155 replies to this topic

[llo1]

Wow, interesting stats.......however, my personal experience doesn’t tally with one aspect that immediately jumps out at me......and that is the T28 Concept...........

I have yet to see anyone play that good in one......individually that is.....maybe simply having one on your team gives you an advantage in winning even if it is driven by the local inebriated half blind player!

 

And the blog seems accurate, with the “mad mode” adjusted mm.
However, you do say that “There is nothing more frustrating in the game than having to fight against the windmills with consistently bad teammates. Skill-based MM would punish anyone who is decent in the game. The analysis shows how significant “rigging” impact already starts for players with 55% WR and above.

My experience shows that I regularly seem to have a selection of bad teammates....and when I do actually get some better ones, the reds are full of bad players....leading to wipe out for one team  following wipe out for one team....

 

Is there any statistical analysis that can possible demonstrate that this increase in total wipe outs across the board are the result of any influence being used?

Following your previous request for just Mad games results, and as I change type of game (and rating) regularly I never submitted anything....would a series of different games (all wipe outs) be of any use to your possible analysis?

[jylpah]

llo1, on 10 November 2020 - 12:09 PM, said:

Wow, interesting stats.......however, my personal experience doesn’t tally with one aspect that immediately jumps out at me......and that is the T28 Concept...........

I have yet to see anyone play that good in one......individually that is.....maybe simply having one on your team gives you an advantage in winning even if it is driven by the local inebriated half blind player!

 

 

Hello llo1,

 

T28 Concept's performance is mostly driven by its relatively better performance for below average players as you can see from the Relative WR vs. player skill level graph. It has rather poor mobility so it limits good players' ability to extraordinary well in it. As you can see the Relative WR graph starts to decrease as the player skill level increases.

 

llo1, on 10 November 2020 - 12:09 PM, said:

Is there any statistical analysis that can possible demonstrate that this increase in total wipe outs across the board are the result of any influence being used?

Following your previous request for just Mad games results, and as I change type of game (and rating) regularly I never submitted anything....would a series of different games (all wipe outs) be of any use to your possible analysis?

 

Initially I had issues in selecting replays of certain type, but that got sorted out. Therefore it does not matter in which order or which replays you send. The scripts can select replays per battle type (Regular, Rating). It is therefore possible to run the scripts for a set of replays and use --extra battle_result parameter. Unfortunately I have to do the analysis myself, since the filtering code relies on MongoDB's query filters to select replays. But as I was writing this, I came up with a feature that could probably let anyone to filter replays based on the "Battle categories". Ah, the endless opportunities to improve code.

 

 

 

 

[llo1]

Jylpah, do I have to send replays individually to wotinspector etc, and then copy each and every link to you, or is there a better way - unfortunately I have an iPad and no access to a pc/Mac at the current time!

[jylpah]

llo1, on 10 November 2020 - 01:07 PM, said:

Jylpah, do I have to send replays individually to wotinspector etc, and then copy each and every link to you, or is there a better way - unfortunately I have an iPad and no access to a pc/Mac at the current time!

 

Just post the replays to wotinspector. You do not need to send me the replays or even inform me (unless you want me to do a custom analysis based on your replays)

[gallabru]

Some times ago, there was no correlation to be seen between wr difference of teams, from one game to the next.
Wipe outs were happening, maybe correlated with wr difference between teams, but nothing more than what could be expected from a random MM.
Things may have changed, it is worth investigation, but I doubt it.

Edited by gallabru, 10 November 2020 - 11:38 AM.

[jylpah]

gallabru, on 10 November 2020 - 01:38 PM, said:

Some times ago, there was no correlation to be seen between wr difference of teams, from one game to the next.
Wipe outs were happening, maybe correlated with wr difference between teams, but nothing more than what could be expected from a random MM.
Things may have changed, it is worth investigation, but I doubt it.

 

So far Regular Battles' MM has been random.

[gallabru]

The surprising part to me is to see that a 2% wr difference between teams is enough to make a 65%er lose 60% of his games.
I would not have thought that it would be enough.

Edited by gallabru, 10 November 2020 - 11:47 AM.

[jylpah]

gallabru, on 10 November 2020 - 01:46 PM, said:

The surprising part to me is to see that a 2% wr difference between teams is enough to make a 65%er lose 60% of his games.
I would not have thought that it would be enough.

 

Nope 50.3%. Their WR was 49.7% in the replays.

 

2% average difference = 7*2% = 14% = 1 extra super unicum (64%) vs. average player (50%).

 

[gallabru]

I am confused.
In your vlog, the first example is with a 67% wr player winning 38% of his games, right ?
I thought that you were saying that the average wr of the red team was 2% higher than the average wr of the green team (minus said player) not that every red player had an average wr 2% above every green team mate (on average).
Could be any repartition, including 6 at 49% and one at 63% indeed. But said player was at 67%, so why not.  

[jylpah]

gallabru, on 10 November 2020 - 02:37 PM, said:

I am confused.
In your vlog, the first example is with a 67% wr player winning 38% of his games, right ?
I thought that you were saying that the average wr of the red team was 2% higher than the average wr of the green team (minus said player) not that every red player had an average wr 2% above every green team mate (on average).
Could be any repartition, including 6 at 49% and one at 63% indeed. But said player was at 67%, so why not.

 

That was the very first data set based on single player. There is much larger data set below that.

 

I am talking about average over team averages. Giving a single player all that 7x2% just illustrates how big the difference can be.

 

Edited by jylpah, 10 November 2020 - 01:29 PM.

[jylpah]

 

Update 7.3 stats added

 

Final stats for update 7.3 have been added.

• Total 230M battles by 484k players recorded in the database during the update 7.3. Please note, these are note all the WoT Blitz players, just those tracked by BlitzAnalysiz().

 

• T28 Concept and Chimera continue to be the best performing tanks in the game. When looking into the details, it is clear that T28 Concept’s performance in the hands of average & below-average players (strong armor) pushes its average Relative WR high. The tank is performing well, but relatively bit less well in the hands of very good players. This is likely due to its poor mobility. At the moment the Relative WR data is not stratified. Stratified sampling would eliminate the impact of the differences in the tanks’ player populations better and give more comparable view of the tanks’ performance.

 

Swedish heavy tank line

 

Update 7.3 brought the Swedish heavy tanks. As it is often the case with new tank lines, those are first played with more experienced & better players distorting the average WR statistics for the tanks. Emil I player population offers a great example:

 

 

Tier VI Strv 74

 

Of the Swedish line, the tier VI Strv 74 seems performing below the average. I liked it personally, but based on the numbers it does not perform that well.

Tier VII Leo

The tier VII Leo seems performing also below the average for the better players. I liked it personally, but maybe its lowish DPM (still higher than the T20) pushed the Relative WR down.

 

Tier VIII Emil I

 

The tier VIII Emil I is the gem of the line and currently the best performing tier VIII Tech Tree heavy tank. Players have been crying it needs to nerfed, but let’s remember that the Emil I is only the 4th best performing tier VIII heavy tank. Its superior performance on the battlefield is also greatly contributed by its better-than-average player population (3.9% higher average WR vs. all tier VIII heavy tanks)

 

 

Tier IX Emil II

 

The tier IX Emil II performed very much the average. I would say the EMil II is perfectly balanced.

 

 

Tier X Kranvagn

 

The tier X Kranvagn was an disappointment for many and it performs the average (below the average for good players). WG often tends to launch new lines with borderline OP-performance. It might be that WG thought the Kranvagn has so much positives that players would play it in any case, and they decided to launch it as balanced. Launching a new line balanced feels almost like launching it nerfed since the standard is borderline-OP.

 

 

 

 

 

Edited by jylpah, 14 November 2020 - 05:53 PM.

[llo1]

Nice summary jylpah.....I particularly like the one about the lower tier Swedish one, up to the Emil 1 as it seems to match my personal experience.....apart from the gun on the Emil 1....for some reason I find that some good well aimed shots seemingly do not pen, whilst conversely some “chanced” shots that I have taken towards reds, knowing that they will know that I could still have 2 more shots, do sometimes pen the target frontally! (NOT the Tiger II though)

...however, personally I also find that without anyone on your team knowing that the actual reload is getting on for 23 seconds, and supporting you whilst you reload, you are way beyond surviving for too long!

Edited by llo1, 13 November 2020 - 03:58 PM.

[jylpah]

 

 

Update 7.4 preliminary stats added

 

 

• In total, there were 105M battles by 419k players recorded in the database during the update 7.4 so far. Please note, these are note all the WoT Blitz players, just those tracked by BlitzAnalysiz().

 

• Chimera has taken the title of the best performing (most OP) tank in the game just ahead of Maus and Emil I. Fair enough, I was wrong, and WG should go and nerf the Emil I a bit :-)

• In general it is great to see Tech Tree tanks taking three spots out of the TOP 5 best performing tanks. The Relative WR figures tend to be bit lower for the new tank lines since many people have been grinding the tank with stock crews and modules. It is not uncommon to see Tech Tree tanks improving their performance stats 1-2 updates since the release. This is despite the fact that I filter out players with too few cumulative battles in a Tech Tree tank to avoid including too many stock tanks in the performance stats.

 

• And as anyone playing tiers VI-VII knows, the Annihilator is performing great.

 

 

[llo1]

Jylpah, Also wrong thread, 

 

I have a slight problem with your results above....not a major on....specifically with the Emil 1.

I have found that since I first got it, and played it, it really is struggling to pen the targets, and seems to be more easily penned by the opposition.

I know that WG won’t have nerfed it already, but I am thinking that the initial OP’ness of the tank was simply due to players inexperienced dealing with the tank.

To that end, can you somehow track the results over time to see if there has been a deterioration in wr from the early days to current games?

[Just_merpug]

llo1, on 27 November 2020 - 08:04 PM, said:

Jylpah, Also wrong thread, 

 

I have a slight problem with your results above....not a major on....specifically with the Emil 1.

I have found that since I first got it, and played it, it really is struggling to pen the targets, and seems to be more easily penned by the opposition.

I know that WG won’t have nerfed it already, but I am thinking that the initial OP’ness of the tank was simply due to players inexperienced dealing with the tank.

To that end, can you somehow track the results over time to see if there has been a deterioration in wr from the early days to current games?

I'm struggling too, mate. Finally had a decent game on Dead Rail, despite our meds not going C and one of our heavies camping spawn. Keep away from the front line, use the ridges, make sure you can hide for the interminable reload. 

[jylpah]

llo1, on 27 November 2020 - 10:04 PM, said:

Jylpah, Also wrong thread, 

 

I have a slight problem with your results above....not a major on....specifically with the Emil 1.

I have found that since I first got it, and played it, it really is struggling to pen the targets, and seems to be more easily penned by the opposition.

I know that WG won’t have nerfed it already, but I am thinking that the initial OP’ness of the tank was simply due to players inexperienced dealing with the tank.

To that end, can you somehow track the results over time to see if there has been a deterioration in wr from the early days to current games?

Initially, new tank lines’ relative WR tends to be less than where it settles since players are still grinding it and many battles are played with less than maxed-out tanks. I try to compensate this by setting a minimum number of total battles in a tank by tier to reduce the impact of the stock-grind battles. 
 

on the other hand, new tank lines average WR tends to be far higher than where it settles for the simple reason that the good, experienced players get the tank first. This can shoot the average WR levels through the roof and people load too much expectations to the tank.
 

Later  the average WR goes down as tank’s player base becomes more normal. On the other hand since larger share of the players have ground through the stock-grind, the relative WR tends to get bit up. I think the Emil I is as good now as it was in 7.3, but as a auto-loader it requires bit skill to manage the reload. The pen is poor, but good that there is at least something “bad” in the tank. 

Edited by jylpah, 27 November 2020 - 10:49 PM.

[jylpah]

UPDATE 7.4 FINAL STATS ADDED

 

Final stats for update 7.4 have been added.

• In total, there were 208M battles by 491k players recorded in the database during the update 7.4 so far. Please note, these are note all the WoT Blitz players, just those tracked by BlitzAnalysiz[].

 

• Chimera has taken the title of the best performing (most OP) tank in the game and T28 Concept maintained its 2nd place. In the final stats Emil I dropped to 5th, but was still ahead of notoriously well-known OP tanks like Annihilator and Object 252U.

 

• Emil I and Emil II got nerfed in update 7.5. It will be in the interesting to see how will those perform during update 7.5.
• Kranvagn remains meh. Maybe the Minsk folks are not fans of Sweden and have decided to leave it at its bit disappointing level.

 

Edited by jylpah, 11 December 2020 - 07:25 PM.

[O_o__O_o__O_o]

keni and b2 only at 60% ? that is a surprise i expected 70+

[erwin10001]

Great work on the stats.

 

What I found interesting are the most played tanks by players in tier 6 and tier 7 are KV-2 and SU-152 respectively. Basically, tanks for camping/sniping. 

 

Weirder still is that AT8 is the top tier 6 tank by relative winrate

[Selthae]

erwin10001, on 12 December 2020 - 12:52 AM, said:

Great work on the stats.

 

What I found interesting are the most played tanks by players in tier 6 and tier 7 are KV-2 and SU-152 respectively. Basically, tanks for camping/sniping. 

 

Weirder still is that AT8 is the top tier 6 tank by relative winrate

 

Most people have no clue how to counter an AT8. If the AT8 gets engaged by 2 or 3 greens they manage. But usually, they try to attack it from the front and meds/lights (and bad heavies) just explode.

 

(45 degrees into the drive wheel, bye bye)

 

Edited by Selthae, 12 December 2020 - 04:21 AM.