Jump to content

MaxL_1023

Mathematics Contributor
  • Content Count

    1,494
  • Joined

  • Last visited

  • Days Won

    5

Everything posted by MaxL_1023

  1. Original article by @MaxL_1023 World of Tanks: MaxL_1023′s Strategy Series Dark Green Level Tactic Guide #4: Module Targeting The strategy series intends to help tankers build a comprehensive mechanical, tactical and strategic background. As opposed to a few large articles the focus will be on a series of smaller guides covering individual aspects of gameplay. Divided into three primary levels, the series articles will be tailored to players of varying skill to allow a smooth progression into the upper echelons of player ability. This is a dark green level guide representin
  2. Original article by @MaxL_1023 World of Tanks: MaxL_1023's Strategy Series Green Level Tactic Guide #3: Weakspots The strategy series intends to help tankers build a comprehensive mechanical, tactical and strategic background. As opposed to a few large articles the focus will be on a series of smaller guides covering individual aspects of gameplay. Divided into three primary levels, the series articles will be tailored to players of varying skill to allow a smooth progression into the upper echelons of player ability. This is a green level guide intended for newer players
  3. Original article by @MaxL_1023 World of Tanks: MaxL_1023's Strategy Series Green Level Tactic Guide #2: Target Selection The strategy series intends to help tankers build a comprehensive mechanical, tactical and strategic background. As opposed to a few large articles the focus will be on a series of smaller guides covering individual aspects of gameplay. Divided into three primary levels, the series articles will be tailored to players of varying skill to allow a smooth progression into the upper echelons of player ability. This is a green level guide intended for newer pl
  4. Original article by @MaxL_1023 Green Level Tactic Guide #1: Armor Facing The strategy series intends to help tankers build a comprehensive mechanical, tactical and strategic background. As opposed to a few large articles the focus will be on a series of smaller guides covering individual aspects of gameplay. Divided into three primary levels, the series articles will be tailored to players of varying skill to allow a smooth progression into the upper echelons of player ability. This is a green level guide intended for newer players who are not yet familiar with intermediate level pla
  5. T28C would be better off using the average of the AT-7 and T-25AT - it is basically halfway between those two tanks. 4005 should use the 183's values for now - it is nothing like the 263. The Conway would be better off using a blend of the T30 and 122-54 values.
  6. We had to fight off the Bulbians who wanted 'Ultra Unicum" added for WN8 > 3900 or some random meaningless number only attainable by intentional padding. I doubt we will be able to get wide acceptance for dropping labels altogether.
  7. The issue is that any descriptor used that is not directly statistically based would have the same subjectivity. If we wanted to stay purely statistical, we would need to label the categories something like: Bottom Fifth Bottom Third Below Average Average Above Average Top Fifth Top Twentieth Top Hundredth Top Thousandth Top Ten-Thousandth I don't find that any better. Switching Very Good to Good and Good to Decent doesn't do anything but diminish players in that region with no actual gain in accuracy.
  8. According to your OP: 1.) To get 1600 WN8, the vast majority of tanks require less than 1 kill/battle and less then their HP in damage 2.) This equates to not carrying their own weight According to Statistics (GASP) 1.) The vast majority of players with 1600 WN8 have a WR significantly higher than the ~49% average value Therefore, based on your logic: ?) The vast majority of players with a win rate significantly higher than 49% (the 60-95th percentile range) do not carry their own weight. Does that make any sense?
  9. I have explained that "not being a burden to your team" and "doing your HP in damage" are actually quite far apart when you examine battle statistics. Only 5% of players have a WN8 above 1600. This roughly equates to dealing your HP in damage. 95% percentile is "very good" under any measure - if it was easy to reach that level of skill it would be more common. The fact that there is a stretched tail of unicum players does not diminish the accomplishments of this skill range, it simply skews perspective.
  10. I truly think that Xen is just intentionally stirring up shit now. If you insist that 1600 is not "very good" then you might as well make light purple "Merely Decent" - that is the slope you are falling down. The whole point of using a relative scale was to remove such subjective judgement. I welcome you to come up with something better.
  11. I suspect that people with under 5000 battles still get the rank for their own benefit, but it does not push anyone else down. It is probably just a logic bug - if you have less than 5k games it should return 0 or something.
  12. Use Warpack+. It misdirects the shots of friend and foe alike, making the user the only player on the battlefield capable of dealing damage. It lets just enough shots through to divert the suspicion of the masses.
  13. Player WN8 follows a gamma distribution. There are very few players near 0 (bots and the like), the majority of players near 800-1100 and the population tails off thereafter. The nature of the distribution means that there are more players with very high WN8 than very low, and the mean WN8 is higher than the median. The color scale is currently a piecewise linear system based on WN8 calculated by plugging in various ratios of the expected values. Performance near 1 in everything gives you about ~1565. This is qualified as "very good" because it is higher than the majority of the pl
  14. This is why I can't have nice things. Xen, look up gamma distributions and get back to me.
  15. The whole thing is fucked and always has been fucked. I can guarantee you that if I was not on hiatus when WN8 was developed I would have had some choice words during development. In fact, the more I learnt about WN8 the lower my opinion of it has become. It is not so much because it is inaccurate in all situations - it is a decent approximation when the central limit theorem bails out bad statistical estimates. However, this is why the original efficiency still correlates well to win rate. My problem is that the metric got far more respect and weight than it ever deserved and t
  16. I advocate investigating using a data transform on the various vehicle damage,kill, etc stats to convert them to gaussian distributions.
  17. The point I am trying to make is that we should figure out why our bad data is causing the problems we have and work to fix the method AND the data so that the system works better. It is also a lot easier to get better data (over time we get more and more) then it is to try and work with a bad method.
  18. I don't think fixing these expected values will help much at all if we just do so by adjusting them manually. It won't change stats significantly and we can't even agree on what to adjust them to. I want to see a lot more hard data regarding tank vs tank per account before even starting to make these kind of manual changes.
  19. I would rather be a consultant as opposed to the main architect - I am better at tearing things down than building them up. Also, I am also sick and tired of people with no idea what they are talking about continuously screwing up the development of every metric in existence.
  20. If you guys don't calm down I am going to get someone to nuke this entire thread. At this point substituting subjective expected values is a hotfix using scale analysis. It gets the right order of magnitude and can accurately estimate which tank values should be higher or lower than most comparable examples. However, it not a permanent fix. It will not analytically determine anything, it won't "prove" relations using objective data and it can't be used to improve the accuracy of WN8 beyond removing a source of error which is barely significant anyways. We are approaching this from th
  21. Would we see any benefits from switching over from a linear regression to a polynomial? Tanks with different "skill ceilings" would be fitted better assuming that relative performance is not a fixed ratio everywhere. I can see curvature in that M48 graph - it looks like the values under 1.0 RDamage are consistently low and those above end up diverging high.
  22. Stay on topic. There doesn't need to be a comprehensive list of tanks which needs adjustment until we can figure out how we can adjust them.
×
×
  • Create New...