Jump to content
Praetor77

WN8 breakthrough

Recommended Posts

See?  Pointing out the flaws of your system and making a suggestion for improvement = whining and being an annoying and insulting troll.  Its clear you are too caught up in the perfection of your baby to see clearly.  Not only do you not feel the need to prove your metric's validity, at this point you are full on attacking anyone who dares question it.

 

Just stop.

 

I am not attacking anyone except the people who behaved like asses in a discussion. I admit I can also be an ass sometimes.

 

 

Of course WN is not inmune to criticism.

However, for exmaple for the two point system problem. You were an ass about it, and besides displaying a good point (that a two poitn system would be better), you also spewed a whole bunch of fallacies in that thread. I tried to explain why it wasn´t THAT big of a deal to use a one point system, though I admitted from the start that two points would be better. However, like any other possible improvement, it is ME who has to put in the man-hours to make it happen. It´s easy to criticize when that takes you 5 mins of your time. It´s not so easy when it takes you days of analysis and work.

I get that, but putting in work for free doesn't make your project exempt from critique.  And for the record, I've never seen anyone attack you...yet I see you attack people regularly.

 

REALLY? Wow... ok.

 

Anyways, I will try to be more friendly in the future concerning critique. I am definitely not a people-person, or good at foruming. I also have a low tolerance for stupidity, and people spewing fallacies left and right. Properly dealing with people´s concerns in a friendly manner? That was what Crab was on the team for... too bad he´s AWOL. :sad:

 

 

But I would like to repeat this point:

Imagine you have a Masters in science and a high school student comes and starts criticizing your work left and right. Will most of his criticisms be of any actual use for you to improve your work? Hardly. There may be one or two decent ideas you could implement, but most of his objections would stem from his lack of knowledge and understanding of Biology. Also, 90% of his objections you would have already thought of and accounted for. 

 

I don´t pretend to be a statistics expert, and any someone with math/statistical knowledge has jumped on board willing to help I have more than welcomed him and embraced all of his ideas and proposals (in fact WN8 stems from an original idea of bjshnog, who at the time had nothing to do with WN).

Share this post


Link to post
Share on other sites

That you would even say that shows you don't have the education to comment. 

 

Funny you say that. From each of Praetor's posts one can see his scientific training, while your posts...well, you are fantastic at this game. I grant you that.

Share this post


Link to post
Share on other sites

Funny you say that. From each of Praetor's posts one can see his scientific training, while your posts...well, you are fantastic at this game. I grant you that.

Again, that shows your lack of understanding.  Guess which of us is better educated?

 

 

Imagine you have a Masters in science and a high school student comes and starts criticizing your work left and right. Will most of his criticisms be of any actual use for you to improve your work?

Oh, you mean like how retards tell me how I am winning due to luck/hax/warpacks/gold spam/platoons/etc despite thousands of hours of posting replays, public challenges, arguments, and so on?  I do have an idea of trying to reason with the window lickers in this very community.  But not everyone who critiques Wn8 is a window licker.  In fact, most of the objections to Wn8 I have read are more or less valid.

Share this post


Link to post
Share on other sites

I think part of the problem here is that a very lengthy and exhaustive process was recently completed to come up with WN8.  Now it's finally over but the process is starting up again.  That's why there is more resistance to it this time.

Share this post


Link to post
Share on other sites



Oh, you mean like how retards tell me how I am winning due to luck/hax/warpacks/gold spam/platoons/etc despite thousands of hours of posting replays, public challenges, arguments, and so on? I do have an idea of trying to reason with the window lickers in this very community. But not everyone who critiques Wn8 is a window licker. In fact, most of the objections to Wn8 I have read are more or less valid.


Most of them are statistical window-lickers. And 90% of suggestions fall into one of these three categories:

1-Known problem. No way to fix with data available in API.
2-Known problem. Would take many man-hours to fix/improve.
3-Statistically retarded suggestions.

Even if the suggestion does fall into those three categories, and even if the poster is a statistical window-licker, I always respond gracefully as long as the objecter is polite. However, if they start accusing me of being a shithead (which I actually am sometimes), or go into a heated up discussion mode, I normally follow their lead and go into rage-mode. Personal defect, I know. Once gain, it is why Crab wsa so valuable to the WN team aside from his decent statistical knowledge.

Also, like I said before, WN is the best I can do.I know its not perfect, and there are a million ways to improve it. Whenever I have time to dedicate to it I do, when I think a suggestion will improve it big time, I go for it if I have the knowhow to do it. However, like I said, it is pro-bono work, so it is up to me to decide what to implement and what not, due to the work eating up my own time and no one else´s.



I think part of the problem here is that a very lengthy and exhaustive process was recently completed to come up with WN8. Now it's finally over but the process is starting up again. That's why there is more resistance to it this time.

People don´t like change. However, don´t think this is what grew into this shitpoasting feast (I include myself in there).

IMHO, it is mainly people discussing and criticizing something they know very little about. Amongst others, Xelos, KraftLawrence and others have very high statistical knowledge, probably more than myself.  However, they are not criticizing the statistical methodology. Don´t you think they would raise their voice if my methodology was shit/biased/arbitrary?

 

 

Also, bjshnog is intuitively good with numbers. There are a bunch of unicums with high knowledge regarding gameplay, who can directly (although subjectively) assess how good/precise it is. Are they bringing up the same critique over and over? No, they can see that result-wise WN8 is the best we have. In fact I would appreciate MORE feedback regarding WN8. Properly thought out feedback. I asked for feedback from the first moment we started on the WN8 road. I recieved a lot of it from Folter, and other purpls who were willing to chip in and give their two cents about something they understood and knew about (game mechanics).I will need more of this to review the two new exp1 and exp2 tables. 

Share this post


Link to post
Share on other sites

However, if they start accusing me of being a shithead (which I actually am sometimes), or go into a heated up discussion mode, I normally follow their lead and go into rage-mode. Personal defect, I know.

Maybe I just don't see it, but I can't think of a time I've seen anyone attack you.  Everyone I know of appreciates the time you put into WN8, because its the first ratings system I know of to have proven validity.

Share this post


Link to post
Share on other sites

I´ve been accused at least by 50 people of being biased and coming up with a "personally tailored" metric.

 

I´ve been accused by yourself of using incorrect statistical methodology, although I humbly think you are clearly less experienced in statistics than I am. Amogst other things, you also called me a deniar, basically implied I pull numbers out of my ass. It´s all good to cut the crap and try to play nice from here towards the future, but please don´t play dumb. 

 

I quote my first post on your "problem with WN8" thread:

 

 

Hi guys! For some reason this thread escaped me. Garbad I think you bring up valid points, though like Crab has been trying to explain, we already know about them, and others are just based on wrong information. Also, through this thread you have answered yourself, and why it would not be a good idea to balance tank expected stats using a small number of players (or just the top one as you suggested somewhere).

 

 

Since reasonable and/or reachable stats on each tank depend so much on many variable factors, it would likely be a VERY bad idea to base those expected stats on just a few number of people. For example, there were a BUNCH of people who only played the M48 when it was OP, and never played it again. Comparing everyone´s M48 stats to those would be folly.

 

Almost by definition the top 1/top 10 players of each tank are outliers... so basing an expected value on them is not very statistically sound. In a perfect world, the exact same player playing a WTFE100 and a Tetrarch to the exact same effectiveness would get the exact same WN8 score. Sadly, limitations are all over the place, though we do try our best. As Dlur, or maybe precam, or Gloating brought up, the expected stats can be modified any time if we spot aberrations. That is I think an advantage of WN8, we can easily correct and improve the metric, without having to do much work.

 

 

 

This is a biased comparison. If you look at a per-player basis, most players have a tiny amount of games in tier 1-2, which also were played a LONG time ago, at the beginning of WOT careers. Those stats are hardly comparable with their recently played tier 10 tanks.

 

Crab explained it quite well in page 2.

 

The idea behind WN8 was not for the server average of WN8 to be the same for all tanks. That would be easy to do, but it would make a terrible metric. What we tried to do was, ideally, have the exact same player get the exact same WN8 scores for every tank he plays. Kinda. In a nutshell.

 

Does that make sense? Crab is so much better at explaining things than I am...

 

 

Hardly dismissive or irreverent, I think... however, down the line...

 

The point is that average should be average.  The problem isn't that people are performing worse in lights, by definition, the average player performs at an average skill level.  But they get ~40% lower Wn8 scores than an average TD playing at an average skill level.  That's the crux of the issue -- you DEFINED average in a light as worse than average in a TD.

 

And that's why WN8 is basically a contest to see who plays the most high tier TDs.

 

 

This is the kind of thing that lead me to pulling my hair out later in the thread. I really don´t care much about personal attacks, but spreading fallacies like this made me lose my temper. Also, stating something that is completely contrary to what statistical analysis is all about.

 

 

Anyways, not justifying my actions here, I know I lose my temper, and I know why:

 

Most of people who criticize WN are statistical window-lickers. And 90% of suggestions fall into one of these three categories:
1-Known problem. No way to fix with data available in API.
2-Known problem. Would take many man-hours to fix/improve. Some I deem worthy to spend my time on. Others I don´t.
3-Statistically retarded suggestions.

 

People who do this, and also spread fallacies and/or go into rage-mode, makes me also go into rage-mode.

 

 

Also, English is not my first language, and I have seen that many times people attach emotions or attitudes to my posting, that I honestly did not intend to be there. And things can escalate quite quickly from there.

Share this post


Link to post
Share on other sites
There are a bunch of unicums with high knowledge regarding gameplay, who can directly (although subjectively) assess how good/precise it is. Are they bringing up the same critique over and over? No, they can see that result-wise WN8 is the best we have. In fact I would appreciate MORE feedback regarding WN8.

 

Problems with Wn8:

 

1.  The community uses it for things its not intended for (ie, to compare people at a very fine level, daily stats, etc).  This is what most people actually want wn8 to do, even if its not your goal.

 

2.  I've never seen evidence of its reliability in XVM predictions.  I suspect it gives retards too little weight (once you hit a certain mass of retards, your team WILL lose regardless of how skilled the remaining ~5-7 are).  Also too much to low tiers and not enough to top tanks.  Tier 8s just don't carry tier 10 often even if highly skilled.  Since this is the goal of wn8, this needs to be the key measure, not correlation with acct win rate.

 

3.  No way to measure tactics/scouting/exp.

 

4.  I still believe there is class/tier bias -- specifically, because wn8 is primarily driven by dpg/kpg it rewards tanks and players that excel at these things.  But those tanks/players do not necessarily correlate so perfectly with winning.  This leads to wn8 farming in tds/certain op heavies.

 

5.  Judging scouts based on DPG makes as much sense as judging arty on lights per game.

 

6.  Degenerate padding by using bad tanks that unics know they can do far above average in (tiger 1, etc, etc, etc).

 

7.  The two points are too high.  average and top 10% might be better.  I'd like to see some numbers, as this could solve the problem of average being below average, especially on low tier tanks.  Or better yet 6 points -- bot (0 wn8), bad (600), average (1200), good (1800), unicum (2400), best in server (~4000).

 

That's all I can think of atm.

Share this post


Link to post
Share on other sites

I´ve been accused at least by 50 people of being biased and coming up with a "personally tailored" metric.

Deniers gonna deny.  I don't blame you for hating on those tards.

 

 Amogst other things, you also called me a deniar, basically implied I pull numbers out of my ass.

Which...you did.  The expected values are fundamentally made up.  And it IS a problem that the average low tier, light tank player gets a below average score.  That's the whole point of this thread -- you know that as well, which is why you moved to the two point scale (we will see to what degree it fixes it, again, I think you put the two points too high).

 

So its made you angry enough to call my insistence that average should be average is a fallacy...so what?

Share this post


Link to post
Share on other sites

Problems with Wn8:

1. The community uses it for things its not intended for (ie, to compare people at a very fine level, daily stats, etc). This is what most people actually want wn8 to do, even if its not your goal.

2. I've never seen evidence of its reliability in XVM predictions. I suspect it gives retards too little weight (once you hit a certain mass of retards, your team WILL lose regardless of how skilled the remaining ~5-7 are). Also too much to low tiers and not enough to top tanks. Tier 8s just don't carry tier 10 often even if highly skilled. Since this is the goal of wn8, this needs to be the key measure, not correlation with acct win rate.

3. No way to measure tactics/scouting/exp.

4. I still believe there is class/tier bias -- specifically, because wn8 is primarily driven by dpg/kpg it rewards tanks and players that excel at these things. But those tanks/players do not necessarily correlate so perfectly with winning. This leads to wn8 farming in tds/certain op heavies.

5. Judging scouts based on DPG makes as much sense as judging arty on lights per game.

6. Degenerate padding by using bad tanks that unics know they can do far above average in (tiger 1, etc, etc, etc).

7. The two points are too high. average and top 10% might be better. I'd like to see some numbers, as this could solve the problem of average being below average, especially on low tier tanks. Or better yet 6 points -- bot (0 wn8), bad (600), average (1200), good (1800), unicum (2400), best in server (~4000).

That's all I can think of atm.

I like this.

1- It says so in the wiki and the WN8 doc, it is an approximate measure of skill. Daily and 60 day stats should be much better after the implementation of two point system. Not much more I can do I think.

2- The prediction formula is all xvm developers work, I have nothing to do with it actually. I agree with almost everything you are saying.

3- Nothing can be done about it. I would love to.

4- Will be quite less of a problem (if in fact it is a problem), when two point system is implemented. The exp damage for tier 10 TDs goes up quite a bit for purple level.

5- Damage importance has been shown multiple times. Also, aside from your 5 shell challenge and Kewei´s winrate during his tank marathons despite him ignoring everything else(704, 268, etc.), I recently did a study on vbaddict battle result data that indicates that even when playing light tanks, damage is the main contributor to winning. Much more so than spotting damage.

6- Diminished importance iwth two-value system. Also, over time is corrected, and not much of an issue unless games on that tank represent a HUGE chunk of the total games played.

7- I tested a three point system, and it offers absolutely no advantage over two points. The data is quite linear. I will post screenies later showing this.

 

 

Which...you did. The expected values are fundamentally made up. And it IS a problem that the average low tier, light tank player gets a below average score. That's the whole point of this thread -- you know that as well, which is why you moved to the two point scale (we will see to what degree it fixes it, again, I think you put the two points too high).

Nope. Expected values are the result of linear regressions on the data. They may be inaccurate or seem like shit due to trying to fit two different populations with one number, but they aren´t made up. For certain tanks that showed particularly bad results with this method, I did borrow linear models from other tanks/group of tanks to fit the data instead of using the tank data to build the linear model.

 

I moved to the two point scale because I feel it will be a significant improvement, and because I was able to think of a way of applying it (which I hadn´t before).

Share this post


Link to post
Share on other sites

As someone who is a very average player (if that), I want to say THANK YOU to the developers of WN8; Praetor77, bjshnog, Crabeatoff, Mr. Noobmeter, Phalynx and the rest. I have been watching the development for several months.  It is sad to see the shit-storm that the last several pages have been comprised of. You have put too much of yourselves into this to be so disrespected.  Sadly, I don't have the education or math skills to be a contributor, but I appreciate what you are doing. 

 

While some suggest that this should be aimed at the top 25%, please don't ignore the rest of us. Since I started paying attention to stats, my play has improved. Granted, I am but a faint shadow of most who post here, but playing the game gives this middle-aged man some pleasure.

 

Please continue your good work.  Even with my limited math skills, I can see that WN8 is superior to WN7 in many ways.

 

Until someone comes up with a better system, I’ll support your efforts. Even now, you continue to try to find ways to improve WN8 and I applaud the efforts. A project that it ever improving and evolving is the best kind. It shows just how much you care about the results of your work.

Share this post


Link to post
Share on other sites

As someone who is a very average player (if that), I want to say THANK YOU to the developers of WN8; Praetor77, bjshnog, Crabeatoff, Mr. Noobmeter, Phalynx and the rest. I have been watching the development for several months.  It is sad to see the shit-storm that the last several pages have been comprised of. You have put too much of yourselves into this to be so disrespected.  Sadly, I don't have the education or math skills to be a contributor, but I appreciate what you are doing. 

 

While some suggest that this should be aimed at the top 25%, please don't ignore the rest of us. Since I started paying attention to stats, my play has improved. Granted, I am but a faint shadow of most who post here, but playing the game gives this middle-aged man some pleasure.

 

Please continue your good work.  Even with my limited math skills, I can see that WN8 is superior to WN7 in many ways.

 

Until someone comes up with a better system I’ll support your efforts. Even now, you continue to try to find ways to improve WN8 and I applaud the efforts. A project that it ever improving and evolving is the best kind. It shows just how much you care about the results of your work.

Thanks, appreciate the appreciation. :)

Share this post


Link to post
Share on other sites

I like this.

1- It says so in the wiki and the WN8 doc, it is an approximate measure of skill. Daily and 60 day stats should be much better after the implementation of two point system. Not much more I can do I think.

2- The prediction formula is all xvm developers work, I have nothing to do with it actually. I agree with almost everything you are saying.

3- Nothing can be done about it. I would love to.

4- Will be quite less of a problem (if in fact it is a problem), when two point system is implemented. The exp damage for tier 10 TDs goes up quite a bit for purple level.

5- Damage importance has been shown multiple times. Also, aside from your 5 shell challenge and Kewei´s winrate during his tank marathons despite him ignoring everything else(704, 268, etc.), I recently did a study on vbaddict battle result data that indicates that even when playing light tanks, damage is the main contributor to winning. Much more so than spotting damage.

6- Diminished importance iwth two-value system. Also, over time is corrected, and not much of an issue unless games on that tank represent a HUGE chunk of the total games played.

7- I tested a three point system, and it offers absolutely no advantage over two points. The data is quite linear. I will post screenies later showing this.

 

 

1.  But its still what the community does.  Not your fault, perhaps, but if the community misuses to tool almost all the time it has to be addressed.

2.  Why not?  If the goal is rough glance meta predictions, xvm reliability should be key.

3.  I know, but still a flaw.

4.  We will see.

5.  Yet I bested keweis win rate over an extremely large sample with much lower dpg.  Its clearly not as linear as wn8 makes it out to be.  I think exp/game is the key, and I would just run it using normalized exp/game and assume all blue+ players have premium.

6.  I doubt that.  People will be choosing to play crap tanks with low expected values to farm wn8 (is7, etc)

7.  Ok.  I would predict this isn't true for certain tanks, but I'll take your word for it.

 

 

Nope. Expected values are the result of linear regressions on the data. They may be inaccurate or seem like shit due to trying to fit two different populations with one number, but they aren´t made up.

 

The assumption being linear, which we will see if two points fixes it.

Share this post


Link to post
Share on other sites

As someone who is a very average player (if that), I want to say THANK YOU to the developers of WN8; Praetor77, bjshnog, Crabeatoff, Mr. Noobmeter, Phalynx and the rest. I have been watching the development for several months.  It is sad to see the shit-storm that the last several pages have been comprised of. You have put too much of yourselves into this to be so disrespected.  Sadly, I don't have the education or math skills to be a contributor, but I appreciate what you are doing. 

 

While some suggest that this should be aimed at the top 25%, please don't ignore the rest of us. Since I started paying attention to stats, my play has improved. Granted, I am but a faint shadow of most who post here, but playing the game gives this middle-aged man some pleasure.

 

Please continue your good work.  Even with my limited math skills, I can see that WN8 is superior to WN7 in many ways.

 

Until someone comes up with a better system I’ll support your efforts. Even now, you continue to try to find ways to improve WN8 and I applaud the efforts. A project that it ever improving and evolving is the best kind. It shows just how much you care about the results of your work.

 

I'm sorry to say this, but the average player does not matter to the WN development at all. WN development is pretty much entirely focused on squashing the newest form of stat-padders, if you are not a stat-padder, then it doesn't matter to you.

 

All fitted coefficients in the WN system are positive, meaning if you have more of them the better you are, it's the same as all other systems, so if you just looked at the old fashioned system of win rate or efficiency the better you are the better outcome, there really isn't anything special about WN8.

Share this post


Link to post
Share on other sites

I'm sorry to say this, but the average player does not matter to the WN development at all. WN development is pretty much entirely focused on squashing the newest form of stat-padders, if you are not a stat-padder, then it doesn't matter to you.

 

All fitted coefficients in the WN system are positive, meaning if you have more of them the better you are, it's the same as all other systems, so if you just looked at the old fashioned system of win rate or efficiency the better you are the better outcome, there really isn't anything special about WN8.

 

This is an absolute fallacy. There isn´t a single element in the WN8 calculation formula that attempts to single out or treat statpadders any differently. Please, explain to me what you are talking about, because it seems we are looking at different formulas.

 

If you think WN8 is as good as efficiency or winrate, we can stop discussing here, and agree to disagree.

Share this post


Link to post
Share on other sites

1.  But its still what the community does.  Not your fault, perhaps, but if the community misuses to tool almost all the time it has to be addressed.

2.  Why not?  If the goal is rough glance meta predictions, xvm reliability should be key.

3.  I know, but still a flaw.

4.  We will see.

5.  Yet I bested keweis win rate over an extremely large sample with much lower dpg.  Its clearly not as linear as wn8 makes it out to be.  I think exp/game is the key, and I would just run it using normalized exp/game and assume all blue+ players have premium.

6.  I doubt that.  People will be choosing to play crap tanks with low expected values to farm wn8 (is7, etc)

7.  Ok.  I would predict this isn't true for certain tanks, but I'll take your word for it.

 

The assumption being linear, which we will see if two points fixes it.

 

1- We attempted to address it b creating the wnefficiency wiki. Again, not much more can be done. Perhaps add something about small samples to the FAQ?

2- Well, I really don´t find myself knowledgable enough to properly develop such a formula. Maybe Xelos can do something about that. He recievd a bunch of data from Phalynx that may help him come up  with something.

5- WN8 gleaned the relationships from data. XP /game cannot be used. A large number of players have played at some time without premium accounts, and a rating based on that would be absolutely nonesense IMHO. Also, different tanks get very different XPs for doing the same thing, for example, KV-5 vs IS-3. XP is not the solution.

6-Over time, the expected value methodology would increase the values as more people get higher and higher rSTATS. Can´t do anything about Wargaming´s changing meta and buffing/nerfing tanks. I also have no additional data that would allow me to adjust for something like that.

7- I can show you the graphs, they are all very linear. However, I cant produce 350 graphs, tell me which tanks you would like to see.

Share this post


Link to post
Share on other sites

Let me say a few last words and I will let it rest and allow the continual improvement (6 sigma anyone?) to carry on.

 

Something the community at large must certainly realize if they wish to retain any validity to the complaints.......Praetor (et al) is doing the best he can with that tools he has both developed and has been given (I think everyone acknowledges that, even Garbad).

 

Something the community also has to acknowledge is the fact that the complaints, at least these here recently DO hold weight in the absolute spectrum of perfection (if that is even possible, which it isn't).

 

Garbad, and Praetor have both pointed out truths. I certainly see and fully acknowledge the limitations and have never held back the appreciation for his efforts nor anyone else's in the pursuit of making metrics better. In the end, myself, Garbad, and everyone else that has complaints or concerns about the current system agrees that we will have to disagree but in the effort of the greater good, accept the things we can't change and adapt the ones we can.

 

Again, THANK YOU to the WN ratings team for all the thankless work you guys pour into this monster to continue to provide something to base opinions on.

Share this post


Link to post
Share on other sites

I must admit that I never use WN8 ratings when I am comparing myself to other players.

WN doesn't directly tell me anything. Even if I have someones WN I still need to check their stats myself on noobmeter and wotlabs to actually find out what is that player doing. I don't know a single serious clan that uses it for recruitment, most of them rely on CW experience with DPG and WR on CW viable tanks.

 

When using XVM (I don't use XVMs chance to win prediction, it sucks) I only look at players win rates. Yes, WR can be inflated, but why should I care? Teams are full of 45-49%-ers, and if there is someone in my team that has a recent WR of 60% then that is the person I will rely on the most. Is he a fake unicum? Probably. Do I care? Hell no, why should I, everyone else in the team is too dumb to be even a fake unicum.

Share this post


Link to post
Share on other sites

I must admit that I never use WN8 ratings when I am comparing myself to other players.

WN doesn't directly tell me anything. Even if I have someones WN I still need to check their stats myself on noobmeter and wotlabs to actually find out what is that player doing. I don't know a single serious clan that uses it for recruitment, most of them rely on CW experience with DPG and WR on CW viable tanks.

 

When using XVM (I don't use XVMs chance to win prediction, it sucks) I only look at players win rates. Yes, WR can be inflated, but why should I care? Teams are full of 45-49%-ers, and if there is someone in my team that has a recent WR of 60% then that is the person I will rely on the most. Is he a fake unicum? Probably. Do I care? Hell no, why should I, everyone else in the team is too dumb to be even a fake unicum.

 

Well you are entitled to your opinion, of course. However, winrate´s flaws are quite evident for use in XVM threat assessing.

 

Is _o7_NUMBER_1_WINRATE_NA_ a better player and a higher threat than kewei? 85% winrate and 1350 WN8 vs  67% winrate and 3700 WN8.

Or comparing two players with lots of battles in low tier, _o7_NUMBER_1_WINRATE_NA_ or Autismspeaks?

 

I guess to each his own. Nevertheless, the WN8 doc on the wnefficiency wiki itself states that neither WN8 nor any other composite metric is a replacement for manual dossier parsing. And many clans use it as a screen for applications, including my own, BADGR. Also, I myself use avg tier+WR+WN8 for threat assesment. Again to each his own. Relevance of your post to the topic and the thread? 0.

If you find WN8 unuseful why come in and post about it? We now know your opinion, have a good day.

Share this post


Link to post
Share on other sites
Back ontopic...
 
exp1 and exp2 table for all tanks, preliminary but pretty close to final, at least with current methodology. Would have to use this to iterate and calculate WN8 and re calculate exp values though, and see if there were any substantial changes.
 
The final column shows approximately how changed the value is, compared to the one data-point method (the current one used in WN8). So if a 1 is there, it hasn´t changed. Values below 1 indicate a reduction and above 1 an increase. Remember exp1 values are supposed to show average stats for the average player, and exp2 is supposed to represent the stats a unicum is expected to get.
 
 
 
 
The rSTATS formula would be something like this for all players who have rSTATS up to 1 (neither the equation, nor the baselines should change much, we will see):
 
(avgSTAT-baseline)/(exp1-baseline)
 
For players above rSTATS 1 the equation would be:
 
(avgSTAT-(exp1 * 2 - exp2)) / (exp2 - exp1)
 
This supposes rSTATS follow a distribution which is modeled quite accurately by two straight lines. One with intercept = baseline and slope = (exp1-baseline), and another with slope=( exp2-exp1) and baseline =(exp1*2-exp2). This works quite well and the plots are quite linear. I will post some plots later, since I have to recalculate all the rSTATs with the new exp table. Regarding rSTAT variation amongst tanks when comparing an average player and a unicum, you can check this table:
 
 
It presents exp2 values, and then simply does a ratio of those values to the exp1 for the same tanks. Ideally a unicum will always do 2 times the amount of rSTATS as an average player on all tanks. Obviously there is variation amongst tanks, but the standard deviation for rDAMAGE (most important stat) and rFRAG is 9% of the mean (0.18). This means at least rDAMAGE and rFRAG are highly linear, and the large majority of tanks have quite balanced exp1 and exp2 stats.
 
 
 
 
I need feedback on the expected tables. Most values look extremely reasonable to me, except for some I mentioned before which I highlighted in yellow. Please look for other oddities, and also please comment on the ones I pointed out, and how you think we should deal with them. My initial opinion is we should leave them the way they are, and let the values drop as more and more "normal" people play those tanks (talking about high exp values for conq GC, STB-1, etc. all the "new" tanks).

Share this post


Link to post
Share on other sites
I need feedback on the expected tables. Most values look extremely reasonable to me, except for some I mentioned before which I highlighted in yellow. Please look for other oddities, and also please comment on the ones I pointed out, and how you think we should deal with them. My initial opinion is we should leave them the way they are, and let the values drop as more and more "normal" people play those tanks (talking about high exp values for conq GC, STB-1, etc. all the "new" tanks)

 

 

Not sure about the expected 7201 values, they'll probably need to be pulled down to E100-esque values to make sense. I don't reasonably see how its expected for a bad E-100 clone to do so much more damage, but I've been wrong before. Seems like an oddity to me; the other CW reward tank (m60) has pretty reasonable expectations.

Share this post


Link to post
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.


  • Recently Browsing   0 members

    No registered users viewing this page.

×
×
  • Create New...