039HT18654.3%55.5%1285.81813.51721.31766.61766.61.031.121.071.180.47
AMX_50_1209HT17156.7%59.0%1262.41805.51728.51769.11769.11.101.141.110.940.78
T11010HT12451.6%55.7%1187.61737.31679.31718.81718.81.041.051.011.361.06
AMX_50_68t10HT12456.5%60.3%1264.32004.41803.01828.71828.71.121.161.011.251.24
E-759HT12060.0%60.1%1308.11988.91725.41751.11751.11.121.121.140.921.31
T328HT10956.0%55.1%1396.12064.51742.21800.51800.51.031.171.410.840.14
PzVIB_Tiger_II8HT10664.2%64.1%1550.82766.12045.02084.72084.71.201.301.411.300.80
ARL_446HT9458.5%57.9%1214.91691.21392.01411.31411.31.080.880.960.751.94
AMX_M4_19457HT8647.7%48.3%1282.41744.51562.41587.91587.90.901.031.000.652.16
BDR_G1B5HT5853.4%50.2%1340.81498.81107.41182.41182.40.940.760.970.520.04
T1506HT5759.6%54.4%1338.51685.61301.81323.51323.51.010.850.680.692.11
M66HT5756.1%52.8%1252.01570.21233.61259.71259.70.980.780.760.781.55
T145HT5766.7%63.2%1446.11958.21669.01658.01658.01.181.041.000.983.04
T1_hvy5HT4461.4%56.7%1413.91689.61322.41375.71375.71.060.890.910.690.30
GB08_Churchill_I5HT2646.2%42.3%1344.01694.11390.21409.31409.30.790.910.620.911.97
B14HT1776.5%74.3%1507.11661.91243.71281.51281.51.390.880.800.600.01
KV26HT1457.1%54.3%903.91266.31314.01309.31309.31.010.790.371.712.69
T57_5810HT1369.2%72.0%1016.11531.01545.21394.11533.51.341.030.700.654.44
GB63_TOG_II6HT1060.0%59.6%723.21041.6909.4965.4965.41.110.590.810.271.55
GB09_Churchill_VII6HT1050.0%48.4%1205.91497.11294.51325.61325.60.900.870.641.18-0.04
VK3001H5HT450.0%47.2%-440.0261.055.667.767.70.880.07-0.05-0.08-0.03
Link to post
Share on other sites

Can't seem to edit my post (too much data in spoiler probably).  Adding in that my 28.01 and T69 have low battle counts so they shouldn't impact me much.  They just stood out.  Arty also stands out as a group so maybe that's the biggest hit.  I do not play an excessive amount of arty so if that's the culprit is seems suprising that it would not specifically hit me at tier (the 53/55 has been played a lot recently for missions) but would hit me overall.  

Also I found a wn7 scale compared to wn8 prior to adding additional tiers to wn8.   I would seem to be inline with my win-rate if the scale is going to be the same so no real 'complaint' about c2.  Just more curiosity. 

NKuTEkW.png

 

Link to post
Share on other sites

WR: 55.4%

AWR: 54.0%

PlatoonR: 1.44

WN8: 1645.1

WN9c1: 1500.5

WN9c2: 1431.1

My thoughts: this seems fairly accurate about the overall performance of my account.  I'm a very social player, so the amount of platooning seems accurate. I get bored/lonely solo-pubbing.  Comparing my top ten tanks by the three metrics, 5 out of 10 are top 10 in all 3: Cromwell, Spersh, 3001P, lloyd carrier, Church III. Lloyd gun carrier aside, all four of those tanks I have gotten rather recently and reflect the effort I have put in recently (as opposed to the vast majority of my games being play4fun yolos). Most other tanks in the top ten still fall just outside it. There are three odd pieces of data related to the top ten tanks that I notice.

My top ten tanks by WN8, WN9c1, and WN9c2:

47692a7553.png

9ac01da4b1.png

8d7ad90256.png

First, the Wespe. I know it was said earlier in this thread that its values need a nerf. It is my #1 tank by Wn9c1 and #3 by c2 as opposed to #43 by wn8. It was probably punished by wn8, but to have that big of a discrepancy between the two metrics is odd. All of my arty games are post 8.6. I imagine people who played it pre-8.6 will see an even bigger buff to it. Also consider that the Lloyd GC, which came post 8.6, has relatively similar rankings across the three metrics (#4, #9, and #4 respectively).

The second is the Panzer IV. Ranked #9, #28, and #3 respectively. Not sure why it is penalized by so heavily by WN9c1.

I guess the last oddity is the M7 Priest. Ranked ~#85, #12, ~#60.  Those

 

4f5ebf8ea4.png

 

I'm a layperson as far as statistics go, but overall, to me this looks good as far as curbing the problems of WN8. I'm looking forward to how this is implemented. There seem to be some good points to abandoning using WN9 as an overall measure and evaluating accounts' overall performance with PR. I'll leave it to you pros to sort that one out.

Link to post
Share on other sites

 

First, the Wespe. I know it was said earlier in this thread that its values need a nerf. It is my #1 tank by Wn9c1 and #3 by c2 as opposed to #43 by wn8. It was probably punished by wn8, but to have that big of a discrepancy between the two metrics is odd. All of my arty games are post 8.6. I imagine people who played it pre-8.6 will see an even bigger buff to it. Also consider that the Lloyd GC, which came post 8.6, has relatively similar rankings across the three metrics (#4, #9, and #4 respectively).

The second is the Panzer IV. Ranked #9, #28, and #3 respectively. Not sure why it is penalized by so heavily by WN9c1.

I guess the last oddity is the M7 Priest. Ranked ~#85, #12, ~#60.

If you played all those tanks post-nerf then you should use the "WN9c2 unmod" column for them. In the case of the Wespe, it looks like you really did have an exceptional performance over those 29 games. This isn't too unusual for artillery because it's so map-dependent: Roll a lot of Prokhorovka and your damage output shoots up.

WN8 treated artillery harshly because all tanks were assumed to have the same skill scaling. In practice it's much easier for a good player to put out exceptional numbers in a medium or light than in a TD or arty. There is a possible issue with WN9 where the true skill scaling of arty is somewhat underrated because arty skill is somewhat orthogonal to skill in other classes. I have a theoretical method to fix that.

 

Anyway would appreciate someone explaining the 104 point dropoff between c1 and c2 and someone bringing out a color scale to help faciiliate discussion between non-unicums would be nice.

There's a possible issue with the normalisation in c2 which may make the overall come out a bit low, but the scale is generally unfinished. No-one put forward an opinion on what numerical scale they'd prefer.

 

Not sure how much my metrics are worth, but the AMX 13 57 for me has a Win% of 62.9% BUT an adjusted Win% of 93.9%. It is also one of my top tanks in WN8 and WN9 1 but not WN9 2, I'm guessing because I don't use it as a proper scout. My low tier and arties are affected by WN9 as others have said, but strangely my 112 spikes rather high on WN9 2

AMX 13 57 adjusted winrate is an actual bug, will be fixed. 112 looks fine. Compared to WN8 you'd expect it to go up relative to the 110 because of the removal of stock bias, while Chinese tanks in general will benefit from the removal of recency bias.

Link to post
Share on other sites

Let´s see...

WR: 61.8%

AWR: 61.6%

PlatoonR: 0.28

WN8: 3030.3

WN9c1: 2318.6

WN9c2: 2201.3

 

Looking at my WN9 sorted by tanks, SPGs seem to dominate the top of my list (The FV304 with almost 900 battles and almost 3k WN9 probably increased the results quite a bit).

That´s it boys, the era of the medium tanks is over!

Stop padding, start clicking!

Link to post
Share on other sites

Ok, some arty stuff. I previously mentioned that the expected values method assumes that skill is single-valued: A good arty player is a good medium tank player, and vice versa. It's likely that this isn't a valid assumption, so I had a look at the damage standard deviation of different tier 10 classes, filtered to post-8.6 players:

Class       Post-8.6 SD     03/15 slope
HT          1.75            1.69   
MT          1.96            1.96
TD          1.58            1.59
SPG         1.30            1.18

The SD results are scaled to fit the slope values for comparison. To reduce the potential problem of very good players avoiding artillery, I filtered to players who played a significant quantity of battles in all classes. The difference wasn't huge, though.

Obviously the main difference is that the SPG skill scaling comes out around 10% higher with the standard deviation method, which is a good excuse for increasing all artillery slope values by the same amount. Other classes seem fine, as other data suggests that the HT skill-scaling falloff is a result of the recent increase in arty counts. RU has even lower HT skill-scaling.

I may also switch arty values back to the (higher) pre-personal mission results, although that depends on whether the current popularity of artillery is temporary. It's possible that a combination of mission-conditioning and that viewpoint mod have caused a permanent change.

 

Link to post
Share on other sites

Well I guess I'm one of the last people on wotlabs to notice this thread - great to see some movement on WN9.

Here are my numbers:

Winrate  =  53.5%
Adj. Winrate  =  52.7%
Platoon ratio  =  0.30
TEFF  =  1152.9    
Personal Rating  =  7159
WN8  =  1485.7
WN9 c1  =  1430.0
WN9 c2  =  1342.7

I dont platoon much, I play lights and mediums a lot, and I never play SPGs. I cant argue with my WN9 c2 overall, but I agree with Crab that the focus needs to be on Recent, not overall. Also damn WG for not publishing assisted damage raw values in the API, as coupled with a 'recent' solution it would remove all the problems and perceptions associated with lights. There is a big difference in winning between those who spot well, and over large durations of the battle, from those who damage farm.

As you finalize this, please think about how it will be sustained, as reverse engineering WN8 to generate expected values after Praetor left wasn't fun, and neither is generating ev's for new tanks. If new tanks require no action, great! 

In any event, dont strive for perfection before rolling something out, as it will never be perfect. WN8 is ready to be retired. Please..!

Link to post
Share on other sites

A question I have of this, is it possible for Richard to have his experimental website show Wn9 for "recent" battles of some sort, whether that be the last 1k battles, or preferable to me the last 5k battles?

There's much talk in this thread of using WGPR for overall and wn9 for recent.  I think it might be interesting to see what the recent would look like.

Link to post
Share on other sites

A question I have of this, is it possible for Richard to have his experimental website show Wn9 for "recent" battles of some sort, whether that be the last 1k battles, or preferable to me the last 5k battles?

The conventional method of generating recent values requires a very different sort of website, with server-side API-pulling code and databases. I'm too cheap for that, although I might generate recent values for all NA & EU players locally and upload them. I only took the first sample last week though, so it would take a while to get anywhere near 5k battles.

For most players, the recent values would look very much like the (un-nerfed) values of the tanks they played most recently, although if you're no longer grinding tanks then it's a little more interesting.

 

Link to post
Share on other sites

The conventional method of generating recent values requires a very different sort of website, with server-side API-pulling code and databases. I'm too cheap for that, although I might generate recent values for all NA & EU players locally and upload them. I only took the first sample last week though, so it would take a while to get anywhere near 5k battles.

For most players, the recent values would look very much like the (un-nerfed) values of the tanks they played most recently, although if you're no longer grinding tanks then it's a little more interesting.

 

this is where we get @never involved

Link to post
Share on other sites

this is where we get @never involved

Not sure you did that @ thingy right. Lemme try @Never

Link to post
Share on other sites

Yes?

The conventional method of generating recent values requires a very different sort of website, with server-side API-pulling code and databases. I'm too cheap for that, although I might generate recent values for all NA & EU players locally and upload them. I only took the first sample last week though, so it would take a while to get anywhere near 5k battles.

For most players, the recent values would look very much like the (un-nerfed) values of the tanks they played most recently, although if you're no longer grinding tanks then it's a little more interesting.

 

I do believe this is what CrabCatMan was summoning you for, the implication being that he was hoping you'd get involved in... something to do with lots of numbers being hosted, I assume. I have no idea, I'm just a pleb

Link to post
Share on other sites

What do you guys need me to do exactly? Didn't quite understand what Richard wants.

It wasn't my intention to summon you, but I guess people really want their recent values.

I guess the idea is that you implement (recent) WN9 with the player data that you already have. No reason why that wouldn't work given that it's designed work on the same data as WN8: Battles per tank, plus damage/frags/spots/def per player. I'm not sure whether you'd want to do that on a separate page or add it to the main page.

This is the current implementation for recent & overall stats in javascript:

var tierAvg = [
	{ win:0.4845, dmg:98.1, frag:0.760, spot:0.984, def:0.477, cap:1.0, weight:0.36 },
	{ win:0.4930, dmg:128.0, frag:0.771, spot:0.993, def:0.623, cap:1.0, weight:0.38 },
	{ win:0.4955, dmg:157.7, frag:0.674, spot:1.125, def:0.553, cap:1.0, weight:0.40 },
	{ win:0.4960, dmg:215.6, frag:0.566, spot:0.852, def:0.588, cap:1.0, weight:0.39 },
	{ win:0.4965, dmg:378.4, frag:0.710, spot:0.999, def:0.652, cap:1.0, weight:0.58 },
	{ win:0.4965, dmg:558.6, frag:0.715, spot:0.991, def:0.608, cap:1.0, weight:0.66 },
	{ win:0.4960, dmg:775.7, frag:0.754, spot:1.046, def:0.641, cap:1.0, weight:0.80 },
	{ win:0.4960, dmg:1081.1, frag:0.798, spot:0.964, def:0.615, cap:1.0, weight:1.0 },
	{ win:0.4955, dmg:1434.7, frag:0.862, spot:0.839, def:0.588, cap:1.0, weight:1.23 },
	{ win:0.4945, dmg:1905.3, frag:1.022, spot:0.959, def:0.619, cap:1.0, weight:1.58 } ];
		
// inputs:
// tanks is array of objects with tank_id and battles vars
// accstats is object containing total battles, damage, frags, spots and defence
// expvals is array indexed by tank_id, containing wn9c2inter, wn9c2slope and wn9c2tier vars
 
function calcWN9(tanks, accstats, expvals)
{
	var edmg=0, efrag=0, espot=0, edef=0, bc=0;
	var wn9c2inter=0, wn9c2slope=0, wn9c2weight=0;
	for (var i=0; i<tanks.length; i++)
	{
		var exp = expvals[tanks[i].tank_id];
//		if (!exp) { }       // chuck out a warning about missing tanks

		var tankbat = tanks[i].battles;
		bc += tankbat;

		var avg = tierAvg[exp.wn9c2tier-1];
		edmg += avg.dmg * tankbat;
		efrag += avg.frag * tankbat;
		espot += avg.spot * tankbat;
		edef += avg.def * tankbat;
		ewin += avg.win * tankbat;
		wn9c2inter += exp.wn9c2inter * tankbat * avg.weight;
		wn9c2slope += exp.wn9c2slope * tankbat * avg.weight;
		wn9c2weight += avg.weight * tankbat;
	}

    var batmod = bc / accstats.battles;
	var rdmg = batmod * accstats.damage / edmg;
	var rfrag = batmod * accstats.frags / efrag;
	var rspot = batmod * accstats.spotted / espot;
	var rdef = batmod * accstats.defence / edef;

	var wn9c2 = 0.7*rdmg + 0.25*Math.sqrt(rfrag*rspot) + 0.05*Math.sqrt(rfrag*Math.sqrt(rdef));
	wn9c2slope /= wn9c2weight; wn9c2inter /= wn9c2weight;
	wn9c2 = (1565/1.306) * (wn9c2 - wn9c2inter) / wn9c2slope;
	return wn9c2;
}

 

The current expvals are available here:

https://docs.google.com/spreadsheets/d/1aBbeJgvH5bETAabx-2YURIXF-gMqLwdFppDDZg8s47o/edit?usp=sharing

You'll only need the wn9inter, wn9slope, wn9tier and IDnum columns.

 

Link to post
Share on other sites

Also fixes added today:

  • Arty gets general slope nerf based on true skill scaling.
  • Arty gets general intercept nerf based on pre-mission values.
  • Fixed typo in overall WN9c2 calc, results should come out a bit higher.
  • Fixed AMX 13 57 adjusted winrates.
  • Added new tanks from patch 9.9.

Data-based platoon ratio rework probably incoming.

Link to post
Share on other sites

Many tank's stats are adjusted up and down during their lives in WOT.   Many players base when they have played many tanks on periods where a specific tank's stats met their personal likes and dislikes.  You use a current 6 week sample of Euro players to set baselines.  How do you address the fact these baselines many times fail to accurately represent the conditions during which players actually played the tanks being rated? 

How do you account for the differences in tactics used on EU and other servers; differences which skew baselines when only EU data is considered?

How do you filter out rerolls and adjust for them?

How do you continue to rationalize not using Spotting Damage?   It has been available for a long period of time now.  I know of no other endeavor where they, when given the chance to create a more accurate (though still inaccurate in this case) measurement,  fail to create the new metric and move forward from that point with the enhanced number.  

Given each class of tank has separate and sometimes quite unique roles, how can you accurately rate all of them using the same formula? 

You state the metric has 2 problem classes of tanks, so why do you present the metric as a reliable indicator of overall skill when you know this isn't true?

Link to post
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.


  • Recently Browsing   0 members

    No registered users viewing this page.

×
×
  • Create New...