Main page

I think he is asking for those stats on a per player basis as opposed to the numbers in the OP which are based on battles played. Might be worth looking into.

Snib, on 07 June 2012 - 03:52 PM, said:

That's the point. I already found out the average number of different tanks a player have by dividing the total number of tanks with the total number of players. But still looking for the average number of tiers a player played.

Would be nice if you can find it out.

So, umm... why isn't this thread stickied anymore (again)? There are exactly two useful threads on this forum. One is Overlords Q&A thread and this is the other. Might I be so bold enough to ask the mods/admins to again sticky this one, please?

Swuul, on 07 June 2012 - 05:13 PM, said:

Signed

Sticky this please.

Let's just keep posting here so it stays on the top... One of the few useful threads on this forum.

Agreed, was looking for this thread the other day and couldn't find it.  Why unsticky one of the few truly useful threads.

I was wondering where this thread went... This needs to be stickied again asap

Swuul, on 07 June 2012 - 05:13 PM, said:

Yup, Its an interesting snapshot of the game.
Unless its providing unhelpful facts to the player-base...

ahh my Tin foil hat, why thank you.

Lord_WC, on 07 June 2012 - 11:02 AM, said:

On the other hand, to take another hypothetical extreme, if you have one guy playing 1000 battles with 75% winrate (as he does well in a tank and likes it), and nine guys playing 100 battles each with 40% winrate (as most people do poorly in the tank) then battle weighted-average result would be 58.4% winrate (while player-weighted one would be 43.5%).

Which one better reflects the average winrate that a person buying this tank would expect to get?

What he's asking for isn't wrong, just a different view on things. Stop being so ignorant, please.

P.S. Should probably have lower limit on battles played in this case to remove outliers due to small sample sizes.

Can i add my voice to those requesting that this thread is pinned again? Why was it unpinned?

I find this data very interesting and refer to it often.

ogremage, on 08 June 2012 - 11:30 AM, said:

Nope, you don't understand the point of weighted averages and forget the size of the data.

First, some guys, who like one tank and play it exclusively is pretty low. Nevertheless WG wants to make a balanced game - meaning the choice of tank doesn't affect your chances to win. Now if we take your above example is that tank OP? Well, yes, by all means it is. Since it can be used to achieve great results. If you start to reach a high sample size (for example if the server in your example had only 100 players) you really could nerf that tank since it's OP. People will flock to it, people will achieve great results with it since that tank is clearly superior to anything else. Time to learn it doesn't matter.

There's absolutely no reason to disregard battles played - first the sample size is large enough and you can safely disregard those childish claims that 'pro' players will flock to a certain tank to achieve high WR while noobs use some other so they have low WR. In reality the average skill level for each tank is roughly the same.

Second think a bit about matchmaking. In the above example of yours the players had 43,5% WR, so the tank is UP in your books. However in reality their opponents met those 9 guys 900 times and won 540 matches against them but in return they met the one guy 1000 times and lost 750 matches to him.

And in your stead I would be careful lobbing around that ignorant flag. You supporting the idea of throwing out all the basic statistical ideas then interpreting them the same way as before makes you pretty much something you just called me.

Anyway, why don't you start and deduct every second player's WR from the grand total for german players? That would be a low number as well. And you can interpret it the same way as before - as a win ratio. However it would be not.

JayJay4466, on 08 June 2012 - 11:42 AM, said:

What he said, I used the values to justify my opinions on the 0.7.4 patch on my blog ONCE i managed to find the post again...

"....the French premium SPG takes the expected massive hit as it becoms available to the unwashed masses..."

I will not read your comments about tank's performance while drinking coffee.........

Lord_WC, you are wrong or irrelevant in most of what you write, but I cannot be bothered to dissect it paragraph by paragraph, as noone else would find it interesting, and it would derail the thread for no good reason.

Player-weighted statistics provide interesting information, just like battle-weighted ones do. Great Snib agrees. You are the only one who doesn't as you seem to have some strange bias against them and you hide that behind some appeal to fake authority - "basic statistical ideas" and "every law in statistics".

Sorry, there is no law in statistics saying you are right and we're not.

ogremage, on 08 June 2012 - 12:37 PM, said:

I don't disagree with that statement, but I am after all for the most part providing battle-weighted statistics, as you call them, so I'd think that shows what I believe makes more sense.

I am always only looking at the number of battles played during a snapshot. There are plenty players with very low numbers of battles played in specific tanks during a snapshot, i.e. players sitting at 0% or 100% win ratios during a specific snapshot. You don't really want them weighted the same as your exemplary guy with 1000 battles, trust me. In particular, because due to psychological reasons the guys with the 100% are more likely to keep playing while they are winning while the guys with the 0% are more likely to stop trying due to frustration. That's why there are more 0% players in the sample than 100% players (it's  also the reason why the majority of those players having quit the game permanently after a few battles only have 0% or very low win ratios - nobody likes losing).

Statistics are about big enough sample sizes, and since e.g. win ratio is wins per battle, we need to get sufficiently large numbers of battles to have meaningful data.

What could be done is compare the battle-weighted numbers to player-weighted numbers. However, although I haven't run the numbers, for the reasons stated above I'd take a bet that the player weighted numbers will be lower in nearly all cases.

ogremage, on 08 June 2012 - 12:37 PM, said:

Yet you can be bothered to write a whole post about how you can't be bothered to answer. Really, if you ran out of arguments just don't post anymore. Posting about how you cannot be bothered to post while dismissing someone elses arguments as 'mostly irrelevant' is pretty ignorant you know. Reeks of strawman and ignorance and stuff.

The thing is that for balancing you cannot use player weighted data; it's a weak indicator at best either generally or about a tank's capabilities.

And again just as in my first post, I do understand how people too lazy to revise their own claims have to cherry pick their data to show us how they are right, but in reality it's only pretty funny and noone will take you seriously.

Hey, you still missed 'but the XY nation's tanks are piloted by real pros and the russians are driven by utter idiots that's why they have similar statsqq' part.
Edited by Lord_WC, 08 June 2012 - 03:32 PM.

Lord_WC, on 08 June 2012 - 03:30 PM, said:

"Yeah, well, you know, that's just, like, your opinion, man."

ogremage, on 08 June 2012 - 04:34 PM, said:

And apparently it's the same as WGs and about anyone's who has the slightest understanding of statistics. And that's why people will dismiss your oppinion regarding the indicators by which the game should be balanced.

We have a million useless stickies, but this is the one that's removed? Return this topic to where it belongs, ffs.