Jump to content
You need to play a total of 20 battles to post in this section.
n00bot

Win Rates and Number of Games

38 comments in this topic

Recommended Posts

639
Members
1,170 posts
11,397 battles

We all know WoWS can be streaky, winning a lot of games one night and losing lots the next.  So how many games does it take to get a "settled" win rate?  20? 100? 1000?

Here are the 95% confidence intervals by number of games.  That means if you get a 50% win rate after the given number of games, your true win rate could be anywhere within the given range, and there's still a 5% chance you're even better or worse than the given range.

 

For a 50% Observed Win Rate

Total Games True Win Rate (95% confidence)
30 31.3% - 68.7%
100 39.8% - 60.2%
1000 46.9% - 53.1%
5000 48.6% - 51.4%

 

 

Basically, after 100 games, your win rate could still be +/- 10%.  You could be trash, or you could be unicum.  100 games is simply nowhere near enough to tell.

Even after 5000 games, there's a 3% swing.  Of course this assumes your "true" win rate doesn't change, but most players get better with experience.  You're not going to play 10,000 games without getting better, and your old games will bring down your average.  Reroll accounts seem to be popular for vanity... Just keep resetting your account until you luck into the upper range of your true WR.

IMO the stats sites and things like Matchmaking Monitor are misleading at best, and trash at worst, because they really only report data but don't actually do any statistics on that data.  Let's just call them data sites instead.  Next time you see a bunch of 40% players on your team using Matchmaking Monitor, just ignore it.  I think MM only uses the most recent 100 games.  Next time some jerk brags about being "#1 in NA for Des Moines" just ignore it.  WoWS-numbers only uses a 30-game minimum for their leaderboard.  That player almost certainly got lucky, and won't touch that boat again for fear of dropping off the Leaders of Luck board.  Notice how all the "top" players for a ship have near the minimum number of games to be listed.  Please don't call them stats sites.  They are merely data reporting sites that do zero math or analysis.

  • Cool 8
  • Funny 3
  • Confused 1
  • Boring 2

Share this post


Link to post
Share on other sites
7,860
[PSP]
[PSP]
Members
12,625 posts
33 minutes ago, n00bot said:

Please don't call them stats sites.  They are merely data reporting sites that do zero math or analysis.

Finally someone who understands how statistics work!

People around here seem to think that a player's "color" actually has statistical meaning, rather than just being an arbitrary cutoff point that the host of a data farming website thought up. For instance, there almost certainly is no statistical significance between these classes: 52.94% Good 51.44% Average

Edited by Snargfargle
  • Cool 1

Share this post


Link to post
Share on other sites
Members
1,493 posts
10,830 battles
1 hour ago, n00bot said:

That player almost certainly got lucky, and won't touch that boat again for fear of dropping off the Leaders of Luck board.

everyone is equal skill levels, we just got lucky. The classic forum post! nobody is better then me they just got lucky.

1 hour ago, n00bot said:

Notice how all the "top" players for a ship have near the minimum number of games to be listed

yes because one bad game could knock you down to rank 2 that is how the ranking system works, your PR value is calculated by the set algorithm of the wows-numbers site which takes in to the equation win rate, kills and average damage, i dont see a luck factor in this equation though i am afraid sir.

win rate is something that is judged differently by different skilled level players, for example yourself just looks at overall win rate but this can be easily manipulated by playing in divs. A 3 man div should be winning 85-90% of games allowing for that 1 game where your team just bends over and theres nothing you can do. A true giga brain chad will look at solo win rates to get an honest visual of how good a player is, making in impact to win a game and carry the "team"  as a solo player requires a far higher skill level then joining a 3 man div and clicking battle to win.

anyone with around a 60% or higher solo win rate is a good competent player able to achieve wins even with his "team" resisting attempts to win the game.

b678deaa8a69ab9250a7d79d7f5a75e0.pngya boi is rank 1 yahagi world wide btw image.jpeg.5467f3c296b882983b578011211ccf27.jpeg

1 hour ago, Snargfargle said:

People around here seem to think that a player's "color" actually has statistical meaning

this is very true, i would say that 70% of unicum+ players are not good at all, i have seen them play in clan battles which is the best tester for skill levels and they crumble. simply put random battles is not a deep pool of competent players, it is a mixed bag of people that just play for fun and dont want to make it competitive so generally they are not highly skilled players. Basing an opinion of someone being a good player based on random battles is just not accurate, throw them into clan battles and they will reveal just how good they are.

 

 

  • Cool 1
  • Confused 2
  • Boring 1

Share this post


Link to post
Share on other sites
639
Members
1,170 posts
11,397 battles
10 minutes ago, ITZ_ACE_BABY said:

b678deaa8a69ab9250a7d79d7f5a75e0.pngya boi is rank 1 yahagi world wide btw 

 

 

 

I can see this touched your nerve.  I'm talking about Win Rate here. Damage (PR) is another subject...  (PR is trash, IMO)

Your chart proves my point quite well... of the "top" players you list here, the Win Rate ranges from 54% to 77%.  If it's truly the best 9 players, we should see a much MUCH narrower range of win rates.  Also, it is very very few games.  The player with the most games, 81, has an observed win rate of 64.2%... That computes to a range of 53.8% to 74.6% with 95% confidence.

Don't worry you are still #1 on this list.  You are the best.

Edited by n00bot
  • Cool 2
  • Funny 1

Share this post


Link to post
Share on other sites
Members
1,493 posts
10,830 battles
46 minutes ago, n00bot said:

I can see this touched your nerve.  I'm talking about Win Rate here. Damage (PR) is another subject...  (PR is trash, IMO)

Your chart proves my point quite well... of the "top" players you list here, the Win Rate ranges from 54% to 77%.  If it's truly the best 9 players, we should see a much MUCH narrower range of win rates.

Don't worry you are still #1 on this list.  You are the best.

well no you entirely missed my point, PR is most efficient to be improved playing solo as normally div mates are good and take your damage so my stats i can confirm were all solo, 60% solo win rate in a yahagi which has negative carry potential is a very impressive achievement, especially considering alot of people cant average 75k in tier 10 ships.

i dont see anywhere you touched a nerve i simply pointed out that you were wrong in your assumptions and i corrected you.

yes i know i am number 1 worldwide thats what that list says.

Edited by ITZ_ACE_BABY
  • Funny 1
  • Boring 2

Share this post


Link to post
Share on other sites
10,611
[WOLF3]
Members
27,038 posts
23,838 battles

The more battles the merrier in indicating a player's general competence.

 

IMO, where a player sits at 1k Random battles is important.  He's no longer the newbie scrub learning how to log in and has gained a fair amount of experience to have a number of the fundamentals down, but still has a lot to learn.  At 1k battles, IMO, they're at an important threshold.  Players tend to still be pretty mediocre at 1k battles, but there's room to grow and improve, or they just keep doing what they're doing and never get better.

 

Personally, at around 1.2k Random Battles I sat at around 43-45% WR.  I had my good moments but in general I was worse than I could admit.  But it can be turned around still at 1k battles.  With a lot of work I did exactly that.  One day I got tired of being someone's pinata and decided to change that.

Edited by HazeGrayUnderway

Share this post


Link to post
Share on other sites
7,118
Members
6,858 posts
15,330 battles
4 hours ago, n00bot said:

I think MM only uses the most recent 100 games.  Next time some jerk brags about being "#1 in NA for Des Moines" just ignore itWoWS-numbers only uses a 30-game minimum for their leaderboard. 

2 assumptions that are factually innacurate. For a post based on "data" and "facts" it is very poorly researched.

 

 

  • Cool 1

Share this post


Link to post
Share on other sites
639
Members
1,170 posts
11,397 battles
2 hours ago, Ducky_shot said:

2 assumptions that are factually innacurate. For a post based on "data" and "facts" it is very poorly researched.

 

Please enlighten me.  What does MM use?  Someone told me 100 recent games but I don't use it.

Wows-numbers literally says "Minimum battles: 30" right above the leaderboard.

Share this post


Link to post
Share on other sites
Members
1,493 posts
10,830 battles
2 minutes ago, n00bot said:

Please enlighten me.  What does MM use?  Someone told me 100 recent games but I don't use it.

Wows-numbers literally says "30 games minimum" right above the leaderboard.

there are different minimum games per couple of tiers, so if you look at tier 10 leaderboards minimum is 80 battles each tier has its own minimum required number of battles as its easier to get fluctuating results at higher tiers.

matchmaking monitor uses every game in that boat you are currently playing in as a display, by default the settings hide boats under 10 battles as this is a small sample size and obviously a new boat so they are learning how to play it but there is an option to show under 10 battles. Also if you make your account hidden it will not show, so if you are this conscious about your low win rate just hide your stats. 

Share this post


Link to post
Share on other sites
7,118
Members
6,858 posts
15,330 battles
6 minutes ago, n00bot said:

Please enlighten me.  What does MM use?  Someone told me 100 recent games but I don't use it.

Wows-numbers literally says "Minimum battles: 30" right above the leaderboard.

Download it and find out, its very customizable. Also, there are multiple options for those types of programs and they each have their own thing.

Do some research and try and figure out the difference here on WoWs Numbers and why it isn't 30 like you think.

image.png.c046459ed8943f8326517a69ce78ae70.png

image.png.122b686fdae7b8b1be2987c0ae8ce5ad.png

Share this post


Link to post
Share on other sites
7,118
Members
6,858 posts
15,330 battles
10 minutes ago, n00bot said:

You have 22 wins in 31 battles for an estimated true win rate between 55.0% - 87.0%.  Hardly a large enough sample to make any scientific claims.

What about when you compare the virtually identical Sims and the extremely similar Mahan? Both of which I am top 5 NA.

Also, I thought your advice was to ignore people who made those assertions. Why aren't you following your own advice and why should we put any stock into anything you said now?

Edited by Ducky_shot

Share this post


Link to post
Share on other sites
5,439
[WOLFG]
Members
28,979 posts
8,218 battles
22 minutes ago, Ducky_shot said:

2 assumptions that are factually innacurate. For a post based on "data" and "facts" it is very poorly researched.

 

(#1 Sims B worldwide here BTW)

As #660 for Normandie, (I have low expectations) I concur!

Share this post


Link to post
Share on other sites
7,118
Members
6,858 posts
15,330 battles
2 minutes ago, Skpstr said:

As #660 for Normandie, (I have low expectations) I concur!

play 5 more games in the Friant dude!!!

Share this post


Link to post
Share on other sites
5,439
[WOLFG]
Members
28,979 posts
8,218 battles
2 minutes ago, Ducky_shot said:

play 5 more games in the Friant dude!!!

I just may do that. Although, now that I'm looking at the leaderboard, watch me lose 5 in a row lol.

Share this post


Link to post
Share on other sites
36
[OO7]
Members
23 posts
5,588 battles
33 minutes ago, Ducky_shot said:

2 assumptions that are factually innacurate. For a post based on "data" and "facts" it is very poorly researched.

 

(#1 Sims B worldwide here BTW)

Agreed with my #77 Kiev :sadpepedab:

 

Edited by Soviet_Bagutte

Share this post


Link to post
Share on other sites
1,415
[SI-YC]
Beta Testers
3,013 posts
5,943 battles

The way I figure it, you're one player out of about a dozen on a team.  This is gonna fairly significantly lock you down to the 50% level.  Over a thousand battles, that's gonna sway a percent or two and can influence it a bit either way.

But you have to be exceptionally good over the long haul to move it too much more than that.  Because of the other dozen scrubs on your team.  50-60% is what I consider reasonable human levels over the many thousand battle ranges.  Anything above that is probably a bit of div type stuff or possibly even stat padding.  The rare actual chad unicum happens but it is rare and stat padding is maybe not so rare, so...

They are as they do.

Share this post


Link to post
Share on other sites
7,118
Members
6,858 posts
15,330 battles
1 minute ago, CommodoreKang said:

The way I figure it, you're one player out of about a dozen on a team.  This is gonna fairly significantly lock you down to the 50% level.  Over a thousand battles, that's gonna sway a percent or two and can influence it a bit either way.

But you have to be exceptionally good over the long haul to move it too much more than that.  Because of the other dozen scrubs on your team.  50-60% is what I consider reasonable human levels over the many thousand battle ranges.  Anything above that is probably a bit of div type stuff or possibly even stat padding.  The rare actual chad unicum happens but it is rare and stat padding is maybe not so rare, so...

They are as they do.

Nice that the api tracks solo stats then, eh? 

Share this post


Link to post
Share on other sites
1,415
[SI-YC]
Beta Testers
3,013 posts
5,943 battles
Just now, Ducky_shot said:

Nice that the api tracks solo stats then, eh? 

Oh sure.  If you have the spare time to flip that switch.  I'm far too busy for that with all the posting on the forums and such!

Share this post


Link to post
Share on other sites
Members
1,493 posts
10,830 battles

my solo win rate is nearly 60% but i think based on 90th guys "true" win rate scheme i am only 45% player 

c0462436f472668ce5e0318954ae6af5.png

  • Cool 1
  • Funny 1
  • Boring 2

Share this post


Link to post
Share on other sites
7,118
Members
6,858 posts
15,330 battles
2 minutes ago, ITZ_ACE_BABY said:

my solo win rate is nearly 60% but i think based on 90th guys "true" win rate scheme i am only 45% player 

c0462436f472668ce5e0318954ae6af5.png

Yeah, he's way better than you with his adjusted rate of 65%

  • Cool 3

Share this post


Link to post
Share on other sites
1,359
[TWFT]
Members
1,345 posts
37,383 battles

For me it is a herculean task to get my WR up with so many battles.  I have always wondered what my WR would be if I reset it.

 
  Overall Recent Last 7 days Last 21 days
Battles 24 353 10 90 259
Win rate
Next level: Great (+1.9%)
54.15%
Very Good
70%
Super Unicum
+ 0.01% 65.56%
Super Unicum
+ 0.04% 61%
Unicum
+ 0.07%
Personal Rating (PR)
Next level: Good (+63)
1 287
Average
1 324
Average
- 1 669
Very Good
+ 1 1 603
Very Good
+ 3
Battles survived 29.75% 30% - 47.78% + 0.07% 43.63% + 0.15%

Share this post


Link to post
Share on other sites
30
[OO7]
Members
36 posts
10,573 battles

That helps my self esteem since I've gotten my #1 spot taken in Bayard, Colbert, and Flint and top 10 in plenty others. Now if people ask me, I'm always #1! 

Edited by GetRekt_X9

Share this post


Link to post
Share on other sites
88
[B-Z]
Members
157 posts
22,031 battles

I agree that 100 battles is random enough to produce a false result. And that result might take another 200-300 battles just to recover to equilibrium. However by 1000 battles I believe your WR clearly indicates your contribution. Sure the number may be off slightly. But not by much.

Share this post


Link to post
Share on other sites
Guest
This topic is now closed to further replies.

  • Recently Browsing   0 members

    No registered users viewing this page.

×