Jump to content
RichardNixon

WN9 aesthetics poll

What numerical value should WN9 use for players who currently have ~2500 WN8?  

230 members have voted

  1. 1. What numerical value should WN9 use for players who currently have ~2500 WN8?

    • Around 400
      126
    • 2500
      85
    • Around 4000
      19
  2. 2. How difficult should it be to get purple account WN9?

    • More difficult than purple overall WN8
      81
    • Same difficulty as purple overall WN8
      132
    • Easier than purple overall WN8
      17


Recommended Posts

5 minutes ago, Yuri_Yslin said:

Purple should be the 99 percentile. If that means 500 WN9, so be it. Less "fake cums" around = good. Makes the true ones rewarded better :)

Purple is the .1% and people just learned in WN7 and WN8 how to farm and inflate your stats to purple level.

Share this post


Link to post
Share on other sites
22 minutes ago, Trobs said:

Purple is the .1% and people just learned in WN7 and WN8 how to farm and inflate your stats to purple level.

That's why the ranking needs an update. One can pad only so much. Padding 2500 recent is really easy, padding 4500 recent is not. There, problem solved :awyeah:

Share this post


Link to post
Share on other sites
18 hours ago, bjshnog said:

I think I'd actually put 900 WN9 at around 900 WN8, because I think 900 WN8 was placed at around 900 WN7.

Then again, if I had designed the scale from the beginning, I would probably have put either the mean or median at either 1000 or 100, to make it easy to see the ratio of contribution between the player in question and an average player.

Note that the concept of an "average player" is ambiguous, tricky to calculate, and the results tend to be counter-intuitive, due to the tier system and a large proportion of irregular & casual or temporary players. Regular players see a very different average in their games, so I don't consider the skill of the average player to be conceptually useful.

The issue with putting 100 & 1000 anywhere near the distribution centre is that they're a waste of a digit. I did consider putting tier 10 average = 500 to give a more natural base point, but then a handful of players will top 1000. The current unicum = 420 scale is actually tier 10 average = 300.

Some sort of player/performance average = 1000 is an option, although that intersects with WN8 so it's like 2500 = 2500, except with more players taking a numerical hit. Your 900 WN9 = 900 WN8 suggestion is in the same ballpark. This is what 900 = 900 actually does:

300 WN8 -> 490 WN9
600 WN8 -> 700 WN9
900 WN8 -> 900 WN9
1200 WN8 -> 1090 WN9
1500 WN8 -> 1270 WN9
2000 WN8 -> 1540 WN9
2500 WN8 -> 1760 WN9
3000 WN8 -> 1940 WN9

7 hours ago, lt_lolcat said:

What does being purple actually mean?  The top 1%?  The top .1%?  I don't think it should just be arbitrarily easier or harder.  The color ranks should represent something concrete mathematically. 

Percentiles aren't a concrete baseline because the top few percent are massively skewed by reroll popularity. This isn't even a server constant: NA has the highest proportion of rerolls, followed by EU. RU is far behind.

Using percentiles from recent data is an option, although the results are heavily dependent on where you put the activity filters. In the end, all scales are arbitrary.

Share this post


Link to post
Share on other sites
8 hours ago, affenbirne said:

short question here: spotting damage is not available, ok. what about bounced damage? from my POV, spotting is basically as important as bouncing shots, because those shots you bounced never hit an ally.

is bounced damage meanwhile available via api?

The problem with your argument is this:

1. get IS-3

2. yolo Prokhorovka mid

3. get purple WN9

Share this post


Link to post
Share on other sites

This might be an oddball question but since the colors are based on percentile-based sections of the population, why not just have the WN9 scale be a percentile instead of just a raw number?  If you're in the top 0.1%, you're dark purple and the number displayed is 99.9%.  A 43% player is in whatever percentile category he's in - bottom 20% or whatever - and so his WN9 would read as 20% and he's deep red.

Share this post


Link to post
Share on other sites
1 hour ago, RichardNixon said:

Note that the concept of an "average player" is ambiguous, tricky to calculate, and the results tend to be counter-intuitive, due to the tier system and a large proportion of irregular & casual or temporary players. Regular players see a very different average in their games, so I don't consider the skill of the average player to be conceptually useful.

Well you have to do something about it since player stats are all relative. Nobody will average 2k+ dpg if everyone playing the game wasn't turbo trash.

And mapping the "average" player to 500 seems pretty nice since if people do break 1k, then it'll really stand out. Of course, the rating might (and probably will) creep like WN8.

(You could have the player with the highest wn rating be a fixed number and compute people's stats based off of that, but that seems a bit too pessimistic, and I doubt you have the computational power for it.)

Share this post


Link to post
Share on other sites

a s t h e t i c s
s s              
t   t            
h     h          
e       e        
t         t      
i           i    
c             c  
s               s

Spoiler

def aes(strg):
    out=''
    for i in strg:
        out+=i+' '
    out+='\n'

    form=''
    for i in range(len(strg)):
        form+="{}{}{} ".format("{",i,"}")
    form+="\n"

    mold=[" " for i in range(len(strg))]

    for i in range(1, len(strg)):
        mp=mold[:]
        mp[0]=strg[i]
        mp[i]=strg[i]
        cur=form.format(*mp)
        out+=cur

    return(out)

 

 

Share this post


Link to post
Share on other sites
31 minutes ago, Mami_Momoe said:

Well you have to do something about it since player stats are all relative. Nobody will average 2k+ dpg if everyone playing the game wasn't turbo trash.

WN9 does have an internal lock to the tier 10 performance average. That doesn't mean that the definition of skill won't drift, but then that's not constant across servers anyway. The real goal is to prevent the expected values from drifting numerically.

If you really wanted a constant-skill definition then you'd need an adjustment per server at least, and any method requires controversial assumptions. A top-1% player on RU is not necessarily the same "skill" as a top-1% player on NA.

Share this post


Link to post
Share on other sites
4 hours ago, Cunicularius said:

a s t h e t i c s
s s              
t   t            
h     h          
e       e        
t         t      
i           i    
c             c  
s               s

  Hide contents


def aes(strg):
    out=''
    for i in strg:
        out+=i+' '
    out+='\n'

    form=''
    for i in range(len(strg)):
        form+="{}{}{} ".format("{",i,"}")
    form+="\n"

    mold=[" " for i in range(len(strg))]

    for i in range(1, len(strg)):
        mp=mold[:]
        mp[0]=strg[i]
        mp[i]=strg[i]
        cur=form.format(*mp)
        out+=cur

    return(out)

 

 

All the programming skills in the world won't save you from your poor spelling.

Share this post


Link to post
Share on other sites
8 hours ago, RichardNixon said:

Note that the concept of an "average player" is ambiguous, tricky to calculate, and the results tend to be counter-intuitive, due to the tier system and a large proportion of irregular & casual or temporary players. Regular players see a very different average in their games, so I don't consider the skill of the average player to be conceptually useful.

The issue with putting 100 & 1000 anywhere near the distribution centre is that they're a waste of a digit. I did consider putting tier 10 average = 500 to give a more natural base point, but then a handful of players will top 1000. The current unicum = 420 scale is actually tier 10 average = 300.

Some sort of player/performance average = 1000 is an option, although that intersects with WN8 so it's like 2500 = 2500, except with more players taking a numerical hit. Your 900 WN9 = 900 WN8 suggestion is in the same ballpark. This is what 900 = 900 actually does:

300 WN8 -> 490 WN9
600 WN8 -> 700 WN9
900 WN8 -> 900 WN9
1200 WN8 -> 1090 WN9
1500 WN8 -> 1270 WN9
2000 WN8 -> 1540 WN9
2500 WN8 -> 1760 WN9
3000 WN8 -> 1940 WN9

Percentiles aren't a concrete baseline because the top few percent are massively skewed by reroll popularity. This isn't even a server constant: NA has the highest proportion of rerolls, followed by EU. RU is far behind.

Using percentiles from recent data is an option, although the results are heavily dependent on where you put the activity filters. In the end, all scales are arbitrary.

This is actually quite close to the WN6/7 scale if memory serves, which was not too bad since it capped around 2200 and people couldn't really get any higher. Is there a reason for choosing 65% for the "account" WN9?

Share this post


Link to post
Share on other sites
40 minutes ago, Kitten said:

This is actually quite close to the WN6/7 scale if memory serves, which was not too bad since it capped around 2200 and people couldn't really get any higher. Is there a reason for choosing 65% for the "account" WN9?

give me a tldr that i would understand at this moment.

Share this post


Link to post
Share on other sites
11 hours ago, Kitten said:

This is actually quite close to the WN6/7 scale if memory serves, which was not too bad since it capped around 2200 and people couldn't really get any higher.

Yeah, WN7 was roughly contribution-linear. WN8 is the odd one out with its accidental quadratic scaling, but it poisoned that scale for WN9. Re-use it, and anyone above ~1k is going to be wondering why they lost a few hundred points. Likely responses are going to be "Fucking wotlabs elitists" and "What am I doing wrong?".

 

12 hours ago, Kitten said:

Is there a reason for choosing 65% for the "account" WN9?

Not a good one. With lower percentages there's an increasing proportion of stagnant-skill accounts with account WN9 > recent WN9, which is a bit counter-intuitive. There's also the principle that if account WN9 replaces overall WN8, it should act more like a real overall.

I should put it to a poll really, but that's more writing.

Share this post


Link to post
Share on other sites
4 minutes ago, RichardNixon said:

Yeah, WN7 was roughly contribution-linear. WN8 is the odd one out with its accidental quadratic scaling, but it poisoned that scale for WN9. Re-use it, and anyone above ~1k is going to be wondering why they lost a few hundred points. Likely responses are going to be "Fucking wotlabs elitists" and "What am I doing wrong?".

 

Not a good one. With lower percentages there's an increasing proportion of stagnant-skill accounts with account WN9 > recent WN9, which is a bit counter-intuitive. There's also the principle that if account WN9 replaces overall WN8, it should act more like a real overall.

I should put it to a poll really, but that's more writing.

Yeah, but chances are those reactions will happen anyways. Base it off of what you think should be correct, not what the public thinks; it is your metric afterall.

Hmm. The concept seems solid just the number arbitrary. On a side note, if you want the documentation translated to different languages I think that's doable into at least German and Hungarian on these forums.

Share this post


Link to post
Share on other sites

I'd really like 4000 for improvement and it being different but seems like not many people want 4k so I opted for 2500, because I don't like working with small numbers, might be masochism tho.

 

And yes I'm a turd who hangs out on forums, surprising huh?

Share this post


Link to post
Share on other sites
On 5/27/2016 at 4:52 PM, Ezz said:

It's a pretty ironic question at this point, given WG have now found it in their wisdom to incentivise platooning, while WN9 attempts to disincentivise it.

pretty sure they took out the platoon padding element. Can anyone confirm this?

Share this post


Link to post
Share on other sites
On 28/05/2016 at 7:16 PM, Kitten said:

Yeah, but chances are those reactions will happen anyways. Base it off of what you think should be correct, not what the public thinks; it is your metric afterall.

The scale's just an arbitrary multiplier, so there is no right or wrong. Internally it's just tier 10 average = 1.0, AFK = 0.

Personally I found anything overlapping the WN8 range to be pretty annoying due to the mismatch. If WN8 and WN9 fitted reasonably well on the same scale, I'd have just used that and not bothered polling, although I have no particular attachment to the old scale. It was just a least-resistance option.

In retrospect, I should have had an option of returning to the ~WN7 scale. It's the maximum-grief option, but that shouldn't rule it out of a WoTLabs poll. Might re-run it.

Share this post


Link to post
Share on other sites

Ok, let's answer some semi-topical stuff in more detail:

On 27/05/2016 at 1:22 PM, Sarzan said:

Silly question.  Are legacy accounts (not active after x amount of time) excluded when evaluating ranges for the scale and for the derivation of the algorithm for it?

Well, the average tier 10 performance was determined with a random sample over a couple of months. This was quite tricky to implement. Amusingly, it's not even possible for WoWS because you can make your stats private.

For percentile distributions I usually exclude accounts with fewer than 100 recent battles and fewer than 2k battles overall. Percentiles for overall stats are very different on RU, although this may be less true for recent stats.

For reference, EU currently looks like this:

Top 0.1%: 3000 WN8 overall, 3560 WN8 recent.

Top 0.01%: 3650 WN8 overall, 4330 WN8 recent.

These values are heavily inflated by WN8 tank-selection padding. Factor that out, and the 0.1% recent is around 3000 WN8. This does change the meaning of the second question: If it's similarly difficult for a non-padder to get purple WN9, then it's a lot tougher for an E-50 player.

 

On 27/05/2016 at 2:53 PM, affenbirne said:

short question here: spotting damage is not available, ok. what about bounced damage? from my POV, spotting is basically as important as bouncing shots, because those shots you bounced never hit an ally.

is bounced damage meanwhile available via api?

It's available in a typically broken way. If per-tank spotting was available then I'd check if armour efficiency was worth using at the same time. I'm not too concerned because the formula already works well for heavies, and bounced damage has some useless-padding potential.

I'm not too convinced about the merits of tracking damage either, given what I've seen of 3rd-mark play.

Share this post


Link to post
Share on other sites

  I like the dropping of the worst 35%. It reminds me of how the USGA handicapping formula works for golf in the US. A players index is their potential, not what they normally shoot for a score. They look at the best 10 of your last 20 posted scores and there's a .96 multiplier. So if my index is 20, the USGA thinks I have the potential to be 20 over par. In reality I'll probably spend more time around 24 over. I have no clue how valid the system is. There's plenty of padding on the golf course for sure...

  

Share this post


Link to post
Share on other sites

A few visual representations of possible scales might go a long way in helping you decide.  The current scale on your WoT stats site( http://jaj22.org.uk/wotstats.html ) for WN9 is an effective comparison, as is your 900 WN9 = 900 WN8 example earlier in the thread.  Perhaps pick a few scales of your choice to compare to WN8 and WN7 and have another poll/discussion.

Share this post


Link to post
Share on other sites
On 28.5.2016 at 0:00 AM, nemlengyel said:

The problem with your argument is this:

1. get IS-3

2. yolo Prokhorovka mid

3. get purple WN9

I thought a bit about it, but would it really be a problem?

Fast & fragile tanks would get a bonus thru spotting, while slow & sturdy tanks would benefit from bouncers. Both contribute something to their team, which should be honored. Obviously Spot/Bounce should be factored lower than dmg dealt, but anyways.

This kind of teamplay is desired, isn't it? And it would honor those more than the others that sit back in a bush on Prok, dealing dmg unspotted which wouldn't be possible w/o those spotting or taking the hits on the front line.

But that's just a green guy's thinking.

Share this post


Link to post
Share on other sites
10 minutes ago, affenbirne said:

I thought a bit about it, but would it really be a problem?

Fast & fragile tanks would get a bonus thru spotting, while slow & sturdy tanks would benefit from bouncers. Both contribute something to their team, which should be honored. Obviously Spot/Bounce should be factored lower than dmg dealt, but anyways.

This kind of teamplay is desired, isn't it? And it would honor those more than the others that sit back in a bush on Prok, dealing dmg unspotted which wouldn't be possible w/o those spotting or taking the hits on the front line.

But that's just a green guy's thinking.

I don't know if you can still find it, but Garbad did a challenge/experiment where he only allowed himself 5 shots a game, to see if he could quantify the value of spotting/tanking/positioning. He pulled something like a 54% winrate. Whatever you think/thought of Garbad, for a guy who routinely pulled high-60s winrates on solo challenges, it's pretty clear that using your gun dwarfs the impact of anything else in almost all cases.

Obviously there are anecdotal counterexamples. Everyone above green has vision-carried a match on Prok or Mali at least once, but WNx is designed to rate over many battles, and over many battles your gun activity is by far the most important. 

Share this post


Link to post
Share on other sites
9 hours ago, BoilerBandsman said:

I don't know if you can still find it, but Garbad did a challenge/experiment where he only allowed himself 5 shots a game, to see if he could quantify the value of spotting/tanking/positioning. He pulled something like a 54% winrate. Whatever you think/thought of Garbad, for a guy who routinely pulled high-60s winrates on solo challenges, it's pretty clear that using your gun dwarfs the impact of anything else in almost all cases.

Obviously there are anecdotal counterexamples. Everyone above green has vision-carried a match on Prok or Mali at least once, but WNx is designed to rate over many battles, and over many battles your gun activity is by far the most important. 

I found this really old thread thread about it, miraculously untarnished.

Share this post


Link to post
Share on other sites
On 2016-05-28 at 8:08 PM, RichardNixon said:

Yeah, WN7 was roughly contribution-linear. WN8 is the odd one out with its accidental quadratic scaling, but it poisoned that scale for WN9. Re-use it, and anyone above ~1k is going to be wondering why they lost a few hundred points. Likely responses are going to be "Fucking wotlabs elitists" and "What am I doing wrong?".

On 2016-05-30 at 11:26 PM, Rexxie said:

I personally really like the 3-digit scale. It helps convey that this is not WN8, while also looking much cleaner and meaningful.

A 3 digit scale should help for exactly that reason (not WN8). Its different enough to avoid a knee jerk reaction and should force people to actually look up why their stats changed.

On 2016-05-28 at 8:16 PM, Kitten said:

Yeah, but chances are those reactions will happen anyways. Base it off of what you think should be correct, not what the public thinks; it is your metric afterall.

Hmm. The concept seems solid just the number arbitrary. On a side note, if you want the documentation translated to different languages I think that's doable into at least German and Hungarian on these forums.

@Kolni knows Swedish and I know a bit of Danish.

 

Share this post


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...