Garbad

The problem with WN8

  

198 members have voted

  1. 1. WN8?

    • Will always be flawed, like any composite. But its the best we have.
    • IS THE TRUE AND LIVING WORD OF GOD


623 posts in this topic

 

The API doesn't actually report what stats come from what tank though. Only dossiers do, and they don't record when the stats were recorded.

 

So you'd need a frozen image of vbaddict's database from the night before every single patch (completely seperate entities) so you could compare the differences.

 

That's not actually that hard to do and not that much data.  Holding per-tank data on a ongoing basis is tough, but doing a one-time thing per patch is viable.  

 

The problem is evaluating anyone that was not part of the database at that point, since their data won't have been collected.  One possibility is for sites (or a site) to do a sweep of all players before each patch and maybe provide that via API.  Again, hard to do on an ongoing basis, but not so bad to spend the week or so before a patch getting a snapshot.

 

And Gryphon_, it is true that trying to adjust for patches would never end, but it also isn't a huge amount of work.  Maybe 8 patches per year, and once you have everything in place it's mostly an automated process anyway.

 

Since that is the single biggest flaw with WN8, it's certainly worth considering.

Share this post


Link to post
Share on other sites

I disagree.  I think DPG in a light is pretty irrelevant.  Scouting is what counts, its just a lot of maps scouting doesn't really help and so the scout can't really do much.  But whether you average 300 dpg or 800 dpg doesn't make much difference in win rates against 2500 HP tanks.  That's a role flaw, not an average skill level flaw.

 

I should have made it clearer ("set things right with lights") - t4-5 light in their current form, dont have buisness in t10(9) battles anymore. But give me a Chaffee in t7-8 games on the right maps I m sometimes even able to carry and contribute something meaningfull in t9 battles. Chaffee for example just needs a few more HP and MM from t7-9 and it would work. After that bring down TD camo significantly and viewranges a little - perfect.

Share this post


Link to post
Share on other sites

I have another question...

 

Which source is the most trusted for your WN8 score? I have noticed disparities between WoTLabs.net, Noobmeter.com, and VBAddict's ADU...

 

My 12 game session tonight, for example, has a WN8 of 2506 according to the ADU with noobmeter showing 2997...

 

Gqrl0gI.jpg

  LQlAEHW.jpg

Share this post


Link to post
Share on other sites

The ADU just averages the WN8 scores of each individual battle. Noobmeter shows a proper rating for the period.

Share this post


Link to post
Share on other sites

The ADU just averages the WN8 scores of each individual battle. Noobmeter shows a proper rating for the period.

 

Thank you!

 

I wasn't prepared for the answer to be the higher one!

Share this post


Link to post
Share on other sites

The ADU just averages the WN8 scores of each individual battle. Noobmeter shows a proper rating for the period.

 

What about the differences among Noobmeter, Wotlabs and Orrie's browser script for overall?

Share this post


Link to post
Share on other sites

What about the differences among Noobmeter, Wotlabs and Orrie's browser script for overall?

 

Orrie's, as far as I know, is 100% accurate, as is Noobmeter's and MyWOTStats'. WoTLabs seems to have slightly less precise rounding, so it may be a few points off.

Share this post


Link to post
Share on other sites

Orrie's, as far as I know, is 100% accurate, as is Noobmeter's and MyWOTStats'. WoTLabs seems to have slightly less precise rounding, so it may be a few points off.

 

Sweet thanks  :thumbup:

Share this post


Link to post
Share on other sites

I mentioned this in a different thread, but it probably bears repeating here since it's more appropriate - Perhaps it would be worthwhile to run the tool (Eureka?) that determined the correlation between the various contributions (damage, spots, kills etc) and win rate against only light tank/scout data? It may just be that the WN8 formula weightings are significantly off for light tanks, and while having a separate calculation for light tanks would increase the complexity, it doesn't seem like it would be significantly more challenging than the current model for calculating WN8.

 

I obviously do not know if there's enough data just for light tanks for it to be a meaningful exercise, but given that light tank ratings seems like one of the weak points on WN8, perhaps it's worth at least exploring? If the weightings come in close enough to the overall ones, it's probably not worth incorporating a change like that. But if the weightings for damage and spots are significantly different from the current general ones, it could be a step towards improving WN8. At that point, you'd probably also want to re-run the correlations for the general part with the light tank ratings removed.

Share this post


Link to post
Share on other sites

I mentioned this in a different thread, but it probably bears repeating here since it's more appropriate - Perhaps it would be worthwhile to run the tool (Eureka?) that determined the correlation between the various contributions (damage, spots, kills etc) and win rate against only light tank/scout data? It may just be that the WN8 formula weightings are significantly off for light tanks, and while having a separate calculation for light tanks would increase the complexity, it doesn't seem like it would be significantly more challenging than the current model for calculating WN8.

 

I obviously do not know if there's enough data just for light tanks for it to be a meaningful exercise, but given that light tank ratings seems like one of the weak points on WN8, perhaps it's worth at least exploring? If the weightings come in close enough to the overall ones, it's probably not worth incorporating a change like that. But if the weightings for damage and spots are significantly different from the current general ones, it could be a step towards improving WN8. At that point, you'd probably also want to re-run the correlations for the general part with the light tank ratings removed.

 

Actually, that's not a bad idea, if only for a temporary fix. WN8 (or WN9) could take the form of (WN8 * battlesMT, HT, SPG, TD + WN9scouts * battlesscouts) / totalBattles.

 

After this, VBE needs to be done, so I need to contact someone...

Share this post


Link to post
Share on other sites

stop playing lt at low tiers all together

 

That's a pretty arrogant reaction. Come on, he is new to this forum and deserves a better welcome than this.

 

GloatingSwine on the other hand hits the sore spot: surviving.

 

The art of good scouting is to minimize your risk while still providing spots. That means learning your maps and good passive scouting bushes, getting crews with 6th sense, minimizing your exposure time while active scouting, ...

 

But honestly, the best advice would probably be to switch to other tank classes. Light tanks are the most challenging tanks in the game. You may want to try out more newbie friendly classes - for example those TDs Mr. AUS_Stevie/Barrowton seems to favour so much.

Share this post


Link to post
Share on other sites

If you want a higher wn8 and you are not in the top 5% of players, never play a high skill cap tank.

In order for wn8 to seem like a good system it will have to set the bar very high on high skill cap tanks. This

has and will have (after they re balance obviously easy tanks to pad.) the effect of severely punishing

the vast majority of players who drive said tanks. Never play in a tank that has had serious nerfs in the past either.

 

Btw has anyone been seeing how easy wn8 is to pad? Take a look at some of the top wn8 recent on noobmeter and

compare that to those players full record. I would use numbers , but since wn8 uses the scale it does, its difficult to quantify

how much padding is possible. Top players that pad hard, seem to be able to get around 1k extra wn8 at this time from exploiting

the system. The community needs to get out a better version or revert back to a wn7+ (like an earlier version of wn8 was to be.) style metric.

Share this post


Link to post
Share on other sites

I don't know how people have the will to pad in light tanks. My best Wn8 atm is in the T21 and I hated every minute of running it. It has a great gun and towards the end of my run in it I had a bunch of 1600+ damage games in it, just couldn't seem to carry those games where it came down to me vs. a few other heavy/mediums left. 52% win in that tank is expected so my win rate was nothing special. So not sure if expected damage value too low, or if I am still slowly getting better at the game (even if only at mid level tiers), or if it was because the tank plays a lot like a psuedo-medium. But I can definitely see how players can farm wn8 in lights with good guns.

 

Sold it the second I had the T71 unlocked however.

Share this post


Link to post
Share on other sites

Right, but the top WN7 scores were mostly rerolls, so was that actually more valid?

 

I was talking recent results. Reroll is irrelevant when discussing recent results.

Not to mention, wn7 and wn8 can both be boosted by rerolling.

Give it time assuming wn8 is not rejected and you will see a ton of rerolls on the top of the wn8 overall charts as well.

(as if they were not there already, lol.)

 

In fact take a look at the best wn7 rerolls, they are the ones who are really exploiting wn8 the most. It will just take time to see

a full migration.

 

If the community had simply adjusted wn7 to give specific damage expectations per tier instead of average tier, most of the padding of wn7 would have

been eliminated. Considering preator has said wn8 is only a few percent better on average than wn7, and this is before the padding of wn8 had started,

an updated wn7 approach is looking pretty sweet right now. imho

Share this post


Link to post
Share on other sites

If the community had simply adjusted wn7 to give specific damage expectations per tier instead of average tier, most of the padding of wn7 would have

been eliminated. Considering preator has said wn8 is only a few percent better on average than wn7, and this is before the padding of wn8 had started,

an updated wn7 approach is looking pretty sweet right now. imho

 

That's effectively saying that because 5% of WN8 can be padded and it is 10% inaccurate, we should switch to something more similar to WN7, while ignoring the fact that 10% of WN7 can be padded and it is 20% inaccurate. Also, as you don't seem to realise, based on your last two posts, the upper end of the WN8 scale stretches, which is what causes most of the error at that end (if you can actually call it error). Part of the error is caused by major differences in expected damage between tanks, but compared to how WN7 uses damage, WN8 is more accurate for the vast majority of players.

Share this post


Link to post
Share on other sites

That's effectively saying that because 5% of WN8 can be padded and it is 10% inaccurate, we should switch to something more similar to WN7, while ignoring the fact that 10% of WN7 can be padded and it is 20% inaccurate. Also, as you don't seem to realise, based on your last two posts, the upper end of the WN8 scale stretches, which is what causes most of the error at that end (if you can actually call it error). Part of the error is caused by major differences in expected damage between tanks, but compared to how WN7 uses damage, WN8 is more accurate for the vast majority of players.

I said, "I would use numbers , but since wn8 uses the scale it does, its difficult to quantify how much padding is possible.".

That should make it clear I do understand how the scale is not linear. Why do you guys have to resort to treating everyone who

posts criticism like they can't understand BASIC math. Considering all the posts we have traded, I would have thought you would

have given me some degree of respect concerning my understanding of wn8.

 

You are just making numbers up when you say 5% and 10% on wn8. Please show the calculations

you did to establish the amount wn8 is superior to wn7 please! Wn7 can be padded far more than 10% so why did you say 10%?

 

If you even had taken the time to read my post you would have seen I was suggesting wn7 be modified to eliminate the high/low tier trick. Its easy to remove that from the

system, just establish damage per tier similar to how wn8 does per tank. (I am sure you are familiar with this approach as I believe wn8 was going down that route at an early stage.)

Share this post


Link to post
Share on other sites

I said, "I would use numbers , but since wn8 uses the scale it does, its difficult to quantify how much padding is possible.".

That should make it clear I do understand how the scale is not linear. Why do you guys have to resort to treating everyone who

posts criticism like they can't understand BASIC math. Considering all the posts we have traded, I would have thought you would

have given me some degree of respect concerning my understanding of wn8.

 

You are just making numbers up when you say 5% and 10% on wn8. Please show the calculations

you did to establish the amount wn8 is superior to wn7 please! Wn7 can be padded far more than 10% so why did you say 10%?

 

If you even had taken the time to read my post you would have seen I was suggesting wn7 be modified to eliminate the high/low tier trick. Its easy to remove that from the

system, just establish damage per tier similar to how wn8 does per tank. (I am sure you are familiar with this approach as I believe wn8 was going down that route at an early stage.)

 

If what you're trying to get at is for us to calculate a set of confidence intervals, then I'm sure that can be done. Aside from that, the correlation coefficient has already been calculated and posted, and is something like 5-10% (I haven't seen the result in a long time) higher than that of WN7.

 

I am aware of the high/low tier 'trick' solution method you've come up with, and would like to say it was actually exactly the same as the very first thing I sent to Praetor77 concerning WN8. Yes, it may very well reduce the 'paddability' of some outlier tanks (light tanks and a little bit SPGs), but it would also make padding using OP tanks such as the WTE100, or the KV-1S, extremely easy, and would make achieving a high rating in light tanks difficult, which would seem to reinstate one of the issues WN8 was specifically designed to fix. However, the only way to fix both at once is to use a system I've already proposed multiple times. The only drawback to it is that would probably increase the processing time 10-fold. That would render it totally impossible to use with XVM, and possibly many stats sites, unless they got some MAJOR upgrades.

Share this post


Link to post
Share on other sites

I don't know how people have the will to pad in light tanks.

 

 

 

I keep meaning to try a game in my AMX 13 90 with a full APCR load.

 

I have a feeling it would be trolltastic.

Share this post


Link to post
Share on other sites

I keep meaning to try a game in my AMX 13 90 with a full APCR load.

 

I have a feeling it would be trolltastic.

 

Shoot all your ammo, get over 9000 WN8, spend 173K credits on more APCR. Repeat until sleeping under a bridge. :thumbup:

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now

  • Recently Browsing   0 members

    No registered users viewing this page.