Jump to content
EchelonIII

Marks of Excellence

Recommended Posts

Amazing post. I love graphs and well thought through analysis. :3

 

I think we can all safely assume that unless Wargaming releases the formula, the MoE will forever be a mysterious entity that only a few select people are allowed to have at third rating - based solely on the will of RNGesus. Similar to personal rating. Which is very unfortunate as I like the idea behind the MoE a lot.

 

They could've/should've made it related to ace tanker badges instead. Way easier to understand since everyone 'mostly' understands how and when Ace Tanker badges are given out.

Share this post


Link to post
Share on other sites

They could've/should've made it related to ace tanker badges instead. Way easier to understand since everyone 'mostly' understands how and when Ace Tanker badges are given out.

 

I actually rather prefer, that the markers reflect any sort of average performance of a player, than anything even the reddest tomato can get by pure luck, given enough time.

Share this post


Link to post
Share on other sites

You could argue that even the reddest tomato can get the current mark of excellence already as it appears to heavily favour games played and only to some degree average damage. Pick your poison I suppose.

Share this post


Link to post
Share on other sites

You could argue that even the reddest tomato can get the current mark of excellence already as it appears to heavily favour games played and only to some degree average damage. Pick your poison I suppose.

Well, can't argue with that one, of course. However, Ace Tanker needs luck just once and you don't get anything out from repeatedly getting Ace Badge. On the other hand, Mark of Excellence is considerably harder to get at least. It's something and better than nothing (and infinitely better than the WG Rating), I guess. And, to be fair, just playing loads of battles doesn't make the marks appear automagically. I'm stuck on 85.77% on my 59-16 since over 30 battles by now and the last time I got an increase at all (a whooping 0.12%), I had to do 3,000 damage and more than 1,000 spotting damage in a Tier 8 game (we still lost, but that's besides the point I was trying to make).

Share this post


Link to post
Share on other sites

Yeah just had 7 ace tankers in my last 9 battles in my churchill III, all over 2.2k damage. Still didn't budge from 87.1%. Not patting myself on the back but I'm sure if that's ever been done it hasn't happened often. Still moe didn't make a move. Waiting for a complete overhaul in 9.2 at this point.

Share this post


Link to post
Share on other sites

 It seems that the more popular tanks requirements are changing sporadically. I sat at 64.9% in the is7 for two days with averages ranging from 2200 to 2400, before a 5k battle pushed me over to 65. My average before the round was 2213 (lower than the previous day by at least 100). It was frustrating to say the least  

Share this post


Link to post
Share on other sites

So I thought about the previous problem a bit more, and decided, hey, why not actually go ahead and figure out how the service tab percentages should function if the system made sense out of 100 battles. We actually have a damage based performance metric, WN8, with a distribution graph, so reverse engineering this shouldn't be that difficult. Sure, WN8 isn't the same as damage+assists, or whatever the heck MoE are built from, but the two items should correlate incredibly close together. It's not so much apples to oranges as it is apples to... sour apples, sweet apples... WN8 multicolored apples.

 

apple_logo_rainbow_6_color.jpg

 

Errr, not this kind.

 

 

 

Now that graph I linked doesn't have an accompanying chart to grab percentiles, but one can be reverse engineered through photoediting software, by measuring the height of each portion in pixels relative to the population:

 

wn8dist.png

 

So now that we have the distribution, the levels for WN8 to reach each percentile for what should could be in the Service->Awards Tab is pretty simple. The percentiles just need to be plotted against the expected total WN8's of each, which, considering its supposed to be out of 100 battles, its merely the halfway WN8 for each group x 100 (example: The 1500-1599 group should mean a player playing the E-50M would arrive at 1550x100=155,000 after 100 battles).

 

equation.png

 

Thus, we can now use Excels trendy LINEST functions to get a sixth order polynomial equation that fits into the distribution, that should fairly accurately give us a benchmark to compare the running WN8 of these 100 battles played in the E-50M against other players. Plotting these against the Running WN8 and comparing it to what is seen in the service tab, we get...

 

 

simwn8.png

 

Well that doesn't look the same at all. The Sim WN8 line is just a plot of how WN8 climbs relative to a bell curve distribution, which the service tab marks should be doing as well. Climbing a bell curve by accumulating more data should always look more like the blue line than the red. The reason for it being down near zero for so long is It's not until you've accumulated 55,000 WN8 points from playing that you get to climb out of the "red" WN8 portion, then things start to move as you climb through more and more of the population. Eventually if you happen to be on the right side of the bell curve, your rate increase should slow down, as you asymptote towards 1.0. Under such a system my E-50M would have hit the 1st mark after the 59th battle, the second after the 71st, and the 3rd after the 85th.

 

So with the servicetab marks, there's no... climbing of a bell curve distribution... it looks more like a regressive curve for crew skill training. The only thing that's really comparative to the two lines is the slow down as you approach...65% in the Servicetab marks, which is simply too strong of a rate decrease for climbing through the 65th percentile. It should be nearly as fast as .40->.50. There's no asymptote that should exist there.

 

Well, what does this plotting look like against other simulated WN8's? I can crush the data set down so that the 100 battle's WN8 only goes to the 45, 65, and 85 percentiles.

 

 

wn8simed.png

 

So this shows pretty much the same thing. The data from the new lines are pushed farther down and have their rates changed, since these never reach the other side of the bell to the part that gets most muddy, and the orange line simulates what a 65% increase for your markers in the service tab should look like.

 

Now, maybe there is a whole group of players that sit at the very bottom of the bell, a kind of clump where both good and players play the E-50M once or twice and just let it sit. Climbing over them as you go shouldn't be that hard, which would explain the sudden increase in the service tab % because it messes up the bell, but if that's the case, then it should be easier to hit the 65%, 85%, 95% percentiles to get the tank markings.

 

So what was the whole purpose of the MoE?

 

WG is a business and thus they always try and do things that in some way help them make money. The best way to do this is to encourage players to purchase premium+premium tanks. The problem with MoE is that its a backwards idea: Get people to play their own tanks that they already have over, and over, to try and increase their markings. But in doing so, the players don't have as much of a strong need to grind credits or buy gold; it doesn't cost anything, really, to play your own tanks over and over again. There is no premium necessary, no major credit expenditures, nor things to be bought for gold. I'm not sure how this really helps Wargaming if you look at it from that standpoint.

 

Maybe it was designed for player retention purposes. Perhaps WG ran a study thought up there's an economic barrier to where players drop from the game because they already have a mindset of "I'm going to get X Tank and just stop playing for any more tanks" and then they quit the game after a few battles in the tank they wanted. Giving them something to work towards might help them keep the population around. But the problem with the system is that its unrewarding. Think of it like being a kid and getting good grades on your report card. You might be given some money, taken out to ChuckECheese to eat for the night, but then you're left out on the street corner to sleep for a week. It makes no sense it all, just like how WG lets you earn one, two marks, then totally confuses you as to what you're doing wrong. It's more befuddling and frustrating than it is rewarding, especially just for a tiny mark on your vehicle.

 

unobtanium_1.jpg

Unobtanium, the power source of Marks of Excellence revealed.

 

So the system itself seems possibly backwards from the idea standpoint and totally backwards in its execution. Somewhere along the line, someone didn't think something through properly.

Share this post


Link to post
Share on other sites

Intersting theory for sure. Also, could it be that damage, damage assisted and spotting damage aren't the only deciding factors? I just had a 2k damage game with no spotting in the 59-16, but Top Gun and Orlik's in a T8 fight (and half the victims of the autoloader were T8), while surviving with my typical HP pool (roundabout 100HP) and it rewarded me with 0.99%. Certainly a good fight and damage result, but not that overtly outstanding to suddenly jumpstart me that far, compared to the recent drought. Kills? Medals? Tier-difference? Just other people playing worse in general in the last undisclosed measurement-period? I don't get it.

Share this post


Link to post
Share on other sites

Anyone had any more success with this?  I'm curious as to what  it takes to get the second mark on my E-50.  I've been stuck at a hair over 73% for a long while now, and it never seems to move at all.

 

Admiralkird's research is very interesting. I don't see why WG has to be some mysterious with this feature.

Share this post


Link to post
Share on other sites

Of all my tanks I have one MoE on my Leo 1 and one MoE on my E-50. I would like to know where these top 35% players are all doing 2700 dpg+ because I know I'm not the best leo player but most people suck in this thing.

Share this post


Link to post
Share on other sites

Yeah, WGs calculation is definitely bullshit, I kept bumping around between 61 and 64% in my FV304 while getting near blue stats with >1000 dpg. Within the upper 10% on WN8 rating but not even in the upper 35% on damage done, sure.

Share this post


Link to post
Share on other sites

Okay, 869 DPG on the T25 puts me on page 4 of 160 of all players on vbaddict.com and I would assume the people who upload their stats on that page are above average to begin with.

 

I have trouble getting over 66%. A 1300 dmg + 200 spotting game resulted in a 0.3% increase. So there must a lot of really brilliant T25 players out there.

Share this post


Link to post
Share on other sites

On my E-50

247 Battles, 57% winrate

DPG: 2236

Assist: 438

Avg Frags: 1.4

 

From VBaddict:

You have a recent damage ratio of 74.3%

Required average damage for the next marker: 3,464

 

I'm not sure how VBaddict is figuring that number, but seriously, 3464 is required to get a second mark?  That's a pretty extreme requiement for something that says it only beats 85% of the population. 

 

http://www.vbaddict.net/tank/s-m-o-g-g-ru-8872a6a4f30e1c711e2f42f57840486f/germany-1/e-50-40

 

But then looking at this guy (picked at random) he's at 93% with only an average damage of 3000 (4600win8), so VBaddict must just be guessing?  Unless of course his recent games have been much much higher than his average?  Or maybe MoE only counts your best games and ignores your bad games?  Blah.  Why does wargaming have to be so mysterious.

Share this post


Link to post
Share on other sites

On my E-50

247 Battles, 57% winrate

DPG: 2236

Assist: 438

Avg Frags: 1.4

 

From VBaddict:

You have a recent damage ratio of 74.3%

Required average damage for the next marker: 3,464

 

I'm not sure how VBaddict is figuring that number, but seriously, 3464 is required to get a second mark?  That's a pretty extreme requiement for something that says it only beats 85% of the population. 

 

http://www.vbaddict.net/tank/s-m-o-g-g-ru-8872a6a4f30e1c711e2f42f57840486f/germany-1/e-50-40

 

But then looking at this guy (picked at random) he's at 93% with only an average damage of 3000 (4600win8), so VBaddict must just be guessing?  Unless of course his recent games have been much much higher than his average?  Or maybe MoE only counts your best games and ignores your bad games?  Blah.  Why does wargaming have to be so mysterious.

 

The E-50 is one of two grinds I started since marks of excellence were introduced. I am sitting at 134 battles, 62% winrate and ~2.700 DPG which apparently puts me at 79% on EU (first mark). If I hadn't played the tank half-stock at the start and adhered to a less risky playstyle, I might have reached a somehwat higher DPG but I doubt much more is possible at my current skillevel since it's already outperforming any other tier IX med I have ever driven.

 

MoE are utter bullshit. By now it's pretty obvious that there is no correlation between the percentage given in the tooltip and the actual distribution amongst the server population. I would be willing to accept that there are 5% or maybe even 10% of E-50 drivers out there who can top my numbers. But 21% doing more damage than me? Prey tell where are all those wonderful superunicums hiding?

 

I believe WG tried to make the marks hard to get (which I appreciate) and taunted us with a tooltip that contains a blatant lie (which pisses me off).

Share this post


Link to post
Share on other sites

I would like to once again state how utterly stumped I am by how MoEs might work.

 

The second I reached 50% in my Cent 7 MoE progress suddenly ground down to a halt, games with 3500 combined damage adding 0.5% progress and suddenly a 5k combined damage game increases MoE progress by 4%.

 

o_o

 

Dafuq

Share this post


Link to post
Share on other sites

The number of games played in a tank definitively plays a huge role. ~30 more games in my E-50 (DPG stayed more or less the same ~2700) is suddenly enough for 84%,

 

While they seem to be dependant on your performance to some degree, MoE are also a reward for playing the same tank over and over again. Typical WG!

Share this post


Link to post
Share on other sites

Each time I come up with an idea for how it might work I end up staring into blankness.

 

My best guess at this time (which is also quite poor and meh) is that Wargaming designed the system to artificially linear-fie your % against the population. If the system is truly taking your measurements against the rest of the population, then the distribution of players in dpg+spotting per tank in 100 games has to lie on a bell curve - there's no way around this. with the average player somewhere around 1:1 DMG/HP per game. Since your MoE progress % per game increases linearly up to 100 games, WG's formula it has to be messing with this bell curve to make it linear.

 

I think they did this because they wanted player's ranks against the population to increase in increments, rather the slow starting percentage that would rapidly increase as you approached 100 games. My guess is whoever thought up the formula thought it would look better if players moved up in predictable increments for each game, rather than moving up 1% at a time in the first 40 battles, then accelerating to 5% until their 60th battle, where it would taper back off at 1%. This would look weird to players and to the programmer it was undesirable. Thus, they devised a way to mess with the levels of the population to make it so the player increased more like a ladder. I believe their intention was to make your MoE progress look like this:

 

 

lin1.jpg

 

 

But I think somewhere along the line, the programmer messed up the formula and has unrealistic measures of performance for the population, which crushs your MoE performance downward, to look like this:

 

 

lin2.jpg

 

 

Thus, at Tier X the MoE 65th percentile is realistically the ~97th percentile. The 85th percentile is really the ~99.5th. And the 95th is really the ~99.9999th.

 

All it would really take for WG to notice this would be to look at the server data and notice the discrepancy. "Okay since 9.1 came out, 4,000 people have played the E-100 for 100 battles, how many have a mark of excellence? 35% of them should... ohh... ohhhhhh. Only 3% of them do? That's not right. How about two MoE? only five people? No one has 3 MoE? Oh dear. We messed something up."

 

I don't think anyone is checking the system to see if it was implemented correctly and they don't realize there is a problem because they've already moved on to other things. There's probably a coefficient or exponent in there that's in error.

 

 

The number of games played in a tank definitively plays a huge role. ~30 more games in my E-50 (DPG stayed more or less the same ~2700) is suddenly enough for 84%,

 

While they seem to be dependant on your performance to some degree, MoE are also a reward for playing the same tank over and over again. Typical WG!

 

Maybe you had a bad run earlier that its overriding. I've got something around 400 games in my FCM since 9.1 and it plateaued around 200 games in at two markers. I haven't seen number of battles played past 100 be conclusive in bumping it up unless there was a performance increase.

Share this post


Link to post
Share on other sites

I have a theory on how to find how the average damage per game ratio required in order to reach a specific percentage:

 

In this example, we will use vbaddict stats for the Leopard 1, and for the sake of simplicity, I'll only take damage dealt into account, not damage assisted.

Source: http://www.vbaddict.net/tankstats/germany/medium/leopard-1-57

 
So 1976.1 is the average damage dealt.
In my theory, WG takes the average of all the players' average damage dealt in their last 50-100 battles, and derives from that value all the percentages.
Which means that having 1976.1 avg dmg dealt makes you reach 50% on your MoE ranking.
100% corresponds to: 1976.1 * 2 = 3952.2
65%: 3952.2 * 0.65 = 2568.93
85%: 3952.2 * 0.85 = 3359.37
95%: 3952.2 * 0.95 = 3754.59
 
Sounds reasonable for 65%, but managing to do 3.7k average damage over the course of 100 battles is insane.
If my theory is correct, this explains why there are so few people in the "over 85%" bracket, it's because WG doesn't make an actual ranking of the best players.

Share this post


Link to post
Share on other sites

<snip>

 

Although VBA's average damage numbers are higher than the actual population, and this seems to be a very odd way to do it, it seems to add up. I'd imagine if this is how it works, the actual server average for the Leopard 1 is more like 1750 and it has a higher multiplier behind it of 2.5 or so. Measuring your increase against the threshold itself alone would also explain the linearity of the service tab increase, so it really would never be measuring you against the actual % of the population at all at each point, just these set values.

 

The thing I've also been thinking on is that the developers probably focused on the RU server when making this stuff up. Since the RU has a greater disparity between the haves and the have nots, whether it be the much larger pool of casual players in general, those who have good computers, and those who have good internet, you end up with much lower average expected damage values, and much more of the upper-end population able to reach these marks.

 

So while the 95% level for the Leopard 1 might be around 3750 in the EU and NA, it might be around 3,350 or so in the RU, which is difficult to achieve, but definitely doable. 5% of the population might be able to reach it given that its also easier to reach 3.4K dmg. Thus, when they translate this metric over into the other regions, you run into problems where these calculated levels are too high because the population in hardware/skill is much tighter, and the same formula no longer works to separate the population as intended.

Share this post


Link to post
Share on other sites

I made a quick html/js page to easily calculate the server average following my theory, and applied results I got after 100 battles in my leo1.

I'll make another 100 battles streak, if the results will be similar, this will prove my theory and I'll publish the page somewhere.

Ideally someone could make a WoT mod to save the last 100 battles results and the percentage after each, and show the required averages in the game itself.

 

sPfgyRD.png

 

As you can see the server average is actually higher than the one in vbaddict: 1,995.1 dealt + 591.2 assisted = 2586.3.

I have four explanations for this:

  1. there are hidden coefficients that I couldn't find out.
  2. vbaddict mentions "Data for this statistics is based on [...] the last 30 days", the leo1 got a buff recently so it does not reflect the 9.2 buff very well yet.
  3. WG doesn't count scrubs into the average.
  4. vbaddict users are scrubs.

2nd explanation is the most likely.

But even with that first try the results are pretty convincing, so I believe my theory is on the right track.

 

And if you're curious, the equation is as follows:

your average = (total damage dealt + assisted) / 100

server average = (your average * 100) / (MoE percentage * 2)

Share this post


Link to post
Share on other sites

The percentile can go down. I was at 84% with the T54 after 5 amazing games, and then 3 shitty ones brought it to 80%. 

i play really shitty in the FCM 50t.  but somehow got on a roll and got it up to the first mark of excellence.  

now that I am playing back at my normal shitty level, the percent went down below the first mark threshold.  Its nice that they don't take the mark away for that.

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...