Volks mp40 vs stg44 vs no update

There's a statistical difference because worse players are included. That's the reason.





The "top 200" stat is better because it doesn't include unskilled players muddying the results. I'm not sure why there's a necessity that it "favours a specific faction" before realising that this is a bad thing.

In order for that to happen the "worse" player would have to favor OKW in this case.

You also have to keep in mind that what you call "worse" player are not actually bad player.

In addition top "200" does not mean the same thing for each faction since there is a big difference in the number of player for each faction.

Finally keep in mind that the large the number of games the better the results.

Simply read the thread.

Pointless argument when there's several recent 1vs1 tournaments to draw better conclusions from. I don't get how automatch is more relevant than competitive scene...

Pointless argument when there's several recent 1vs1 tournaments to draw better conclusions from. I don't get how automatch is more relevant than competitive scene...

How many times does it need to be repeated that tournaments are -NOT- reliable source of data, because sample size is smaller then error margin of smallest viable sample size?

Tournaments, at best, can show trend, but will never accurately portray balance, unless you need to identify faction in need of dire buffs, because utter lack of that faction in the tournament is a testament to that.

The fact that okw where only picked in 25% of the game in ML 5 is an indication that they are UP.

The fact that okw where only picked in 25% of the game in ML 5 is an indication that they are UP.

It's even more indicative that they didn't even win once, also I'm sure that it's even less than 25%

It's even more indicative that they didn't even win once, also I'm sure that it's even less than 25%

https://docs.google.com/spreadsheets/d/1hGGFJwPoijh4etnnKVJIAg5sF0BGKb9MtLHsFJM-exU/edit#gid=440039274

https://docs.google.com/spreadsheets/d/1hGGFJwPoijh4etnnKVJIAg5sF0BGKb9MtLHsFJM-exU/edit#gid=440039274

I stand corrected, I thought it was even less by what I recalled
The divide between OKW win rate and expected win rate in the top 16 is impressingly bad

In order for that to happen the "worse" player would have to favor OKW in this case.

Regardless of who it "favours" or does not "favour", it can clearly be observed that polluting the data with low-skill players averages winrates. I see absolutely no reason to assume that by introducing useless data (in terms of determining balance), you might have come up with a more accurate display of factional balance than by using more "pure" data (taken only from people who can be assumed to know how to use tools at their disposal).

The more logical assumption is that winrates are determined more by random factors than the actual strengths/weaknesses of factions in this dataset you keep pointing at.

You also have to keep in mind that what you call "worse" player are not actually bad player.

The other players can be of literally any skill level. It's not as though the "all" scraper only expands the lower limit of skill to, say; the top 300. What happens is that it uses matches that have at least one participant who is in the top 200 for that mode, while every other player's rank is entirely ignored. There is nothing stopping any of the other seven players in the match being literally the bottom of the barrel in terms of skill, and in some cases (such as when a good player queues with his friends who may or may not recreationally huff paint thinner) this is almost certainly the case.

It absolutely does include bad players. This data is not useful.


This is stated in the "about" page for the website. Here is the quote.

Difference between top 200 and regular stats?

The regular analysis takes matches from top 200 players as base input too. However the difference is that regular analysis does not care about the other players. In regular analysis you could have one team be top 200 and other team rank 2000. Or in 4v4, only 1 player from the match could be top 200 and other players might be much lower skill.

You have not produced a single argument why the "top 200" are better instead there is an indication that the number might be inflated for some faction due to lower player count. There has even been a thought that one should use % of the top player instead of rank 200 to reduce that effect.

On the other hand from the stat it already clear that a low number of games result in different stat.

If top player matter more than number of games than stat should be of the TOP 10 player only.

The fact that in some games their might be some player of lower level in all mode and factions does not mean anything unless:
you know in how many games that happened and the number is significant
or
some how provide indication the correlation between lower player and a specific faction improving the win rates of certain faction.

In sort do decided to use one set of stat over the other you need to find a convincing explanation of the difference in win rates.

You have not produced a single argument why the "top 200" are better instead there is an indication that the number might be inflated for some faction due to lower player count. There has even been a thought that one should use % of the top player instead of rank 200 to reduce that effect.

Because better players know how to use their faction, and so provide an actually accurate view of factional balance. I've said this multiple times, I'm not sure how you missed it.

Also: I agree that the top % of a faction should be used, rather than the top 200. Again, I said this previously. This is not currently how the data on the site is gathered though. This is also precisely the opposite approach to including terrible player's data in assessing winrates, so I'm really not sure why you're trying to have your cake and eat it, too.

To retort: You haven't provided a single argument as to why using the "all" data would somehow be a more accurate view of balance, despite including games where one side may very well be made up of people who simply do not know how to play the game.

The fact that UKF is so abysmally underplayed is clearly evidence that the faction needs improvement/reworking, in any case.

On the other hand from the stat it already clear that a low number of games result in different stat.

If top player matter more than number of games than stat should be of the TOP 10 player only.

The lower number of games (Not a "low" number. 6273 games is a reasonable dataset for 1v1s, especially over just a month) result in different statistics because the quality of these games is higher.

Skill matters more than raw numbers, but a large number of games is required to weed out anomalous results. If we had thousands and thousands of games from the absolute best player of each faction then that would be great, but unfortunately we do not.

A large number of datapoints is not useful if many of those datapoints are inaccurate.

The fact that in some games their might be some player of lower level does not mean anything unless:
you know in how many games that happened and the number is significant
or
some how provide indication the correlation between lower player and a specific faction improving the win rates of certain faction.

Absolutely not. The onus isnt on me to prove that the results haven't been majorly tainted by the potential introduction of results that have basically nothing to do with how weak or strong a faction is. If you want to

An unknown variable has been introduced. This invalidates the results. There is no way to tell what is causing an improvement in winrate when using this dragnet-data gathering approach, and that is a black mark against it, not a point in its' favour.

The "All" statistics are absolutely worthless because we do not know the quality of the players involved. The mere fact that we don't know the quality of the data means it can't be used.

I am completely amazed by how dedicated vipper is to holding proven wrong opinion and how relentlessly he will barrage anyone pointing out how wrong he is, completely ignoring any logic and reason.

You're a clown if you think including all players is better way then only taking into account top 200, completely ignoring the fact that sample size for that top 200 is more then enough and still crying "but iz lezz den ALL!!"

Utter clown.

The "top 200" data certainly isnt perfect, but it's a damn sight better than using practically every game scraped off of the bottom of someone's shoe. Hopefully the guy gathering the data does implement something like a "top 10% analysis.

....

To retort: You haven't provided a single argument as to why using the "all" data would somehow be a more accurate view of balance, despite including games where one side may very well be made up of people who simply do not know how to play the game.

....

And I have not claimed that one should should be used over the other. Unless one can provide a convincing explanation for the difference none of them are safe.

But there are explanation why UKF could do look worse for "200":
It could possible that due to lower number of player a UKF player ranked 200 is on average weaker than Ostheer player ranked 200.

In addition the fact that TOP player avoided playing OKW (25% pick rate) is also an indication that OKW are not as good as the "200" game present them in 1vs1 and are closer to the number of all players.

And I have not claimed that one should should be used over the other. Unless one can provide a convincing explanation for the difference none of them are safe.

Neither of them are perfect, but the "all" data is clearly far more flawed than the "top 200" data merely by virtue of the fact that we do not know what is in the "all" data.

If all you're allegedly trying to argue is that both of these datasets are not enough to solely base balance decisions off of, then you're wasting your time (And going about it in a terrible fashion). I already agree with that, and I think most people recognise that this data is somewhat flawed. It is still the best dataset available at this time.

But there are explanation why UKF could do look worse for "200":
It could possible that due to lower number of player a UKF player ranked 200 is on average weaker than Ostheer player ranked 200.

Sure, which is why I stated that this data isnt perfect. Neither of these datasets solve the issue you're suggesting, however.

In addition the fact that TOP player avoided playing OKW (25% pick rate) is also an indication that OKW are not as good as the "200" game present them in 1vs1 and are closer to the number of all players.

A single tournament doesn't provide enough datapoints for more than a broad analysis, but I have previously stated that OKW would benefit from changes, so I'm not sure what you're trying to convince me of.

USF also had a poor pickrate relative to SOV, and UKF's pickrate was practically negligible.

UKF's performance is the one that stands out most in ML5 as being ludicrously poor, and USF's as one that was surprisingly good, incidentally. Despite the pickrate of OKW being low, their winrate in the top 8 was very good, and their upset performance was only slightly below SOV and OST, whereas USF's was comparatively through the roof. UKF's in comparison, was entirely nonexistant (Presumably due to the lack of picks)

The main thing I get from the ML5 data is that USF may be overtuned, OKW slightly undertuned, and UKF being a joke. OST has a very high winrate in the top 8, but their "upset" value is the same as SOV, so I'm not entirely sure what that implies.

I'd love to hear an actual statistician's opinion, because perhaps I'm assessing this incorrectly.

Neither of them are perfect, but the "all" data is clearly far more flawed than the "top 200" data merely by virtue of the fact that we do not know what is in the "all" data.

That is an assumption on you part and not a fact, which is not supported by any argument.

Again that is you opinion and not a fact.

It is a fact. The "all" data has unknown variables, and includes too many confounding factors to be genuinely useful. I really don't know why you're so adamantly against this other than, again, because the "all" data supports what you're arguing more closely.

The only thing it provides is a raw increase in datapoints, and if that's all you want you may as well take the total game's population, which would be similarly too broad.

I've made plenty of arguments, you merely do not like them.

It is a fact.

Simply is not.

Hold on now; if this is the quote:

Difference between top 200 and regular stats?

The regular analysis takes matches from top 200 players as base input too. However the difference is that regular analysis does not care about the other players.

So here it's saying that "regular analysis" includes matches from top 200 players and that "regular analysis" does not care about the other players. This of course would imply that "top 200 analysis" does care about the other players.

In regular analysis you could have one team be top 200 and other team rank 2000. Or in 4v4, only 1 player from the match could be top 200 and other players might be much lower skill.

This seems to be the issue that everyone says top 200 has. And yet the author of the top200 stats says that that's the issue from which regular analysis and top200 differ from.

Is everyone reading not reading into the original top200 post or do you guys know something that I don't?

Hold on now; if this is the quote:


So here it's saying that "regular analysis" includes matches from top 200 players and that "regular analysis" does not care about the other players. This of course would imply that "top 200 analysis" does care about the other players.



This seems to be the issue that everyone says top 200 has. And yet the author of the top200 stats says that that's the issue from which regular analysis and top200 differ from.

Is everyone reading not reading into the original top 200 post or do you guys know something that I don't?

I think you perhaps have misread some posts, unless I've misspoken in a previous post. "Regular analysis" is indeed the one I'm arguing has the flaw of including players who don't potentially don't understand that a Kubelwagen can't actually fight an IS2, or who don't understand that you're supposed to retreat your squads rather than let them get wiped out by a machinegun.

"Top 200" has a different problem in that the "Top 200" players in each faction(per mode) are not necessarily equally skilled. A faction like UKF has far fewer players than a faction like OST, and so the "best" 200 UKF players are not necessarily as skilled as the "best" 200 players from Ost, merely as there are fewer candidates to fill these positions, so to speak.

Perhaps this is where the confusion is/was.

(finally got around to answering)

OST has medkits on all their mainline infantry that can be used on any squad but that doesn't change that it can also have one of the most inexpensive healing bunker. OKW light vehicles aren't any better than T-70, Stuart or M20.
Sturmpioneers medkits used to be unlocked with veterancy, and passive healing for volksgrenadiers used to be unlocked way before than vet 5 so whatever the original design was it doesn't really matter. USF was designed to not have a tier 0 mortar

OST medkits function very differently though. No AoE healing, casting time and you can "only" use it on your main line. Medkit crates can be picked up by anyone and most importantly provide AoE healing, practically cutting the price at least in half. However you are fixed in place, but that is the case with any "medic squad" healing as well, which they are meant to replace.

It really doesn't

Claiming that it worked because Mechanized used to be meta ignores the fact that LVs are ALWAYS the meta for all factions, that the Battlegroup started seeing use because the flak ht was improved after being useless for a lot, as well as several key changes that OKW went through
Mechanized builds were nerfed, and so were Volksgrenadiers, which made Mechanized builds much less powerful. Ultimately there's no "meta" for OKW because the faction is competitively dead with 3 picks and 0% winrate in the last ML5, volksgreandiers late game bleed and inefficient trading being great part of it, which is only made worse by the lack of healing without backtech

OKW is strong in all modes (outside of tournaments), and even in 1v1 does not lack much behind Ostheer if at all. The lack of tournament play is something to be concerned about for sure, but given that larger stat samples show otherwise this does not seem to be a problem outside of the top 20 players.
Regarding LVs: OKW functioned well on T2 builds, even without a heavy to crutch on. Players still seemed to choose T2 builds despite the heavy mun bleed for a long time. Now their T1 got stronger and more accessible. Now there indeed is some powercreep that made other factions easier to play as well. Maybe my view on this is scewed, but I see both T1 and T2 builds regularly in the game.

The only faction in game that has a fully linear teching is Ostheer, because the mod team went out of their way to force a fully linear teching and disincentivize Tier 0 into Tier 2 builds, and still, OST can ignore the Tier 4 if not accessible. UKF has several side techs and Anvil + Hammer. No faction would be similar. Even OKW itself wouldn't change since its bad design basically always accounted for the player fully teching anyway, which works in resource inflated teamgames but gets absolutely stomped in 1vs1

It would still be a ridicolous munition sink. The only good med crate is the 221 because it can be dropped as single one to heal whenever you want field presence, which is a good tradeoff and a useful ability. Healing an entire match off sturmpios med crate *3 in your base costs simply too much and is far from what Relic probably originally designed med crates dropped by infantry for.

The aim of my suggestion is to slightly reduce the mun bleed of OKW from the mid game onwards. This should keep the intended trade off between STGs and healing intact while allowing more munitions float later on. As you said, OKW is supposed to get all trucks out at some point. If the resource pressure is too high in 1v1 but not in team games, SP veterancy might actually be the way to go. SPs shine in 1v1 situations and can level up quicker, while they are worse in team games. Additionally, wipes are also easier in team games due to more arty. This change should therefore benefit 1v1s more than team games and would ease the pressure of getting T1 out after teching T2 and T3.