Jump to content
RichardNixon

WN9 poll 2, scale + account WN9

Which numerical scale should WN9 use?  

233 members have voted

  1. 1. Which numerical scale should WN9 use?

    • Base 300 (unicum 420)
      24
    • Base 320 (superunicum 500)
      66
    • Base 450 (superunicum 700)
      15
    • Base 650 (superunicum 1000)
      128
  2. 2. What percentage of battles should account WN9 use?

    • 50%
      44
    • 55%
      7
    • 60%
      49
    • 65%
      54
    • 70%
      79


Recommended Posts

I voted 650 base because it allows for the highest resolution in the choices you offer. I think this is a matter of balancing "easy to handle" and "accurate". The higher the scale, the more accurate it is (which is good), but it also makes things more difficult to put into perspective. I think 100 WN9 per level is aesthetically beautiful and is more in line with another popular scale (percentage) than the second best choice, IMHO, the 320 base.

I also voted for 70%, assuming what you mean is that the lowest WN9 tanks in the account will not be counted in the total calculation until a point where only 70% of the battles remain.

I have not looked at the actual stats, but in a normal distribution, about 68% of values are within one SD of the mean. This makes for 68% central, 16% above and 16% below. Assuming the average player's tank stats behave sort of like a normal distribution (some bad, lots of average, some good), and that we do not want to remove the top portion, I think a case can actually be made for 84%, which would in effect mean approximately removing everything which is more than one SD below the mean. This would IMHO be more meaningful than just throwing an arbitrary percentage out there. If you think this may be a good idea, it may be interesting to calculate this point with real stats in hand. I'm actually curious to see if it would be close to a normal distribution's 84%.

Share this post


Link to post
Share on other sites

650 because it's nicer number to be super uni at a round 1k.

And 50% easily. Results are compromised in this forum really, way too many reroll unicums.

Share this post


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

I think the account WN9 should be an estimate of recent WN9, so whatever percentage of battles minimizes the gap between those would seem the best to me. Should any considerations be made to players whose battles are mostly in one or a few tanks (with a significant rating gap between those and their other tanks)?

1. It's close enough to be heavily dependent on sample technique, which is pretty tough to control for recents. There's also a general rule that better players have relatively high recent vs account WN9, even on the reroll servers. For top 50% vs 6-month recent EU, the crossover is at ~1.1x base (94th centile). For top 65% the crossover is at ~0.5x base (24th centile).

If you wanted the average player to have account = recent, 55% is probably about right. If you wanted the average WoTLabs player to have account = recent, you'd need to go well below 50%. Apparently the distinguishing quality of good players is that they get better.

2. Currently account WN9 caps the maximum weight (battles) of each tank, so players who only play one tank well get hit pretty hard. I'm scared to put this to a vote. It fits both the achievement metric and skill metric arguments so it's staying anyway.

 

5 hours ago, ZXrage said:

My guess is there's no recent, since they added the 70% thing in order to fulfill the "recent" part

There's still recent. Long recents are a more accurate skill metric (at least for WN9), and shorter recents are handy for checking progress.

 

4 hours ago, SuperTheBoss said:

based on the description I'm a little confused, does it eliminate the lowest 30% of tanks or the lowest 30% of battles?

Sorta both. Lowest 30% of battles, but selected by tank. If you have a small number of battles in a lot of badly-played tanks, then they'll all get dumped.

Share this post


Link to post
Share on other sites

That actually makes a lot more sense. Thanks for the response.

side note, feel free to ignore: As an average level player I feel like I use wn8(9) a bit differently than some of the other people commenting on this. I use xvm to determine expected behaviors in games. For example, if I see a light blue+ player in a heavy by themselves on a flank against a bunch of reds and oranges then I feel more comfortable finishing my flank before flexing in a slower tank or focusing on other things because i have confidence that they'll at least take a while to die. If its a red player im likely to fall back right away and by the time I get turned around they're dead already.

if I see a player with poor stats and under 1k games played I expect them to make a lot if mistakes and be inexperienced but not necessarily make bad decisions and ill try to signal those players to follow me ir fall back or whatever, whereas if you have orange stats with a large number of battles, I treat those players like bots because they have been bad for a long time and the only explanation I can think of is they ignore what's going on around them.

There are plenty of examples but I guess my point is that to me representing as much data as possible is more important to me than how many purples there are. IMO if you are able to reroll purple you're already "good enough" to be relied on in battle and that's all that counts to an average player like myself because I sure as hell couldn't do that.

Anyway, thats my input for what its worth but I feel like a lot of the less great players aren't necessarily going to comment on here so I thought I would put my 2 ¢ in.

Share this post


Link to post
Share on other sites

The thing is that if the percentage is too high the incentive to reroll while still be there (maybe even more) because there won't be anymore statistics to see you're actuall level atm (btw it will probably be harder to spot a reroll with for examples10k battles because you can't compare recent and overall wich are usually really close for rerolls. 

Share this post


Link to post
Share on other sites
25 minutes ago, SuperTheBoss said:

if I see a player with poor stats and under 1k games played I expect them to make a lot if mistakes and be inexperienced but not necessarily make bad decisions and ill try to signal those players to follow me ir fall back or whatever, whereas if you have orange stats with a large number of battles, I treat those players like bots because they have been bad for a long time and the only explanation I can think of is they ignore what's going on around them.

I played a lot of low tiers, and that was my observation too. The low-battle-count players were an unknown though: Often they were rapidly improving players who just played with bad crews and little map knowledge, but some of them were idiots. Throwing away the worst tanks does help to distinguish between the two, although low tier results are unreliable anyway due to newbie MM.

As a side note, XVM win chance was hilariously bad for low tier games because it used battle count in the wrong direction. A team with a couple of 20k battle yellows was grossly overrated compared to a newbie team.

Share this post


Link to post
Share on other sites

For the "account" WN9, I voted for 60%, but would actually prefer that as an upper bound.  If the goal of "account" WN9 is to represent your cumulative ability, I don't think you'd need more than 50% of games to do that.  I guess it depends on how you split up the 50% - if it is the most recent 50%, then 50% is fine.  If it's truly the "best" 50%, I could see some issues.

Would it be possible to do something like most recent 10% plus the best 50% of the remainder?  This would seem to address both the rerolling issue (no need since bad old games will drop off) and the account sale issue (since there's specifically a recent component).  I don't know exactly how you've implemented this, so I don't know if that's even possible (or computationally feasible), but it seems like a discussion maybe worth having?

Share this post


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

Would it be possible to do something like most recent 10% plus the best 50% of the remainder?

Nope. If that was possible I'd be inclined to dump the overall/account and use recents. A hybrid metric has some merits, but using the best tanks is primarily a workaround for not knowing which tanks are recent.

Share this post


Link to post
Share on other sites

I sort of feel like that if you could get 1K to be the equivelent of 5K WN8, it would be good as you could refer to 1K being as the holy grail of skill in WN9. 

Share this post


Link to post
Share on other sites
34 minutes ago, Assassin7 said:

I sort of feel like that if you could get 1K to be the equivelent of 5K WN8, it would be good as you could refer to 1K being as the holy grail of skill in WN9. 

The Kewei/carbon/X3NA spot.

Share this post


Link to post
Share on other sites
38 minutes ago, Assassin7 said:

I sort of feel like that if you could get 1K to be the equivelent of 5K WN8, it would be good as you could refer to 1K being as the holy grail of skill in WN9. 

3 minutes ago, blackzaru said:

The Kewei/carbon/X3NA spot.

They run about 2x average tier 10 recent. The base 450 scale was chosen to keep them barely within 3 digits, and you could go to base 500 to get them on the borderline.

The bands wouldn't be 50s or 100s though, and those seem to be very popular. You'd need to hack two or three colours out of the scale to make it work.

Share this post


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

They run about 2x average tier 10 recent. The base 450 scale was chosen to keep them barely within 3 digits, and you could go to base 500 to get them on the borderline.

The bands wouldn't be 50s or 100s though, and those seem to be very popular. You'd need to hack two or three colours out of the scale to make it work.

Or add more colors to fuck with the color blinds and pubbies. :kwim:

Share this post


Link to post
Share on other sites
28 minutes ago, Assassin7 said:

"man, you're only a dark orange, im a Peach colour therefore Im better obviously."

 

new WN9 colour scale:

mVaeAeK.png

 

Not enough shades of yellow. You are missing "lemon" "mc hammer's pants" and "ripe as fuck banana" at least.

Share this post


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

1. It's close enough to be heavily dependent on sample technique, which is pretty tough to control for recents. There's also a general rule that better players have relatively high recent vs account WN9, even on the reroll servers. For top 50% vs 6-month recent EU, the crossover is at ~1.1x base (94th centile). For top 65% the crossover is at ~0.5x base (24th centile).

If you wanted the average player to have account = recent, 55% is probably about right. If you wanted the average WoTLabs player to have account = recent, you'd need to go well below 50%. Apparently the distinguishing quality of good players is that they get better.

I guess there are arguments both ways. If we wanted to minimize the gap across the skill spectrum (treating it as if it's uniform), then going with a lower percentage is probably right. But if we want to make statistical improvement visible, it might be better to leave the percentage high. Then again, you could also just compare recent WN9 to the WN9 at the beginning of the recent interval, rather than overall.

In that case, I would vote for a lower percentage, and suggest a "pre-recent" WN9 be added to the main stat page of WoTLabs for the purpose of measuring improvement. Or maybe a second recent rating for the interval between 1000 and 2000 battles ago. If the system can handle all that.

It might also be useful or interesting to calculate another "recent" WN9 on WoTLabs using the same weighting technique as the "account" WN9, but for more recent battles instead of best tanks. Then you might have a recent rating that adjusts for single-tank "padding".

18 hours ago, RichardNixon said:

2. Currently account WN9 caps the maximum weight (battles) of each tank, so players who only play one tank well get hit pretty hard. I'm scared to put this to a vote. It fits both the achievement metric and skill metric arguments so it's staying anyway.

What kind of cap is it? Does it cap the weight based on the proportion of battles in the tank or just the raw number of battles? I would recommend the former (because some players play a lot more than others).If you want it to be more robust, you could probably pair the cap with some sort of modifier based on how skewed the battle counts are (though I don't know what kind of measure you could use for that). That might handle cases where nearly every battle is in a single tank. But that's also more inefficient and very few accounts are that way, so it's probably not worth it.

Man, I didn't expect to get back into this stuff. :|

Share this post


Link to post
Share on other sites

Base 320 all the way baby. It combines aesthetics with practicality. Yes, base 650 is a little bit more intuitive, but after a small amount of use I think anyone with an IQ greater than that of a cinder block could get used to base 320.

8 hours ago, RichardNixon said:

Currently account WN9 caps the maximum weight (battles) of each tank, so players who only play one tank well get hit pretty hard. I'm scared to put this to a vote. It fits both the achievement metric and skill metric arguments so it's staying anyway.

Good.

Share this post


Link to post
Share on other sites

Voted 650 base and 50% because 

  1. Less reroll advantage / incentive.
  2. Less effect from historical nerfs & buffs.
  3. More tolerant of stock grinds & play for fun tanks.

 

1&2 are straightforward, 3 people who dump money/tourny/cw gold into free exp can avoid bad tanks and stock grinds entirely(I haven't ground out a stock tank in over 2 years personally probably longer because of skirmishes) and I think that shouldn't be reflected in a skill metric. 

Share this post


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

Voted 650 base and 50% because 

  1. Less reroll advantage / incentive.
  2. Less effect from historical nerfs & buffs.
  3. More tolerant of stock grinds & play for fun tanks.

 

1&2 are straightforward, 3 people who dump money/tourny/cw gold into free exp can avoid bad tanks and stock grinds entirely(I haven't ground out a stock tank in over 2 years personally probably longer because of skirmishes) and I think that shouldn't be reflected in a skill metric. 

I guess it would be quite helpful to know how much % of the average player's battles are in stock tanks , add 15-20% for newbie games and use that percentage. Thats apparently impossible from what i heard though :(

As for base WN9, 420 unicum is dank but now that i rethink it 650 base looks better with the 100 points increments and the 1k superuni.

Share this post


Link to post
Share on other sites

I like the base 450 one.

The base 650 ones, I dunno because super unis out of a thounsad it feels a bit to percentage-y while not accually being a percentage.

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...