Statistical Outlier? My life on the left corner of the Gauss distribution graph

[PrivateBert 261]

Hey guys,

 

first off: you math contributors are doing a fantastic job here on WotLabs, many thanks for that.

Today I want to ask the question if you think it is possible that there are persons / people that are statistical outliers concerning "luck".

You know that guy who always comes out throwing a "1" when rolling dice. I am that guy. My colleagues and my family all notice this too and it is a running gag already.

They say I should write a book titled "My life on the left corner of the Gauss distribution curve". Funny, isn't it?

Of course they are mistaken, because even for me "luck" evens out. I have very very bad luck with all trivial stuff (games etc.) but A LOT of luck concerning the important stuff (job, health, accidents).

Anyway, bear with me

 

How does this relate to WoT? Well, it's a game, so I seem to play right on the left edge in WoT also.

 

For instance many of my tank winrates do not correspond at all with the metrics of my gameplay qualitiy.

An example is my T-34-3 with 85 battles at the moment:

My damage, experience, spotting, WN8 all is very far above average. My winrate is 37%. THIRTYSEVEN.

And I have many more tanks like that. Many more.

I have 3 Ace Tankers in it (85 battles). The last one for a defeat:

http://www.vbaddict.net/battlereport/privatebert-eu-763ad41bee4611a0b887cbd450528e07/ruinberg-china-t-34-3-74830782825358184

 

A typical evening session looks like this:

I am an aggressive player with way above average spotting numbers btw. So no passive damage farming.

 

Of those 18 battles, 12 have been played with Pref MM 8s.

The other 6 battles were:

2 battles with Tiger II, both Tier10 battles

1 battle with Panther II, Tier 10 also

1 battle with Fury, Tier 8

2 battles with TigerH, both Tier 8

Else my WN8 would be higher, but you can only do so much damage when you are low tier all the time...

 

I know: confirmation bias, small sample size Yeah, that#s what I thought.

So I analyse more and check the metrics of my tanks on vbaddict, especially the average battle tier.

And what is the result: ALL my tanks see higher average battletier than the average player. And the unicum player's average battletiers are all lower than that of the average player...examples:

T-34-3

Fury

Tiger II (730 battles...)

 

These are only examples of my currently played tanks. It is true for all my tanks. The differences are sometimes small, but sometimes significant (as you can see).

 

Can these numbers still be explained? Or am I really living on the left edge?

Also: why do unicums generally have so low average battletiers?

Any other comments to what I posted? What is going on here?

 

 

[elFred 3]

Try to play off peak time and avoid EU2 after CW (to avoid facing to much unicum platoon)

 

Plus sample size is likely important for example for my BDR (about 900 games)  I'm below unicum :

 

4.93 for  4.96

 

When with my 50 B I'm over (about 200 games)

 

9.10 for 8.77

 

vbaddict mix stats from other servers than EU ,it has perhaps an impact

[Zinn 1,333]

I'll take a stab and say that your problem isn't mathematical in nature but simply that you are way too aggressive and taking dangerous positions. You are by no means a terrible player but if I were you, I'd start by sorting my tanks by WR, look up some recent replays and see what I did right: Then I'd look up poorly performing vehicles, preferably of similar type and on the same maps to see what doesn't make sense.

 

For instance, why is your Tiger II performing very well over an immense 730 battles with 2k WN8 and a ~58% WR, when your E-75 that's better in every single aspect only gets a 48,3% WR, 1575 WN8 (although that's less relevant for now) and only does marginally more damage in 209 battles? That's well beyond "luck", considering that the E-75 has god-like armor, amazing HP and a punchy gun, even if the DPM is low. The E-100 looks just as bad but that's understandable if you cannot make the E-75 work, since it's more or less the same tanking principle but scaled up.

 

Start making a replay pack of similar vehicles like I mentioned and think carefully about every single thing you do in those battles: Why did you go to that spot on the map? What worked and what didn't? What could you have done better? Why did you win/lose? Also, go to Garbad's and take part in his competition there. It's very educational and you get your answers straight from the horse's mouth.

[PrivateBert 261]

I'll take a stab and say that your problem isn't mathematical in nature but simply that you are way too aggressive and taking dangerous positions. You are by no means a terrible player but if I were you, I'd start by sorting my tanks by WR, look up some recent replays and see what I did right: Then I'd look up poorly performing vehicles, preferably of similar type and on the same maps to see what doesn't make sense.

 

For instance, why is your Tiger II performing very well over an immense 730 battles with 2k WN8 and a ~58% WR, when your E-75 that's better in every single aspect only gets a 48,3% WR, 1575 WN8 (although that's less relevant for now) and only does marginally more damage in 209 battles? That's well beyond "luck", considering that the E-75 has god-like armor, amazing HP and a punchy gun, even if the DPM is low. The E-100 looks just as bad but that's understandable if you cannot make the E-75 work, since it's more or less the same tanking principle but scaled up.

 

Start making a replay pack of similar vehicles like I mentioned and think carefully about every single thing you do in those battles: Why did you go to that spot on the map? What worked and what didn't? What could you have done better? Why did you win/lose? Also, go to Garbad's and take part in his competition there. It's very educational and you get your answers straight from the horse's mouth.

Very helpful answers, at least partly. (Upboat)

I can answer your E75 question: those games are mostly stock grind and then rushing to the E100 as fast as possible. With a fresh crew also (because I kept the good one in the Tiger II).

Also I played that E75 and E100 without having much / any clew. When I play my E75 now, I get much much better results. I just do it very seldom.

E100 is the same: don't play it much anymore, as I am into faster tanks now. When I played it I played badly. Not hiding my lower plate, not sidescaping, not being able to pen anything.

It is better now, but I hate it already. The current meta on EU is yolorush as fast as possible, which leaves me either without damage and a win, or with a loss before I even reach the fight.

 

Since I am at Wotlabs, I read all the guides carefully, watch streams and absorb every bit of knowledge I am presented with. I am here for 2 months now? When I came here my recent was at 1200 WN8 and 48% winrate.

The changes to TDs (which I played well before) broke me and I was trying to get into med and light play.

I think I am slowly getting it, that's why the recent shows blue.

 

TL:DR: I know I did and do a lot of derping and bad play. I am working on it.

But not so bad as to have a 37% winrate in the T-34-3?  So this does not explain that.

Also I tried to be more patient, not be on the front line and snipe with it at the beginning. Does not work at all.

After 2 minutes I am usually 1-5 behind, or worse. I can only somehow make it work when I am aggressive, optimally with a wolfpack of meds and push a flank.

Those are the only games I have a chance of winning. An example is the linked replay above. Even if I lost, it shows the only working strategy.

My team still was so bad to lose anyway, although my wolfpack did a lot of early damage in the critical flank in town and then mopped up the field.

Also it does not explain why my average battletier is so much higher than that of everyone else.

 

Edit: I have replays that show my typical playstyle. There is much Tiger H in it, but I often play it like a med and I always review my sessions:

http://www.vbaddict.net/player-replays/privatebert-eu-763ad41bee4611a0b887cbd450528e07

 

Still, I don't want to stress the "get better, win more" aspect here. I know that and I am doing it.

 

Really, my question is a mathematical one. Am I living on the left edge of the Gauss curve? I think  I do, especially in games. also in WoT.

It is just difficult to grasp. You feel when you are fucked in the ass...I am looking for the numbers to prove it

Questions is: am I fucked by bad luck? Not fucked more than others?

My numbers are off, that's for sure, questions is if it is still within statistical tolerance.

[RichardNixon 835]

The T-34-3 is within tolerance. The 5% confidence interval at 85 games is at least +/-11% winrate, and the T-34-3 is an awful tank, statistically. You'd expect one out of 20 tanks to be outside a 5% confidence interval, so everyone will have a few of them.

 

The Tiger II vs E-75 looks fine. You'd expect 52% with the E-75 with that WN8, with the 5% confidence interval at 200 games being 7%.

 

Seeing a higher tier than average on VBaddict is normal for EU players, because NA has a lower proportion of tier 10 tanks and VBaddict's averages are an EU/NA mix.

 

 

Your overall rStats look like this:

 

rWinC = 1.00

rDmg = 0.95

rFrag = 0.87

rSpotC = 1.21

rDef = 0.84

 

The rDmg/rFrag ratio and high rSpotC indicate a high aggression level. rWinC > rDmg might suggest that's it's effective aggression, but as you have BIA > topgun, the alternative explanation is that you're over-aggressive but platoon-padded.

 

I don't have suitable recent stats for the rStat breakdown, so if you changed your playstyle significantly over time then this may be out of date.

[PrivateBert 261]

Wow, many thanks, greatly appreciated!

Yes, my recent is much higher than my overall. Estimates:

rDmg ca. 1.2 (not sure about that one though, it climbs every day though)

rFrag ca 1.4 (and climbing)

Winrate recent is 54.5%

 

Platoon "padding" goes in the other direction for me. VBaddict says my solo winrate is higher than my platoon winrate.

Only for recent though.

My overall winrate was a bit inflated by company battles and platoons (where I scouted). I played a lot of light tanks.

Also solo I played a lot of Tier6 scout and it saw many T10 battles. Not good for rDmg and RFrag. But I think I was an effective scout.

 

Thank you again for explaining the math behind the numbers. So I currently have a bit of bad luck, but within statistical tolerance.

Bad thing is: my bad luck will stay with me, I will always be on the low side of the statistical tolerance

But I am used to that in games and can live with it, because I have the luck where it matters more .

Also thanks for explaining the difference in battle tiers. I am a bit happier now (Upboat).

[PrivateBert 261]

The last two sessions with ~40 games each, have been absolutely brutal again.

I ran a winrate below 45%. It would have been worse some weeks ago, but my gameplay has improved much and so I was able to win some battles I surely would have lost in former times.

And even though my teams died like flies (or suicidal idiots) all the time, I still managed to run a blue WN8.

Also the matchmaker was hilarious, e.g. I had 26 games in my Fury where I was Hightier ONCE. Else it was always two Tier8 battles, then one Tier7, two Tier8...rinse and repeat.

In around 50 Lorr.40T battles I had Tier10 matches 90% of the time.

People less mentally stable (or less used to living on the left edge) would have snapped. I am sure of that.

My T-34-4 is now sporting a staggering 44% winrate with 1900 WN8. Lorr.40t 47% with 1650 WN8...

 

You really start to doubt the random number algorithm that is used by wargaming. But overall it still seems to even out. But it is hell of streaky,

[SirKnumskull 260]

My T-34-4 is now sporting a staggering 44% winrate with 1900 WN8. Lorr.40t 47% with 1650 WN8...

I may not work for single tanks, but overall your stats don't seem to be so extremely bad, even per session.

 

Look at my stats for some serious mismatch between winrate and WN8 on some tanks. Sherman vs. Pz. IV H for example, very similar stats, very similar tanks, almost 21% winrate difference. Bad luck on individual tanks doesn't matter as long as the overall stats match (and they do in your case IMO).

[PrivateBert 261]

Yeah I know that overall they match. I even wrote that

Still you can get quite mad on bad streaks that last ~100 battles (or more).

[PrivateBert 261]

Today 30 battles, 39% win rate. Fuck my life...

And also can never do all that much, because the battle is mostly over before I can even load a second clip. Just wow.

[PrivateBert 261]

Sorry for the necro But I would please like to ask a followup question.

 

I asked why my average battle tier is so much higher than the average. The answer was:

NA has less Tier 10 than EU, so EU players see higher averages.

 

But that does not make any sense to me.

I am talking average battle tier per tank. For example an Easy8. It really should not matter how many T10 tanks there are.

 

On 90% of my tanks my average battle tier is significantly higher than the average for that tank. Unicums generally have lower battle tier on their tanks than average.

You can really click through ally my tanks and almost every one lies significantly higher than the average.

Something is odd here.