Jump to content
You need to play a total of 5 battles to post in this section.
n00bot

WTR is Trash

108 comments in this topic

Recommended Posts

286
[PLPT]
Members
632 posts
6,808 battles

First of all, I used to be a professional data scientist.

The Warships Today Rating is trash and should be completely ignored, IMHO.  They don't even collect enough relevant data from the API for the rating to be meaningful.  For example, they give no consideration to the team composition.  Activities like smoking for an ally or area denial have zero impact.  Damage dealt is severely overweighted.  My best guess is that they use simple correlation statistics.

As a simple example of WTR getting it completely wrong, look at my own numbers for the Harekaze.  I have a 57% win rate in the boat after 80 battles, with only a 716 WTR.  Every single stat of mine in that boat is well below the average, and yet I consistently win with it.  It's not just DD's either.  My Konig win rate is 55% but again every other stat, including WTR, is average or below.  On the flip side, I have an average WTR in the South Carolina, but only a 37% win rate.  WTR seems to be simply tracking the damage I do.

I guess WTR means something if you don't care about winning.  I think it's probably a good indicator of how well you let other people tank for you, so you can farm damage and steal kills.

  • Cool 3
  • Boring 3
  • Bad 5

Share this post


Link to post
Share on other sites
896
[SYN]
[SYN]
Members
2,253 posts
8,578 battles

My tl;dr opinion on this is

 

Winrate = measure of how you play the game

WTR = measure of how you play a ship

 

Sometimes they go hand in hand, sometimes they don't.

  • Cool 1

Share this post


Link to post
Share on other sites
669
[CAFE]
Members
1,395 posts
10,746 battles
28 minutes ago, n00bot said:

First of all, I used to be a professional data scientist.

The Warships Today Rating is trash and should be completely ignored, IMHO.  They don't even collect enough relevant data from the API for the rating to be meaningful.  For example, they give no consideration to the team composition.  Activities like smoking for an ally or area denial have zero impact.  Damage dealt is severely overweighted.  My best guess is that they use simple correlation statistics.

As a simple example of WTR getting it completely wrong, look at my own numbers for the Harekaze.  I have a 57% win rate in the boat after 80 battles, with only a 716 WTR.  Every single stat of mine in that boat is well below the average, and yet I consistently win with it.  It's not just DD's either.  My Konig win rate is 55% but again every other stat, including WTR, is average or below.  On the flip side, I have an average WTR in the South Carolina, but only a 37% win rate.  WTR seems to be simply tracking the damage I do.

I guess WTR means something if you don't care about winning.  I think it's probably a good indicator of how well you let other people tank for you, so you can farm damage and steal kills.

WTR seems to be a pretty good indicator of usefulness in the battle. It does not always line up and can sometimes be misleading, but it does give a good general idea.

  • Cool 1

Share this post


Link to post
Share on other sites
2,544
[PSP]
Members
6,230 posts
8,863 battles
33 minutes ago, n00bot said:

First of all, I used to be a professional data scientist.

The Warships Today Rating is trash and should be completely ignored, IMHO.  They don't even collect enough relevant data from the API for the rating to be meaningful. 

Warships Today can only work with such data as they have access to. Because WOWS is a business as well as a game there are legal issues with data mining. Therefore, Warships Today has to rely on the data WarGaming allows them to use. They have stated themselves that they would like to use more variables. Apparently some data are not reliably available either.

Edited by Snargfargle

Share this post


Link to post
Share on other sites
1,259
Members
2,992 posts
5,204 battles

I can see your point for smoke, although quantifying that value would be difficult. Team composition becomes negligible fairly quickly.

Share this post


Link to post
Share on other sites
286
[PLPT]
Members
632 posts
6,808 battles
4 minutes ago, Slntreaper said:

"I don't like WTR because I have a bad WTR."

I don't really care about my stats or WTR.  I play for fun.  I YOLO with abandon for the LOLz.

But people seem to take this very seriously, and after hearing so much about WTR I decided to check it out.  I think it's really more like

"My WTR is good so it must be valid"

Uh, no

  • Cool 2
  • Boring 1

Share this post


Link to post
Share on other sites
572
[WOLF4]
Members
1,091 posts
4,681 battles
34 minutes ago, n00bot said:

First of all, I used to be a professional data scientist.

The Warships Today Rating is trash and should be completely ignored, IMHO.  They don't even collect enough relevant data from the API for the rating to be meaningful.  For example, they give no consideration to the team composition.  Activities like smoking for an ally or area denial have zero impact.  Damage dealt is severely overweighted.  My best guess is that they use simple correlation statistics.

As a simple example of WTR getting it completely wrong, look at my own numbers for the Harekaze.  I have a 57% win rate in the boat after 80 battles, with only a 716 WTR.  Every single stat of mine in that boat is well below the average, and yet I consistently win with it.  It's not just DD's either.  My Konig win rate is 55% but again every other stat, including WTR, is average or below.  On the flip side, I have an average WTR in the South Carolina, but only a 37% win rate.  WTR seems to be simply tracking the damage I do.

I guess WTR means something if you don't care about winning.  I think it's probably a good indicator of how well you let other people tank for you, so you can farm damage and steal kills.

As a data scientist, you would know that there is no such thing as a perfect model of a system, except the system itself. Your complaints about WTR are nonsense because while you may come up with a "better" metric, you CANNOT come up with a perfect one. 

All we're doing is arguing about how much accuracy we need in the WTR. 

WTR is "good enough" for what it's designed to do, which is give a general idea of where you rank when compared against the rest of the player base. 

  • Cool 2

Share this post


Link to post
Share on other sites
286
[PLPT]
Members
632 posts
6,808 battles
Just now, DingBat said:

WTR is "good enough" for what it's designed to do, which is give a general idea of where you rank when compared against the rest of the player base. 

After seeing the weights page, I can say without reservation that it would be more useful to look at only average damage and ignore WTR.  Mixing in kills simply rewards kill stealers.

Share this post


Link to post
Share on other sites
2,544
[PSP]
Members
6,230 posts
8,863 battles
1 minute ago, n00bot said:

After seeing the weights page, I can say without reservation that it would be more useful to look at only average damage and ignore WTR.  Mixing in kills simply rewards kill stealers.

I've explained in other posts why damage alone is not a good index of a player's team contribution. A player may only do 50k damage but if that damage takes out three DDs then he has contributed a great deal to the team. Conversely, a player can do 500k of fire damage to battleships and see all that re-healed, leaving the battleships to fight on and mostly be no worse for the wear. 

  • Cool 2

Share this post


Link to post
Share on other sites
1,259
Members
2,992 posts
5,204 battles
4 minutes ago, n00bot said:

After seeing the weights page, I can say without reservation that it would be more useful to look at only average damage and ignore WTR.  Mixing in kills simply rewards kill stealers.

Nah, damage is weighted far more, but kills matter. For 1, getting enemy ships off the board is important. For another, you could solo 3 destroyers and do less damage than soloing a battleship, but the former is better than the latter.

  • Cool 1

Share this post


Link to post
Share on other sites
294
[O7]
[O7]
Beta Testers
1,252 posts
7,734 battles
13 minutes ago, Slntreaper said:

"I don't like WTR because I have a bad WTR."

That's how most of these threads boil down too. Is WTR the god-tier of stats, no but it is at least a decent comparison site.

Share this post


Link to post
Share on other sites
286
[PLPT]
Members
632 posts
6,808 battles
Just now, Snargfargle said:

I've explained in other posts why damage alone is not a good index of a player's team contribution. A player may only do 50k damage but if that damage takes out three DDs then he has contributed a great deal to the team. Conversely, a player can do 500k of fire damage to battleships and see all that re-healed, leaving the battleships to fight on and mostly be no worse for the wear. 

Oh I totally agree.  It's just another way the whole WTR system is completely broken.

Share this post


Link to post
Share on other sites
286
[PLPT]
Members
632 posts
6,808 battles
2 minutes ago, cometguy said:

Nah, damage is weighted far more, but kills matter. For 1, getting enemy ships off the board is important. For another, you could solo 3 destroyers and do less damage than soloing a battleship, but the former is better than the latter.

And you could also do 3 damage and get 3 kills.  Another way the rating broken.

Share this post


Link to post
Share on other sites
174
[KENT]
Members
481 posts
3,167 battles
10 minutes ago, n00bot said:

I don't really care about my stats or WTR.  I play for fun.  I YOLO with abandon for the LOLz.

But people seem to take this very seriously, and after hearing so much about WTR I decided to check it out.  I think it's really more like

"My WTR is good so it must be valid"

Uh, no

So why would you complain about WTR if you don't even play this game seriously?

 

Either you play it seriously and want WTR to get easier, or you play this game for fun and don't care.  

Share this post


Link to post
Share on other sites
286
[PLPT]
Members
632 posts
6,808 battles
48 minutes ago, n00bot said:

I guess WTR means something if you don't care about winning.  I think it's probably a good indicator of how well you let other people tank for you, so you can farm damage and steal kills.

Lest people think I'm just complaining without offering a solution, why wouldn't they simply use the academically proven TrueSkill method?

Edited by n00bot

Share this post


Link to post
Share on other sites
286
[PLPT]
Members
632 posts
6,808 battles
2 minutes ago, Slntreaper said:

So why would you complain about WTR if you don't even play this game seriously?

Either you play it seriously and want WTR to get easier, or you play this game for fun and don't care.  

I care because my clan seems to think WTR matters, and that causes problems.  There are excellent players with bad "stats" who get passed up for the competitive team (I'm not one of them and don't care to be on our competitive team)

It also matters when people don't want to div with me because of some broken stat they see.  It's ignorant, really.

  • Cool 1

Share this post


Link to post
Share on other sites
2,787
Members
9,961 posts
52 minutes ago, n00bot said:

First of all, I used to be a professional data scientist.

The Warships Today Rating is trash and should be completely ignored, IMHO.  They don't even collect enough relevant data from the API for the rating to be meaningful.  For example, they give no consideration to the team composition.  Activities like smoking for an ally or area denial have zero impact.  Damage dealt is severely overweighted.  My best guess is that they use simple correlation statistics.

As a simple example of WTR getting it completely wrong, look at my own numbers for the Harekaze.  I have a 57% win rate in the boat after 80 battles, with only a 716 WTR.  Every single stat of mine in that boat is well below the average, and yet I consistently win with it.  It's not just DD's either.  My Konig win rate is 55% but again every other stat, including WTR, is average or below.  On the flip side, I have an average WTR in the South Carolina, but only a 37% win rate.  WTR seems to be simply tracking the damage I do.

I guess WTR means something if you don't care about winning.  I think it's probably a good indicator of how well you let other people tank for you, so you can farm damage and steal kills.

 

Until something better comes along, it's what we have....which is better than nothing....:cap_hmm:

Share this post


Link to post
Share on other sites
136
[WDS]
Members
381 posts
7,026 battles
2 minutes ago, n00bot said:

After seeing the weights page, I can say without reservation that it would be more useful to look at only average damage and ignore WTR.  Mixing in kills simply rewards kill stealers.

What exactly what are kill stealers ? If u come around an island and spot two boats one at full strength and one at 1/4 strength and someone is shooting at it don't u target the weaker boat am I supposed to not shot at it because it is some one else's kill so now is I have to worry about that too . I thought the idea was to sink ships win the game . I'm just asking not being a wise guy 

  • Cool 1

Share this post


Link to post
Share on other sites
572
[WOLF4]
Members
1,091 posts
4,681 battles
14 minutes ago, n00bot said:

After seeing the weights page, I can say without reservation that it would be more useful to look at only average damage and ignore WTR.  Mixing in kills simply rewards kill stealers.

They "teaked" the WTR version for WoT about 8 times during my time playing. There was only two constants:

1. There were ALWAYS some people who thought it sucked

2. There were always people who found how to manipulate it

 

You are never going to find the satisfaction you seek. 

And "kill stealers"? Really?

Share this post


Link to post
Share on other sites
572
[WOLF4]
Members
1,091 posts
4,681 battles
7 minutes ago, n00bot said:

And you could also do 3 damage and get 3 kills.  Another way the rating broken.

How often do you think that's going to happen? You're not talking like a data scientist. 

  • Cool 1
  • Funny 1

Share this post


Link to post
Share on other sites
286
[PLPT]
Members
632 posts
6,808 battles
2 minutes ago, DingBat said:

You are never going to find the satisfaction you seek. 

TrueSkill is a mathematically correct rating based solely on win rate.  That sounds a lot better to me than "some guy" kindof making up some weights.

Share this post


Link to post
Share on other sites
Members
3,774 posts
8,328 battles

I can more agree with the damage factor you can carry with damage win rates how ever if you have rotten luck send get a hole day of losing matches it can affect on your Win rate.  Not to mention players get sick and tired of being bashed because they have a lower or even bad win rate. 

Edited by 0crazy8s0

Share this post


Link to post
Share on other sites

  • Recently Browsing   0 members

    No registered users viewing this page.

×