Friday, 29 March 2013

The spring Performance Rating rebalance

I got some constructive feedback with examples from players such as NextToYou and Sweetheart about situations where the Performance Rating wasn't behaving as accurately as it could.

Basically, some of the tank parameters that NoobMeter used had become inaccurate, either because WG had balanced or unbalanced tanks, or because they had added whole new lines and classes of tanks, the capabilities of which and whether they are overpowered or underpowered at the time of adding were unknown, and needed to be estimated.

For most players that play a mix of tanks PR would still be accurate as the deviations balance themselves out - in my opinion a much more accurate rating than the alternatives. However for some who only play a "mono-culture" of tanks it could lead to inaccurate results.

So time was ripe for a spring cleanup and rebalance. With the help of my good friend crux2005 and already mentioned NextToYou we reviewed the parameters for all tanks and made a number of adjustments based on available information from a number of sources, some of which weren't available when NoobMeter was initially created.

As a result, the expectations on the following tank classes were on average lowered: t10 TDs (significantly, especially JgPzE-100 and FV215b (183)), t9 TDs (significantly), t8 HTs (slightly), and most SPGs.

The expectations on the following tank classes were on average raised: t10 HTs (slightly), t9 MT (slightly, with exception of T54E1), t8 MT (slightly), and some other lower tier tanks (in nearly all cases slightly).

As a general rule, TDs are still expected to do more damage than HTs and MTs, and SPGs significantly more damage than tanks of the same tier.

What does that all mean?

The database is being recalculated right now. As that is done during today, your PR will change, in most cases very slightly.

The general rule is that if it goes up you have to post on the forums about how PR is the best rating ever, and if it goes down then you have to go on a rage about how it is all a conspiracy against you personally and how the changes were not needed. :)

After that all settles, I hope we will be left with an even more accurate rating to assess player skill and performance in World of Tanks.

As always, if you have suggestions or questions about www.noobmeter.com or the Performance Rating, please remain calm and contact me either commenting on this blog post, or sending me a private message on the forums (username NoobMeter), or by posting in one of the main NoobMeter discussion threads.

Thanks to all those who already do so!

Enjoy your tanking!

P.S. Stay tuned, I have some interesting features for NoobMeter in the works and hopefully launched soon :)

32 comments:

  1. When will changes be visible? (or they are already visible?)

    ReplyDelete
  2. The recalculations are finished. But if you played a "normal" mix then any changes to your PR rating should be small.

    ReplyDelete
  3. My PR went 100 down, my recent PR went 200 down....
    I am still waiting for a rating, which shows your achievements in relation to the tiers youre playing with. That means, its harder to get 1000Xp when you play as tier6 in tier8 battles, what is not visible in ratings like this.

    Also the win rate doesnt always depend on you, but often on your team. There are tanks like KV1S who can win a battle, even outnumbered, but others like AMX12t cant.

    ReplyDelete
  4. Actually, NoobMeter PR does take into account what tiers you play. Very much so.

    ReplyDelete
  5. I dont mean the tiers YOU play, but the tiers youre fighting with

    ReplyDelete
  6. Ah, I see. But that is kind of the same as for everybody. You in your t6 tank get the same amount of t8 battles as anyone else with t6 tank. Unless you platoon with high-tiers.

    ReplyDelete
  7. That depends on the tank. A level6 scout cant be toptier as often as a level 6 heavy. And that doesnt explain why my recent rating-yesterday Iwas playing meds and TDs- decreased by 260.

    ReplyDelete
  8. Mine didn't decrease as sharply, but I still dropped ~60 points, and I'm honestly surprised to see that the expectations for T8 mediums were raised, considering the Centurion 1 tends to have pretty low alpha. I know the average damage stat of the 20 pounder is 230, but in practice it's far lower, usually about 200 even. Perhaps the RNG just despises me, but sub 200 damage shots are commonplace.

    In fact, it seems like the whole equation is so absurdly biased towards alpha that suicide rushing in a T5 medium with a howitzer (M4, PzIV) gets you "Unicum" performance as long as you land at least 2 HEAT shells. Rinse, repeat till 2,000 PR, regardless of how much of a credit to the team you were.

    ReplyDelete
  9. a) Damage requirements in NM for t8 MT are still slightly less than for t8 HT and significantly less than for t8 TDs.

    b) The data that I have indicates that Centurion has one of the highest avg dmg attained from t8 meds, only behind T69.

    It also has one of the highest Tank OP ratings from t8 meds.

    All that despite the low alpha.

    Alpha is not everything, DPM is usually as important and often even more important. If all that mattered was alpha then Stug E-100 would be the best tank in the game :D.

    ReplyDelete
  10. It is true, dmg is highle rated in PR, but it should be differed , on which tank you do it. For example, 150dmg on a löwe with aM4 should be worth more than 400dmg on a Crusader with the same tank.

    By the way, is there a formula for calculating the PR?

    ReplyDelete
  11. There is no data from WG to make such a distinction as to which tanks you make damage on. However, if you keep shooting low-tier tanks it will have a negative impact on WR, so PR will fall anyway.

    There is an algorithm for calculating PR but it is not public.

    ReplyDelete
  12. Oh, I understand that Alpha is not the be-all, end-all of WoT, I'm just commenting that it *seems* like the algorithm is biased towards those high alpha tanks. Obviously, me playing with 2-3 platoon mates in various tanks is far short of an adequate statistical sample, but bear with me.

    On the way to Cent, I would play Comet in a platoon with a friend in an IS- we'd win a lot, and our EXP was usually roughly in the same ballpark give or take a match where one of us got blown up early. At the end of the night, though, her PR would be way higher than mine, by 200+ points. During a crew EXP event, I hopped in my PzIV to grind out the rest of a crew skill. Doing well (but by no means awesome) I found my PR was INSANE, like 2,300 coming from me, a player that's usually around 1,500 (before the spring adjustment, that is). The only thing I can really point to in the stats of those tanks is that they have a disproportionately high alpha (for their tier, at least) of 350 for PzIV and 390 for IS.

    ReplyDelete
  13. Thanks for your answer, but still I have one question? Why is the formula hidden from the public?

    ReplyDelete
  14. a) There is a difference between "hidden" and "not shared" :)

    b) Because IMHO the reasons for sharing it don't outweigh the reasons for not sharing it.

    ReplyDelete
  15. Whilst you're updating the formula to account for the tier tank I was shooting can you update it for how drunk I was when I logged in? ;)

    But seriously PR is by far the best indicator I've seen for what a player is capable of keep up the good work :).

    ReplyDelete
  16. This comment has been removed by the author.

    ReplyDelete
  17. So if the formula is not shared, can you at least say, what has changed, please?
    Im not asking for the whole formula, I just would like to know, what has become more/less important now, in order to increase my PR.

    ReplyDelete
  18. The parameters for different tanks, as described in the blog post above.

    Previously if you played Stug E-100 then you would have a tough time getting a great PR. Now you have a chance, if you play well :).

    ReplyDelete
  19. Does this mean, that "mono-cultures" - as you described them - get a lower rating now?
    I am playing meds mostly, so what will be the changes for me in increasing my PR?
    You mentioned in the text, that the expectations were raised. What does it mean?

    ReplyDelete
  20. No, "mono-cultures" don't have any specific penalties assigned to them.

    But like in real life when just growing one species of crop leaves your farm vulnerable to whoever targets that one species on crop, same here - if you just played this one tank and it was determined that PR had been too generous in what it expects from that tank, then expectations can be raised and your PR gets lowered. :)

    If however you played a mix, then some tanks get changed in one direction, some in others and overall it is usually a wash.

    ReplyDelete
  21. Ah, now I see. It means, when I play Hellcat well, the expectations are raised and I cant increase my PR that fast?

    ReplyDelete
  22. Yes, the basic idea is that just playing an OP tank is not enough to get a good PR rating, you also have to play it well.

    ReplyDelete
  23. I looked at my winrates with some tanks and saw some red numbers. So there is my question: When thy players winrate is lower than the tank`s winrate, the player is bad and his PR goes down?
    Then in order to increase it, I have to win some battles with that tank?

    ReplyDelete
  24. All the red numbers show is that you have a lower winrate in that tank than other players on the server on average.

    It is not directly impacting PR calculation.

    To improve PR you have to win battles and do good damage - doesn't matter in which tank, as long as you do well then your PR will be good.

    ReplyDelete
  25. But its easier to increase your PR playing a tank with red numbers than with green? I mean easier, when we talk about the same tank

    ReplyDelete
  26. In theory it makes no difference.

    A victory in a tank you have 0% victories in can serve as well to boost PR as a victory in a tank you have 100% victories in up to some point.

    But it is easiest to increase PR by playing the tanks you play best. So if you play some tanks better than others and win more with them then continuing to win with them will boost PR more.

    So from that perspective - if you think your good winrates in some tanks are no accident but because you know these tanks well and play them well - you can keep playing them to increase PR.

    Also I can see why some would want to get all green numbers for themselves in the winrate column, just to make it look pretty.

    ReplyDelete
  27. Until you tell us what the formula is: U-S-E-L-E-S-S

    ReplyDelete
    Replies
    1. Yeah, well, that's just, like, your opinion, man.

      Delete
  28. I love the site, like what you have done.
    But time for an however or but. It may be me
    unsure, but I do quite a lot of critical hits
    over actual damage in hp is this factored into
    your rating system?

    ReplyDelete
    Replies
    1. Only to the degree it helps the team win, so is accounted in win rate component.

      Unfortunately WG doesn't include that information in the API feed.

      Delete
  29. This comment has been removed by the author.

    ReplyDelete
  30. Hi man, it is nice to see your commitment to improve your already great site, but:
    Concerning your recent change to the PR I think you made the changes that were not needed:

    a) If you are a good player and you 99% do solo randoms, you have to play a GOOD / OP tank because you will have to carry the game 80% of the time or you have no chance of victory. Very simple, if you even try to think and play something else, it's not gonna end up well :)

    b) For a bad player, having an OP tank is not a big deal. OP tank in the hands of a baddy is so useless if you are on top of your game. Bad players have tactics similar to a drugged-up-headless-chicken.

    Player: RomelMyau

    p.s. Thumbs up for Big Lebowski references in the comments

    ReplyDelete