Jump to content
Sign in to follow this  
You need to play a total of 50 battles to post in this section.
Aetius85

Average calculations in wows-numbers

26 comments in this topic

Recommended Posts

[BATES]
Players
1,120 posts
14,525 battles

Someone can explain how average PR is calculated in wows-numbers?

 

Last 4 games

 

image.thumb.png.5ef910e9124822a9e0265e9acc920d12.png

 

image.png.df42416005a148539bda286f0247662c.png

 

How can be the average PR 1212 if the 4 values are around 1500? 

  • Funny 1

Share this post


Link to post
Share on other sites
[XTREM]
Players
2,258 posts
16,164 battles

It's because recents take into account the combined result of your recent games.

To point to one facet of the calculation, your session W/R is 50%, which would already result in a drop in session PR when you combine the results. Individual ship PR in the same period notwithstanding. 

 

Basically, each ship having around 1500 PR does not necessarily mean the overall PR for the session will be the same. If 10 ships have green PR but also have red W/R, then the overall stat will not be green, but lower.

 

N.B. PR is calculated using dmg - kills - W/R. dmg being the most important. 

  • Cool 2

Share this post


Link to post
Share on other sites
[VIBES]
Players
1,249 posts
23,058 battles
3 minutes ago, Miragetank90 said:

It's because recents take into account the combined result of your recent games.

To point to one facet of the calculation, your session W/R is 50%, which would already result in a drop in session PR when you combine the results. Individual ship PR in the same period notwithstanding. 

 

Basically, each ship having around 1500 PR does not necessarily mean the overall PR for the session will be the same. 

That's...not clear at all, I'm afraid.

  • Cool 1

Share this post


Link to post
Share on other sites
[XTREM]
Players
2,258 posts
16,164 battles
2 minutes ago, tocqueville8 said:

That's...not clear at all, I'm afraid.

 

I added an example. Which part is confusing? 

Share this post


Link to post
Share on other sites
[VIBES]
Players
1,249 posts
23,058 battles
40 minutes ago, Miragetank90 said:

I added an example. Which part is confusing? 

What example?

 

52 minutes ago, Miragetank90 said:

It's because recents take into account the combined result of your recent games.

Right, so an average of 4 games, 2 of which with 1.5k PR and 2 more with an average of 1.5k, should also be around 1.5k, I should imagine...

Clearly the PR of those 2 T-61 games counts more (twice as much, I presume) than the one of the single Siegfried and Scharn B games, but how can the overall average be worse than all of them?

 

52 minutes ago, Miragetank90 said:

If 10 ships have green PR but also have red W/R, then the overall stat will not be green, but lower.

Why? You can have purple PR and red WR, in one game or even over a short session. Are you saying that the "average PR" combines both PR (which also includes WR) and WR again, so WR is effectively counted twice?

Share this post


Link to post
Share on other sites
[TORAZ]
Beta Tester
15,211 posts
21,908 battles

Account (aka global) PR differs from ship PR.

  • Cool 1

Share this post


Link to post
Share on other sites
Players
5,398 posts
24,445 battles
4 hours ago, Aetius85 said:

How can be the average PR 1212 if the 4 values are around 1500? 

It looks like a clear mistake for me on first glance. You can always write to Wiochi and ask him about it. Last time when I had similar thing, he answered me where was a problem.

 

ps. but analyzing other similar cases, average PR may be different then it looks like obvious for us in wows-numbers.com

But ex. in proships.ru is an exact average match of number of games x PR +(..) /all games, so it should be average 1556 for those games.

Share this post


Link to post
Share on other sites
Beta Tester
2,262 posts
20,290 battles
Vor 3 Stunden, Miragetank90 sagte:

Which part is confusing? 

The part where OP's screenshot shows amount of recent games of 4. Each of these 4 games has a higher PR value than the average for the 4 games.

Either the values are wrong or the labels.

Share this post


Link to post
Share on other sites
[BATES]
Players
1,120 posts
14,525 battles
3 hours ago, Odo_Toothless said:

It looks like a clear mistake for me on first glance. You can always write to Wiochi and ask him about it. Last time when I had similar thing, he answered me where was a problem.

 

ps. but analyzing other similar cases, average PR may be different then it looks like obvious for us in wows-numbers.com

But ex. in proships.ru is an exact average match of number of games x PR +(..) /all games, so it should be average 1556 for those games.

Thanks for suggestion. Will try to reach him and will let you know :)

Share this post


Link to post
Share on other sites
[NEXT]
Players
7,159 posts
12,844 battles
6 hours ago, Aetius85 said:

Someone can explain how average PR is calculated in wows-numbers?

 

Last 4 games

 

image.thumb.png.5ef910e9124822a9e0265e9acc920d12.png

 

image.png.df42416005a148539bda286f0247662c.png

 

How can be the average PR 1212 if the 4 values are around 1500? 

Have you checked the explanation? Nvm, someone posted that

 

https://wows-numbers.com/personal/rating

  • Cool 1

Share this post


Link to post
Share on other sites
[ADRIA]
Players
335 posts
11,448 battles

There is no "average PR" in wows-numbers, only "overall PR". The formulas are in the description that's been linked above. "Average PR" is a misleading label.

Share this post


Link to post
Share on other sites
Players
2,943 posts
34,690 battles

Anyone knows how the experience (not pr, wr) is calculated? Does it count the amount of total battles a player has in their account while calculating a game experience at the end of the match?

I played a game on my alt account yesterday with a T10 dd and the after game experience on it seems very low (ss attached). On the other hand, i get higher experience number with even lower tier ships on my main account which has around 27K random battles. That's why i think, experience calcs might have to do with overall amount of battles a player possess, but i'm not sure and may be wrong;

 

 

alt1.JPG

Share this post


Link to post
Share on other sites
[CHEFT]
Players
744 posts
8,032 battles

@Execute0rder66 xp is base xp * premium (if you have it). Therefor the value is pretty worthless, at least the overall one. As long as you know of the person has premium or not, you can calculate the actual recent hase xp.

  • Cool 1

Share this post


Link to post
Share on other sites
[BATES]
Players
1,120 posts
14,525 battles
1 hour ago, Pikkozoikum said:

Have you checked the explanation? Nvm, someone posted that

 

https://wows-numbers.com/personal/rating

Thanks for the link, but I am not asking what is PR, I already know the formulas. My question is about how the "Recent" one is calculated, as I thought it was an average value based on the values of the recent battles detail (second screenshot), as the WR one (4 battles, 2 wins, 2 lost = 50%)

Share this post


Link to post
Share on other sites
Players
2,943 posts
34,690 battles
8 minutes ago, geschlittert said:

@Execute0rder66 xp is base xp * premium (if you have it). Therefor the value is pretty worthless, at least the overall one. As long as you know of the person has premium or not, you can calculate the actual recent hase xp.

Pfft, i just checked how it is calculated after reading your post and it's really not a very meaningful number to see how you did in the battles :cap_haloween:

 

XP   =  XPB • At • (1 + Σ MXP) + ZXP

 

This is the formula and it confirms that it is worthless.

 

Share this post


Link to post
Share on other sites
[CHEFT]
Players
744 posts
8,032 battles
3 minutes ago, Execute0rder66 said:

Pfft, i just checked how it is calculated after reading your post and it's really not a very meaningful number to see how you did in the battles :cap_haloween:

 

XP   =  XPB • At • (1 + Σ MXP) + ZXP

 

This is the formula and it confirms that it is worthless.

 

Blame the API which doesn't provide the actual base xp. I wish it did...

  • Cool 1

Share this post


Link to post
Share on other sites
Players
2,943 posts
34,690 battles
Just now, geschlittert said:

Blame the API which doesn't provide the actual base xp. I wish it did...

Exactly.

Share this post


Link to post
Share on other sites
[CHEFT]
Players
12,666 posts
9,841 battles
39 minutes ago, Aetius85 said:

Thanks for the link, but I am not asking what is PR, I already know the formulas. My question is about how the "Recent" one is calculated, as I thought it was an average value based on the values of the recent battles detail (second screenshot), as the WR one (4 battles, 2 wins, 2 lost = 50%)

 

It adds all values together first. So it adds 2x your T61 games + the siegfied and the Scharnhorst B. Then it does the same for the expected values. And thats why your PR drops:

image.png.7d38eea40f7d37dcc7fe5469a22f578c.png

Thats your avg damage/kills/wins in the session

The expected values for the 3 ships combined is

image.png.f86fab2b6b2cc228d197a5f92c99ea33.png (some rounding errors occured because my excel did something, but its fairly close)

So because your Siegfried damage is rather low, you fall behind in the expected damage. Because you lost 2 games with T61 you also fell behind in the expected wins. Only in frags you are 0,5 ahead. The Siegfried is definetely the one dragging your PR down for that session, because its PR is heavily inflated due to the win and the 2 frags, but lacks damage.

Because you are around the avg expected values, your PR is also average.

Share this post


Link to post
Share on other sites
[THESO]
Players
1,671 posts
19,543 battles
12 hours ago, Aetius85 said:

Someone can explain how average PR is calculated in wows-numbers?

 

Last 4 games

 

image.thumb.png.5ef910e9124822a9e0265e9acc920d12.png

 

image.png.df42416005a148539bda286f0247662c.png

 

How can be the average PR 1212 if the 4 values are around 1500? 

 

First of all I'd like to bring @asalonen into the discussion. He's working on a tool and certainly knows more about PR than me.

 

To explain your question: The PR is calculated based on the performance of all players. Your PR on say the T-61 is based on the specific performance of all T-61 players. They define the expected values for a ship and your performance is calculated in relation to that. Assume for a second they are rather mediocre players, then it's much easier to get a high PR. On the other hand a ship like the Siegfried is played by more seasoned players, since it takes RP to get it and is not the first choice among the RP ships. So the Siegfried likely has pretty good overall numbers driven by the good performance of its above average players. It will be much harder to have a good PR on the Siegfried. In your case your good performance on the T-61 evens out the PR despite the 0% winrate, while it's comparable on the Siegfried even tho you won.

 

The daily or even overall performance however is calculated over all the battles It does not take into account the specific ship types. You can play the most overpowered ship all day long and still your result will be treated as if you had just played the average ship and put into comparison with the results of all players in that time frame.

To put it differently: Your PR on the ships is ship-specific. You daily PR is not calculated as an average of the ships-PRs you played that day, but as an overall number of your battles played in comparison to all the battles played by all players. That is why the numbers differ.

  • Cool 1

Share this post


Link to post
Share on other sites
[XTREM]
Players
2,258 posts
16,164 battles
5 hours ago, Johnny_Moneto said:

The part where OP's screenshot shows amount of recent games of 4. Each of these 4 games has a higher PR value than the average for the 4 games.

Either the values are wrong or the labels.

8 hours ago, tocqueville8 said:

Why? You can have purple PR and red WR, in one game or even over a short session. Are you saying that the "average PR" combines both PR (which also includes WR) and WR again, so WR is effectively counted twice?

 

You can, when it comes to individual ships or even over a session, but not necessarily so, due to the way the recents are calculated with respect to the expected(dmg/kill/winrate) values when you combine the results to make the recents. 

 

Sorry, I'm not very good at explaining technical things. But, others made it clearer above methinks. Like here:

 

40 minutes ago, DFens_666 said:

 

It adds all values together first. So it adds 2x your T61 games + the siegfied and the Scharnhorst B. Then it does the same for the expected values. And thats why your PR drops:

image.png.7d38eea40f7d37dcc7fe5469a22f578c.png

Thats your avg damage/kills/wins in the session

The expected values for the 3 ships combined is

image.png.f86fab2b6b2cc228d197a5f92c99ea33.png (some rounding errors occured because my excel did something, but its fairly close)

So because your Siegfried damage is rather low, you fall behind in the expected damage. Because you lost 2 games with T61 you also fell behind in the expected wins. Only in frags you are 0,5 ahead. The Siegfried is definetely the one dragging your PR down for that session, because its PR is heavily inflated due to the win and the 2 frags, but lacks damage.

Because you are around the avg expected values, your PR is also average.

 

  • Cool 1

Share this post


Link to post
Share on other sites
[SCRUB]
Quality Poster
5,416 posts
20,555 battles

Hi all,

 

A Forumite @Armorin (he even created his own WoWs DataBase for tracking stats long ago as a personal project) wrote about PR calculation some time ago:

 

On 9/27/2018 at 12:31 AM, Armorin said:

I've been taking a closer look at the PR calculations from https://wows-numbers.com/personal/rating

 

For a balanced ship, assuming average 50% win ratio, greater than 35%  win rate is required to start accumulating PR points. 65% win rate creates a multiplier of 2. A 100% win rate would yield a maximum multiplier of 4.33. Obviously beating the odds in a UP ship with less than 50% WR will yield a better multiplier than in an OP ship with greater than 50% PR.

 

Damage requires greater than 40% average damage to start accumulating PR points. 50% greater than average damage yields a multiplier of 1.83. Double average damage yields a multiplier of 2.67. Triple average damage yields a 4.33 multiplier.

 

Kill frags requires greater than 10% average kill frags to start accumulating PR points. Taking the EU server average of 0.68 kill frags (it differs greatly between ships), This yields a 0.64 multiplier. Killing 2 ships gets a 3.16 multiplier, a kraken a mighty 8.06 multiplier.

 

A loss with below 40% average damage, but 2 kill frags obtained gets 948 PR points.

A loss with below 40% average damage, but a lucky kraken obtained gets 2418 PR points.

A loss with 50% greater average damage but no kill frags gets 1281 PR points.

A loss with double average damage but no kill frags gets 1869 PR points.

A loss with triple average damage but no kill frags gets 3031 PR points.

A win, but with below 40% average damage and no kill frags yields 650 PR points.

 

At 65% win rate, only 300 PR points are being obtained from winning. So damage farming and kill stealing really does rake in the PR figures. 

 

 

Leo "Apollo11"

Share this post


Link to post
Share on other sites
[VIBES]
Players
1,249 posts
23,058 battles
28 minutes ago, HMS_Kilinowski said:

To put it differently: Your PR on the ships is ship-specific. You daily PR is not calculated as an average of the ships-PRs you played that day, but as an overall number of your battles played in comparison to all the battles played by all players. That is why the numbers differ.

Ok, that's clear now.

 

Thanks all.

Share this post


Link to post
Share on other sites
[THESO]
Players
692 posts
13,559 battles
5 hours ago, HMS_Kilinowski said:

First of all I'd like to bring @asalonen into the discussion. He's working on a tool and certainly knows more about PR than me.

Not really an expert, but... 

 

I quickly punched the OP's (ping @Aetius85) numbers through a spreadsheet, and got the same results. So on the right you have the ship PR's of 1500+ each and then the overall PR of just 1211. The slight differences in the values are likely due to the expected values having changed between my calculation and the OP's screenshot.

 

https://docs.google.com/spreadsheets/d/1AyLt4DoxtbT_MWr8S1SUHGRE2twucctZxs1hMFrPlaA/edit?usp=sharing

 

It's a bit surpising and somewhat counter-intuitive. :Smile_Default: The ship PR's are all on the lower limit of "very good" (1550) for different reasons. T-61 gets a good PR for having very nice damage and pretty good frags. Siegfrid and Scharnhorst B both get a lot of PR from the 100 % win rate.

 

However, when you sum them up the overall win rate becomes just 50 % -- less then expected with these nice ships. Damage is also average. Frags are still good but not enough to drag the PR any higher than 1211.

 

Moral of the story: Don't do statistics with a small sample size? I wouldn't pay any attention to the "Recent" or maybe even the "7 day" PR. Not the overall PR and especially not the individual ships.

  • Cool 1

Share this post


Link to post
Share on other sites
[BATES]
Players
1,120 posts
14,525 battles
11 minutes ago, asalonen said:

Not really an expert, but... 

 

I quickly punched the OP's (ping @Aetius85) numbers through a spreadsheet, and got the same results. So on the right you have the ship PR's of 1500+ each and then the overall PR of just 1211. The slight differences in the values are likely due to the expected values having changed between my calculation and the OP's screenshot.

 

https://docs.google.com/spreadsheets/d/1AyLt4DoxtbT_MWr8S1SUHGRE2twucctZxs1hMFrPlaA/edit?usp=sharing

 

It's a bit surpising and somewhat counter-intuitive. :Smile_Default: The ship PR's are all on the lower limit of "very good" (1550) for different reasons. T-61 gets a good PR for having very nice damage and pretty good frags. Siegfrid and Scharnhorst B both get a lot of PR from the 100 % win rate.

 

However, when you sum them up the overall win rate becomes just 50 % -- less then expected for these nice ships. Damage is also average. Frags are still good but not enough to drag the PR any higher than 1211.

 

Moral of the story: Don't do statistics with a small sample size? I wouldn't pay any attention to the "Recent" or maybe even the "7 day" PR. Not the overall PR and especially not the individual ships.

Thanks for your explanation (and also to the others who did previously). I don't pay attention too on small sample sizes, but I simply thought that the "Recent" values were calculated as average of the values shown when you open the detail. After checking your spreadsheet formulas now I understood that the "Recent" value is calculated adding all the values from single battles, like if I did a single 183988 dmg battle, with 5 frags and a double win, to be compared with an expected value of 

2.20 189609.40 3.51

 

BR,

 

Aetius85

 

  • Cool 1

Share this post


Link to post
Share on other sites
Sign in to follow this  

×