Praetor77

The new efficiency formula is even worse than the old one. Yes it gives slightly less cap points, and is more difficult to farm through capping, but it is HEAVILY tier dependant. Two players with same skills, but one with avg tier played 6 and another with avg tier played 9 will get EXTREMELY different efficiency scores. Efficiency was already tier dependant and you got higher score with higher average tier played, but Eff2.0 is even worse. Damage/tier calculation is TERRIBLE. You just get more points for playing higher tier tanks, even if you do worse than at low tiers. This is the points OLDeff NEWeff and WN6 award for doing approximate average damages for each tier (data taken from vbaddict.net). WN6 is pretty balanced from tiers 5-10, while oldefficiency favors high tiers, and NEWeff is just terrible. As this following table shows, new efficiency gives you the same points for doing 1120 damage in tier 4, than for doing 1260 in tier 6, 1365 in tier 8 or 1450 in a tier 10 tank.... unbelievable:

Exactly, they care about their own number, and will quickly get used to, and accept, the change. Also, like I said new eff or WN6, they will drop anyways so... What they don´t care about is a slight change in the % player scale. I don´t see the true advantage in slightly changing the metric to adjust to the old eff scales.

I thought both things were a problem for you, the scale and the points drop. I don´t really think anyone cares about the scale being a little different. You can just inform them of this fact by changing the color scale. REALLY, no one cares about the scale, specially if it only changes significantly at low scores 0-1000. I think we all understand the scale problem, but you give it a HUGE importance when the rest of us think it´s a very minor problem.

Yeah, and now all the analysis has to be redone for new Efficiency formula. The new efficiency formula dropped scores for everyone, so I guess, now no reason to complain about WN6 score distribution, lol! :D The WN6 score distribution is a gamma distribution for sure. This is when you use ALL the players in the server. As you raise the minimum number of battles considered, it becomes more and more like a normal distribution, or to speak more correctly, this increases the value of the shape parameter and decreases the scale parameter. I believe the new Efficiency formula is still pretty bad. Still very farmable by capping and defense. Also suiscouting 12 enemies still gives you 2400 efficiency... O_o The score of all the best players on the server has dropped with the new effficiency and the scores of pedotankers has actually increased with the new formula! Maybe this is the perfect chance to change to WN6 for default config! New efficiency would drop scores anyways!

Interesting. Now that NEW EFFICIENCY scores are lower than old eff, no one will care about WN6 drop in scores... :D

Nice analysis seyrich! I knew 200 was too small, because we tested that. WE decided on 2000, but I guess 1000 will work almost as well, and increase the % of players a lot. Anyways, 2000 battles makes sense because those are the players that are here to stay and are dedicated to the game. You will rarely see players with less than 2000 games in tier 5+ battles, and measuring their skill is complicated because it fluctuates too much during that first part of the learning curve. Sirmax, back to the analysis, WN6 does not in fact have a normal distribution, in fact it has a gamma distribution, and we know this for some time now. http://en.wikipedia.org/wiki/Gamma_distribution Even without analyzing a sample of players, we knew it would have this kind of distribution since damage and kills have gamma distributions and not normal distributions (we analyed 200000 battles from vbaddict/wot). We tried transforming the data to a normal distribution and then measure each stat in a "mean + x standard deviations" fashion. However this was a VERY complicated formula, and the gain in accuracy seemed to be insignificant (2-5%). This gamma distribution has to be accounted for when calculating chances to win using WN6. This may be why it performs worse than with eff, and I think eff has a closer to normal distribution because a large percentage of players (about 50%) have a huge linear factor that is not influenced by skill, simply by taking the decision to farm at end of game (cap). This increases the value of the shape parameter and decreases the scale parameter, making the curve similar to a normal distribution. That being said, you can see above that with players with a minimum of 2000, WN6 is much closer to Eff %s of players. Regarding the range, we used a similar range to Eff on purpouse, so interpretation of WN6 scores would be easier. Kind of ironic that you are put off exactly by that reason, in being similar but different at player %s for the server. Seyrich, could you compare WN6 vs Eff with the 2000 game player dataset? PS: Including survival is redundant, because damage and kills depend VERY much on survival. Secondly, for some reason average spots seem to be a very important part of the WN6 formula, and somehow areable to predict how often a player wins surprisingly well. Remember this is for overall account stats. I guess it has to do with aggresive play or something, plus the fact that you cant farm spots and damage/kills at the same time. If you camp and play very defensively, high damage and kills but low spot. If you suiscout, lots of spots, but no damage/kills.

I´m sad you feel that way. I beliee WN6 in default conf would be a HUGE step forward. I do not understand why you have such a problem with very VERY bad players getting 50-150 less points than Efficiency. Just redraw the categories like Fromshadow said and I think everything would be good. Efficiency is an EXTREMELY flawed metric. Just change it to WN6 and let people get used to the new numbers.... Also, I repeat, if the limit of games is placed at 1000 instead of 200, the numbers will be much more similar.

That is what I mean, I think WN6 range of scores is actually one of it´s advantages over efficiency and you plan to change WN6 range of scores to be like Efficiency´s... If you repeat the analysis you did but for players with over 2000 games you will see the difference between Eff and WN6 is much smaller. This is simply because even very new players can understand to go into cap at end of game, so their erff increases. However learning to play tanks well and do lots of damage takes more time. Also WN6 penalizes low avg tier, and new players obviously have low avg tier.

WN6 is a fair, statistically sound formula. You propose changing that, just because bad players get lower ratings than on Eff. I do not think that is a sound reason to change WN6. WN6 is designed to measure player skill as well as accurately measure differences in player skill. Someone with 2400 WN6 does 4 times as much as someone with 600 WN6. You want to raise the score of those 600 WN6 players to around 850... why? This will also make WN6 lose that awesome linear scale which is one of WN6´s best qualities IMHO. Let people get used to the new ratings. Also, if the problem is % of server above X WN6 score... why are you so sure that the %s used by Eff are the ones that are "correct"? If you want modify default color scheme for WN6 to fit with old Eff color scheme. I see absolutely no statistical reasons to change WN6 to WN6´. PS: What I would be willing to do to make the scale more similar to eff, and that makes some statistical sense is to make the last part of WN6 formula, which penalizes low tier players, start at tier 4 and not 5. I mean changing: +(6-MIN(TIER,6))*-60 to +(5-MIN(TIER,5))*-75 That should increase WN6 score quite a lot for players with few battles, which is what you are analyzing. Also, increasing minimum games played from 200 to 2000 and you will see a HUGE drop in the difference between WN6 and Efficiency. A lot of the players in your dataset have very few battles, so their avg tie is still low and WN6 gives them a lower score due to that. As I said please raise minimum games from 200 to 1000 or 2000 and you will see a huge change. It is very difficult to measure new players to the game anyways, so no sense in including them in the analysis of a global account metric like WN6.

One of the advantages of WN6 is that it is much more linear than Eff. With eff, you could not tell if a 800 eff player was better than a 600 eff player. With WN6 you can. I don´t see the problem with having a slightly lower server average score (according to wotlabs.net the average eff is 990 and the average WN is 825). A big reason for that drop in score is becuase many, MANY people focus on capping at end of game, and don´t care about trying to play better and do more damage and get more kills. This was a big influence of Eff on player mentality which is sad. It lead people to trying to stay alive (even if they didnt do damage or get kills) just to cap at end of game and their eff went up even though they were absolutely no help at all to their team during the game. People will get used to the new numbers, and once they learn WN does not reward cap, but damage, they will have no choice but to try to play better, instead of cap whoring. Modifying WN6 to WN6´ to be more like Eff for low score players is completely unfair to the average and good players. I don´t think its a good idea to modify a metric which we all agree is a better measure of player skill than Eff, which we are trying to replace, and then try to make it more like the old one just so people will have bigger numbers on their screens. If you suck, you suck. Who cares if the number is 700 or 500, it´s still terrible. However, I DO think color matters. People care more about the color than the number, so changing the color scale is a much better alternative I think to changing WN6 values. People want the number to be the same as their signature and on all other stat sites. Changing the XVM number will just be confusing for everyone.

I've been pretty busy with work lately, sorry about the extended leave of absence. WN6 range is bigger because I believe it more fairly evaluates player skill, giving less points to those that were just stealing cap points at end of game and artificially inflated their efficiency. The same happens at the high end. Efficiency gave too few points for damage and kills, so really good players were getting artificially low scores if they did not cap a lot. WN6 was designed to have a similar range of scores to efficiency, but the differences that still exist, I believe are due to the differences inherent to the calculations. Again, regarding the huge amount of red WN6 players, I believe efficiency gives some of these bad players 1000-1200 scores but simply based on their capture stealing, and not due to actual skill. I have yet to see a RED WN6 player do anything decent in a battle... Players with negative WN6 are SURELY bots, or at are least using bots for large amount of games. They truly DO deserve the negative scores. Of course, WN6 score doesnt make much sense for players who are very new to the game (less than 1000 games played). I find the statistical analysis interesting, but I do not see the advantage in normalizing to efficiency (the proposed change simply raises the scores incrmentally for sub 1500 WN6 players). Making WN6 more like Efficiency makes no sense to me. We all agree WN6 is better, so why try to make it more like Eff? Also, in those graphs we really cannot see what is happening with higher scores, to see that we would need logarithmic scale for y axis, or simply an Eff vs WN6 graph and a Eff vs WN6' graph. I think it is a bad idea, the linear approximation shows that WN6' is simply lowering WN6 scores and adding an arbitrary amount of score. That score will be similar to whan eff adds for capping... again not a good idea IMHO. I think it makes much more sense to simply change the coloring scale. As I proposed in a post, coloring for WN6 would be changed to: "wn": [ { "value": 500, "color": "0xFF1A00" }, { "value": 700, "color": "0xE77F00" }, { "value": 900, "color": "0xECCD00" }, { "value": 1100, "color": "0xAFFF00" }, { "value": 1350, "color": "0x2FFF00" }, { "value": 1500, "color": "0x00FFB3" }, { "value": 1700, "color": "0x00BCFF" }, { "value": 2000, "color": "0x2121FF" }, { "value": 9999, "color": "0x8C00E9" } ], This was done with a small random sample, but this could be changed to achieve a more similar color scale to Eff: "wn": [ { "value": 200, "color": "0xFF1A00" }, { "value": 500, "color": "0xE77F00" }, { "value": 800, "color": "0xECCD00" }, { "value": 1000, "color": "0xAFFF00" }, { "value": 1250, "color": "0x2FFF00" }, { "value": 1400, "color": "0x00FFB3" }, { "value": 1700, "color": "0x00BCFF" }, { "value": 2000, "color": "0x2121FF" }, { "value": 9999, "color": "0x8C00E9" } ], The idea was to actually assign scores that corresponded to % of server players with that score or more... WN5 Percentage Better than X% of server less than 500 8.0% 0.0% 500-699 17.0% 8.0% 700-899 25.0% 25.0% 900-1099 23.0% 50.0% 1100-1349 17.0% 73.0% 1350-1499 5.0% 90.0% 1500-1699 3.0% 95.0% 1700-1999 1.0% 99.0% 2000+ 0.2% 99.8% Maybe limits could be better calculated with this larger sample, but once again, changing WN scores to be more like efficiency is not a good idea. Just change the color scale to be more similar to efficiency % of server color scale if you wish.

The data sirmax posted for different ranges of WN6 is from a random sample of 20000 players, something much more precise could be constructed from XVM's database. Regarding arty, I agree, they get too high WN6 ratings, mainly due to unfair tiers (tier 6 arty does approx the same avg damage than tier 8 tanks). This will be fixed once WG rebalances arty. WN6 is also unfair to scout tanks (lights). There is no way to fairly measure them without WG including a "spotted damage" stat on the player profile page. Nevertheless, I have managed above 2000 WN in the last light tank I have played over 100 games (WZ-131).

I don´t know if Google Translate was very good, but I think it would be great to have an avg tank tier played macro, displaying this value in XVM. Together with WN6 and Winrate, you have a VERY good idea of a player´s skill. Dimitry, WN6 is just called 6 because we started at version 1, WN1, and we are now at the sixth version, WN6.

WN6 lowers the ratings for games played in tiers 1-5 (tier 5 is a very small amount). WN6 attempts to measure player skill regardless of the tier of tanks played, but obviously it is easier to get higher stats in low tier tanks, therefore frags and damage are less rewarded when playing low tier tanks than when playing high tier tabnks. The MIN(Tier, 6) means that the lower value of either 6 and avg tier is chosen. This ensures that frags give the exact same amount of score for tiers 6-10. Lastly, regarding color catergories, this would be the most precise: // Dynamic color by efficiency (only with xvm-stat) "wn": [ { "value": 500, "color": "0xFF1A00" }, { "value": 700, "color": "0xE77F00" }, { "value": 900, "color": "0xECCD00" }, { "value": 1100, "color": "0xAFFF00" }, { "value": 1350, "color": "0x2FFF00" }, { "value": 1500, "color": "0x00FFB3" }, { "value": 1700, "color": "0x00BCFF" }, { "value": 2000, "color": "0x2121FF" }, { "value": 9999, "color": "0x8C00E9" } ], PS: Sirmax that sounds AWESOME! :D

Seems to be working great despite some players (0-8 per battle) whose WN does not seem to be in DB yet. One big advantage of using WN6 is easily spotting bots. Sometimes bots had efficiency of 400-800, which some bad players have the same ratings without being bots, but in WN6 they almost always have scores below 300. It would be nice to have a sample config file in the docs directory just having the standard configuration but replacing efficiency with WN6.

WN4 tired to rate all tiers equally (did not bash seal clubbers). The problem was the top ratings were all tie 1 and 2 players, so people asked for a more "fair" rating, so WN5 and then WN6 lowered the ratings for pedotankers. Also this might be interesting for you guys to evaluate WN6... http://noobmeter.com/serverTop/ru/WN6/

I was just thinking maybe using normal avg tier might be better, since people will want their WN6 value in XVM to be the same as their signatures... let´s just wait for WG to do the arty rebalance... :D Sorry for the mixup, although that WOULD improve WN6 accuracy, it would be confusing probably.

Yeah that would only be temporary until arty rebalance, and would only be worth it if it was easy to implement.

OH!! Stupid me! Sorry about that. No, sirmax, the damage part has no tier cap, since that adjusts the average damage for each tier. The avg tier is just the normal one, though if we could add 2 extra for each game on arty, that would make WN6 even more precise.Don´t know if you can do that. For example: III Su-26 | 40 battles II T-26 | 30 battles II Bison | 10 battle I MS-1 | 20 battles I Leichttraktor | 10 battles I T1 Cunningham | 10 battle Normal average tier would be 2 ((40*3+30+2+10*2+20+10+10)/120). If we could count Su-26 games as tier 5 and Bison as tier 4, the avg tier would now be 2.83 (40*5+30*2+10*4+20+10+10)/120. Don´t know if this could be done, but it would definitely be an improvement. Otherwise, just normal avg tier would be fine.

I don't quite get your question. Tier comes into play for frags (for which tier 1 and 2 recieve less points, 3 a bit more, 4 a bit more and so forth, but is constant for tiers 6-10). Tier also comes into play for measuring damage. We took average damage done at each tier and adjusted an exponential formula to that data. The damage a player does is related to the damage "expected" of him for that tier, the average damage. If he has the average damage, he gets around 400 points, 800 points for double the average damage at that tier, and soforth. The last part of the formula: (6-MIN(TIER,6))*-60 Penalizes players with low average tier a bit more, to bring seal clubbers or pedotankers rating down. Players with average tier less than 6 get increasing penalizers, 60 points for avg tier 5, 120 for tier 4, and so forth, 300 for average tier 1. Was that your question?

wotlabs.net uses WN6 now, and noobmeter.com is adding it as we speak. Dont have my own website. Like I said the formula is a group effort, and non-profit.

Thanks for translating the basics Twopizza. For an average player, kills and damage give about 40% of final score each, spot and winrate give about 8% each and defense gives 4%. The weight assigned to each parameter was decided by rigorous statistical analysis using evolutionary algorithms. We used a random player sample of 30000 players to arrive at the optimal formulas to balance damage and frags for each tier. The winrate portion is just used as a proxy to evaluate a player´s intangible stats, like map awareness, knwing when to track an enemy (which gives you actually less damage sometimes), helping teammates, stopping scouts from spotting arty, etc. The idea is to reward players for things they do which leads to wins, but does not show up on the profile page stats. Let me say again, that this was worked on for a LONG time. We did not discard cap on a whim. Statistical analysis proved that it was almost impossible to include frags fairly, since 99% of it was sitting on cap at end of game, and so should not be rewarded. Also, I am not respoinsible for noobmeter, in fact, I fight with noobmeter a lot regarding his rating, since his formula is secret, and does not seem to give linear results (compare PR scores for lironman and CTO_... they are almost identical although CTO has MUCH better stats. Lastly, spots is still important in score, but less so than efficiency, which make suicide-scouting a way to farm efficiency but not in WN6.

Примечание модератора: Тема закрыта, продолжение обсуждения рейтингов WN6 и WN7 ведём в новой теме. Note by Moderator: Topic is closed. Please continue discussion about WN6 and WN7 in new topic. Hi guys, Alexander asked me to start a topic here. We have a 65 page long discussion with several members of the NA forum who are statistics experts. We aimed at replacing efficiency with a formula that effectively and accurately measured playing skill. I wanted to submit the wN6 formula to you for approval to insert in XVM. I sincerely think we have a better formula than PR or Efficiency. I think it is the most accurate and fair formula to measure performance and battle efficiency. Also, it only uses data from player profiles and not the API which does not work well for NA server. This is the formula, I will be glad to answer any doubts and help you incorporate this into XVM. WN6 formula: (1240-1040/(MIN(TIER,6))^0.164)*FRAGS +DAMAGE*530/(184*e^(0.24*TIER)+130) +SPOT*125 +MIN(DEF,2.2)*100 +((185/(0.17+e^((WINRATE-35)*-0.134)))-500)*0.45 +(6-MIN(TIER,6))*-60 Would like to highlight that MIN() means the number capped at that value, so MIN(TIER, 6) means avg tier capped at 6, (so player avg tier is used if it is lower than 6, otherwise 6 is used) and MIN(DEF,2.2) means defense is capped at 2.2. Like I said we have 65 pages of discussion of how we improved the formula over time. This was a group effort, there have been about 10 people working on WN formula, 4 or 5 of them are statistics experts (Syndicate, crabeatoff, Neatoman, tpapp157). If you want to see details of how the formula was developed, and how it changed over time you can read here, or just ask me: http://forum.worldoftanks.com/index.php?/topic/184017-actual-statistically-sound-efficiency-formula/page__st__1080 WN6 formula gives equal importance to frags and damage (although frags is a much more sound statistic, we decided to give them equal weight to avoid kill farming), around 40% of the total score each. The damage is carefully tied to average tier so an awesome player with avg tier 7 and another with avg tier 9 will get similar scores for their damages. Kills was modified to give less score to players with average tier played 1-3 (seal clubbers, or pedotankers). Statistical analysis revealed spots is very important to win games. Players with more avg spots in their overall accounts win more. Maybe it is because they are more aggresive players than the ones with few spots. Spots is less rewarded than in effiiency though, so suicide scouting does not raise WN score. In efficiency, suicide-spotting 10 enemies gave you 2000 Efficiency points, which would make you a super-unicum. In WN, that gives you 1250 which makes you an average player. Also, if you suicide-scout, your average damage and kills drop, so you cannot farm WN by suiscouting. Defense also gives some score, although it is normally 5% of the total, and it is capped at 2.2 to stop defense farming. Cap is not included since statistical analysis proved that 99% or more of cap points are useless points due to sitting in cap at end of game, and do not contribute to a team win. Lastly, we have a winrate term, having 48% wins leads this term to have a value of 0. This term is an S-curve, which rewards winrates above 48%, but using winrate as a proxy to reward intangible skills like map awareness, knowing when to track tanks which leads to team kills, protecting arty from scouts, etc. All these lead to team wins but dont show anywhere else. Nevertheless, this term accounts for 0-8% of total WN6 score. WN6 is impossible to farm, and is also more fair than efficiency, because having a low cap meant it was almost impossible to have a high efficiency score, even if you were an EXTREMELY good player. Efficiency depended on capping too much, which also made it incredibly easy to farm, and also unfair to measure the skills of players who prefer to hunt last enemies instead of sitting in cap at the end of victories. I will be glad to answer any more questions, and if you prefer discussion on a forum I would have no problem. This table of the top players of the NA server and a few EU server players, and also some middle and low players from NA server will let you see how WN6 and Efficiency respond to the different stats: http://i50.tinypic.com/efh3s7.jpg WN6 has recently been incorporated into wotlabs signatures, and will also replace WN4 on noobmeter.com very soon.