WarshipsToday

Warships Today Rating - a WoWS rating, and warships.today

425 posts in this topic

We've just launched http://warships.today/ in beta.

It does tracking of player statistics in a user friendly way and also calculates Warships Today Rating (WTR) - a rating that considers stats such as damage, frags (ship and plane) to determine player performance.

The individual player page provides a great overview of how the player is doing in different ships and shows a comparison with server average stats. 

It currently has all players on EU, NA, RU servers in the database, so the vehicle stats are representative (not skewed towards those that use the site).

It would be great to hear what you think and any comments or suggestions.

Regards,

Nathan

Share this post


Link to post
Share on other sites

Paging @RichardNixon (Does this not work on shitvision mobile?)

Questions:

What are the rating cutoffs (I have 1910, is this good, bad or indifferent)?

What's your methodology. I know it's expected value based, but is it similar to WN8?

Are you anticipating modding this into the game client? (Please don't)

Nice idea generally though.

Edited by OnboardG1

Share this post


Link to post
Share on other sites

> What are the rating cutoffs (I have 1910, is this good, bad or indifferent)?

Currently - < 300 very bad, 700 bad, < 900 below average, < 1000 average, < 1100 good, < 1300 very good, < 2000 unicum, > 2000 super unicum. They are bound to change (we will generate distribution across the server, then do proper bins).

> What's your methodology. I know it's expected value based, but is it similar to WN8?

It's the same as WN8 to get the expected values (actually using the server averages for those - here is the raw data http://warships.today/json/wows/ratings/warships-today-rating/coefficients ), and then the final formula is as implemented here (this is the actual code being executed):

http://warships.today/public/warships_today_rating.js

 

> Are you anticipating modding this into the game client? (Please don't)

Not really. This has taken 20x more time than we expected as it is, without getting into modding the game client.

We just hope people use it and like it.
 

> Nice idea generally though.

Thanks, glad you like it.

Edited by WarshipsToday

Share this post


Link to post
Share on other sites

Paging @RichardNixon (Does this not work on shitvision mobile?)

I don't think the @ worked. Forum is slow enough to spot new threads anyway.

Notes:

  1. The "expected values" appear to be averages, so it's trivially paddable by sealclubbing.
  2. Cap coefficient looks like an ideological hangover from WoT. I'm fairly sure capping is more important than that in WoWS. Maybe I'll test it later.
  3. Plane vs ship frags method is interesting but probably weighted too strongly towards plane kills, especially at high tiers.
  4. The winrate factor is inappropriate for per-ship stats because people don't play enough battles. It's mostly adding noise.

Share this post


Link to post
Share on other sites

I find it interesting that I'm riding pretty close to the unicum line.... just as I had in tanks.

Share this post


Link to post
Share on other sites

If I keep playing my Battleships for a bit, I will indeed be Unicum. O wow :)

With a 46% win rate haha. But that's a different story.

Share this post


Link to post
Share on other sites
   Notes:
  1. The "expected values" appear to be averages, so it's trivially paddable by sealclubbing.
  2. Cap coefficient looks like an ideological hangover from WoT. I'm fairly sure capping is more important than that in WoWS. Maybe I'll test it later.
  3. Plane vs ship frags method is interesting but probably weighted too strongly towards plane kills, especially at high tiers.
  4. The winrate factor is inappropriate for per-ship stats because people don't play enough battles. It's mostly adding noise.

Thanks for your comments.

 

1. Yes. What are the best practices towards fixing it? My current thoughts are adding a modifier to how far above "expected" low-tier tank values can go. So if you do 2x more damage in t10 ship it counts as 2x, but if you do 2x more damage in t4 it counts only as, say, sqrt(2) times more WRT damage component for that ship.
 

2. Touché. Hated cappers in WoT. So what weights would you suggest?

3. The "plane frag vs ship frag" factor is right now 10, could be increased.
4. Doing two separate ratings overall vs per-ship will confuse people and complicate things. Skipping winrate altogether is also controversial.

Edited by WarshipsToday

Share this post


Link to post
Share on other sites

I don't think the @ worked. Forum is slow enough to spot new threads anyway.

Notes:

  1. The "expected values" appear to be averages, so it's trivially paddable by sealclubbing

I did notice that my Wyoming stats are enormously high compared to my Tirpitz stats for a similar win rate. Im not sure that should be happening.

Share this post


Link to post
Share on other sites

When will players from SEA can see theirs? 

Soon (TM). The server is actually already added, but not visible as we are still mining data for it.

Share this post


Link to post
Share on other sites

Thanks for the work. :)

And I don't think I wanna see mine at this moment as I'm doing so bad in ships. I think I have 40+% wr in it so far. lol

Share this post


Link to post
Share on other sites

1. Yes. What are the best practices towards fixing it? My current thoughts are adding a modifier to how far above "expected" low-tier tank values can go. So if you do 2x more damage in t10 ship it counts as 2x, but if you do 2x more damage in t4 it counts only as, say, sqrt(2) times more WRT damage component for that ship.
 

2. Touché. Hated cappers in WoT. So what weights would you suggest?

3. The "plane frag vs ship frag" factor is right now 10, could be increased.
4. Doing two separate ratings overall vs per-ship will confuse people and complicate things. Skipping winrate altogether is also controversial.

1. Tier modifiers are only a partial fix because average player skill varies significantly within tiers. The general solution is to make graphs of ship stat vs ship-adjusted player stat and use linear regression. For this to work well, you also need to restrict the data to recent games, because players improve over time. There are two basic methods: A) Use heuristics to guess which ships were played recently, B) Take multiple samples over time, and only use battles played over an interval.

It's all tricky stuff. You can use my numbers if you like:

http://forum.wotlabs.net/index.php?/topic/22555-preliminary-wows-expected-values/&do=findComment&comment=595012

Not brilliant, but should be much better than averages. I do have numbers for cap, def and plane kills, but some of the assumptions in the method appear not to hold for defence, and I'm not convinced about the other two. Averages should be ok there.

I don't have enough data for the new ships yet. Substituting values from older ships is probably better for the moment.

 

2. Cap points were worth ~5% weight even in WoT once people stopped padding them for Efficiency. 5% is probably a safe value, although I suspect they're more important than that.

3. On a second look the method seems reasonable here. Midway comes out as 56% weight for plane kills, Yamato as 38%. If anything the issue is at lower tiers (Fuso 10%), but at least it's low-noise.

4. Well, use solo winrate at least. Don't give people any more encouragement to sealclub in divisions.

Share this post


Link to post
Share on other sites

I'll just enjoy being at the top while it lasts.

grats on being top.  and I wonder what will happen as sample size for the high tiers goes up. (personal example: despite being a top 10 Yamato player on the server according to warshipstats, I only have 1434 rating in it (likely due to high expected values from the few people who have it)).  also, I have my best individual ship ratings in the low tier DD's (though that's not unexpected as anybody who understands the vision mechanics can club hard at those tiers)

Share this post


Link to post
Share on other sites

grats on being top.  and I wonder what will happen as sample size for the high tiers goes up. (personal example: despite being a top 10 Yamato player on the server according to warshipstats, I only have 1434 rating in it (likely due to high expected values from the few people who have it)).  also, I have my best individual ship ratings in the low tier DD's (though that's not unexpected as anybody who understands the vision mechanics can club hard at those tiers)

That is pretty much whats happening, the top scores are further from the average in the lower tiers. There was some discussion about it on the thread they made over on the official forums.

http://forum.worldofwarships.com/index.php?/topic/55276-warshipstoday-new-statistics-website-beta

Share this post


Link to post
Share on other sites

That is pretty much whats happening, the top scores are further from the average in the lower tiers. There was some discussion about it on the thread they made over on the official forums.

http://forum.worldofwarships.com/index.php?/topic/55276-warshipstoday-new-statistics-website-beta

My Wyoming/Tirpitz stats are a pretty good example of that. I'm 6th global for damage in my Tirpitz with the same WR as my top global Wyoming damage. The difference is 20% in favor of the Wyoming. I guess it's maybe easier to fail to more damage in the higher tier ships? 

I'm also top EU in WTR according to the site. If you look at my stats and BGrey's stats we're getting big WTR boosts from our expertise in our preferred low tiers. Our T5s in particular (furutaka, omaha, kongo and NY are the big ones here) are all sitting at 2300+. In comparison we each only have one T6+ ship that's over 2k WTR (Pensa for BGrey, Atlanta for me). 

Neither of us have forgotten how to play at unicum WR at higher tiers so something seems a little off. It might be the averages issue that Richard mentioned coming into play.

 

EDIT: NM mestrith explained it really well in the official forums :P

Edited by OnboardG1

Share this post


Link to post
Share on other sites

grats on being top.  and I wonder what will happen as sample size for the high tiers goes up. (personal example: despite being a top 10 Yamato player on the server according to warshipstats, I only have 1434 rating in it (likely due to high expected values from the few people who have it)).

This effect doesn't tend to change much over time. Even if everyone had all the ships, bad players would need to play a higher proportion of their battles in low tiers to make credits, and are more likely to drop down a few tiers because they're outmatched.

WoT expected values would look similar if you used plain averages. What's easier? 250 damage in a T-34 or 2500 in an Objekt 907?

Share this post


Link to post
Share on other sites

We've had some pretty good discussion about it on the official NA forums: http://forum.worldofwarships.com/index.php?/topic/55276-warshipstoday-new-statistics-website-beta/

A couple of issues prevent it from being as successfully uniform as WN8 for tanks, but they seem pretty open to feedback to iron out the kinks.  So far, using individual ship win rate and using average damage without accounting for tier are the two main issues I see.

Share this post


Link to post
Share on other sites

  • Recently Browsing   0 members

    No registered users viewing this page.