If a player averages at a 70% WR in total, then a vehicle he has 60% in another, then the 60% WR is shown to underperform, since the player using it achieves a below average result relative to his skill level.
Similarly if a player averages at 20% WR, a vehicle he has a 40% WR in is overperforming relative to his skill level.
Filtering pure stats through individual player averages is how you can make sure that player skill is not what drives BRs, but rather just pure vehicle performance.
Sorry for the confusion, but in my first post about the possible solution i mentioned that it should be per vehicle basis.
For example:
Of ALL the players using the KPz-70, you take X % of top performers in it. And use their statistics in it, to determine the KPz-70s overall performance (over- or underperforming).
You DO NOT apply their performance to evaluate the overall performance of other vehicles. Unless they are in the X % of top performers in those vehicles.
Balancing single vehicles just by WR is a very bad idea, which is why Gaijin isn’t doing it.
I’ve played some very, very mid vehicles and had really good WRs in them, because they are in a pretty good lineup and not because they’re so strong.
Guys, there would not be such a big balance problem if Gaijin actually played the game themselves instead of just looking at excel sheets
Gaijin balances primarily based on vehicle earnings, which plays a huge part in the efficiency metric. Winrate is also a bad way to balance, especially in ground RB where there are many factors that affect it.
The top 1000 perform badly different in different types of tanks compared to the average player. They excel in mobile an fast tanks, while a worse player will do better with more armour. I also don’t think it will give a good idea on how balanced something actually is if you only take into account a fraction of a percent of the player base. There’s also issues with what even a top 1000 player is defined as.
Another thing I’d like to add is that the top 1000 players in certain sleal clubbing vehicles have a near 100% (or 100%) winrate and a huge KD. The players who sweat in squads are not the ones you want to balance the game.
This is such an old argument, and frankly people dont seem to realize the scale of the game. The amount of vehicles in the game, and the absolutely monumental amount of time the devs should have to dedicate to each vehicle. On top of all the other work they have to do.
I think the guy meant top 1000 of exact vehicle not just top 1000 of the whole game
Gaijin has always bent over backwards to equalize the lowest denominator, common or not. Most base bombers in air RB and bot scripters in sim for example are just module grinding for ground RB. Ever since ground got added air has been a mess in terms of mechanics and balance.
It may be old but time and time again Gaijin’s actions show that they in fact don’t play their game very often or at all and just look at excel sheets of K/D, WR and SL/RP gain of the vehicle
I’ll just quote myself from a couple of posts back…
What would you suggest? That they constantly play every vehicle of the game for hundreds of battles in all game modes themselves, to determine if all vehicles across the board are balanced?
You really don’t see how that is an absolutely ridiculous proposal?
Even then, my statement applies. The top 1000 M24 players would all have insane KDs and 95%+ winrates. The same applies to a lot of low tier stuff, especially in arcade.
And that exactly shows that in this example, in capable hands the M24 wreaks havoc and makes it overpowered, and it should be balanced. ie. moved up in BR.
Yeah fair, and probably easier for Gaijin to implement than my idea. Though this also lowers the sample size per vehicle, and might still be affected by the individual skill level of the top % players.
Still, it’s better than what we have now
I know, it just makes for a much easier example with simple single % values rather than a combined sum of multiple values, over certain periods of time that Gaijin have already vaguely confirmed they use.
Of course winrates generally suck for vehicle balancing, being the one main stat that doesn’t actually depend on the individual vehicle, but also the team or lineups, but it’s also irrelevant for my idea what value you use. Basically it goes something like this:
Example
With the current system, desired value (D) must be equal to or within a certain margin of tolerance to 50%. For that the players vehicle value (V) is taken and divided by the desired value (D). So if a player has 70% value on a vehicle the calculation simply goes:
V / D → 70 / 50 → 1.4
The resulting value of 1.4 shows the vehicle overperforming by 40%. For this is is absolutely irrelevant that the specific player here has an average value (A) of 80%, meaning in this specific vehicle he is underperforming, even if statistics show the exact opposite. However to filter out player skill is quite simple, by using individual player average (A) in place of a global desired average (D):
V / A → 70 / 80 → 0.875
This then clearly shows the vehicle is underperforming by 12.5% relative to the player skill, and allows the vehicle to be objectively balanced. This can then be applied to players of any skill level equally, and will still give vehicles more objective rating.
The only issue I can think of is that it doesn’t account for differing skill floors and ceilings on vehicles, but that seems like the lesser evil compared to “If played by good players → moved up”, at least in my own opinion.
Now, for a realistic implementation, all of these values would be split up into different subcategories or earnings, kills, assists and/or whatever other values Gaijin use, and also generally being taken only in from a certain recent timeframe, perhaps the last month or two, to reflect only the current game, but none of those things are relevant to the example at all.
Ridiculous proposal to make the devs play their own game? No one is saying to play hundreds of battles on a single vehicle (even though that’s what play testers should do), but you don’t need massive sample size to actually understand the game especially when you are the dev and you don’t have to grind anything. FYI Gaijin is bigger than 10 people and they do a lot of outsourcing often so they have or would have enough man power if they wanted to.
Imagine if any other company didn’t even launch their game before and after shipping it
Screw it, i think Gaijin should return to the form and balance vehicles solely on the basis of their SL/RP gain
Actually, win rates in Arcade (where all vehicles can face all vehicles, so there is no possible nation imbalance) is the perfect example, why including win rates in BR calculations is a completely stupid idea.
My stats in Ground Arcade, using exactly the same M44:
According to win rate logic, German M44 should go up from 4.0 to 4.7 (it’s massively “overperforming” with 77% win rate), French should go up to 4.3, American and Japanese should stay at 4.0, British should go down to 3.7 and Italian to 3.3 (it’s clearly “underperforming” with only 40% win rate).
I use one vehicle lineups, so I only used M44 in these battles (with eventual backup, when I die). These win rates are not affected by any other vehicle I used, these are pure M44 performance stats.
It’s actually pretty funny that some people read things like who is a good player, or which vehicle is better by looking at win rates. Win rates are almost completely random and luck based in 16v16 players battles. That’s exactly why the devs don’t take them into consideration.
Obviously devs dont have to grind vehicles. And im well aware that their company size exceeds 10 employees. But to have a meaningful sample size they’d have to play more than a couple of games, per vehicle, per gamemode, constantly. Because after balance changes, they need to re-evaluate EVERY vehicle affected by BR changes (vehicles now facing the newly moved vehicles etc), not just the ones that have had their BRs changed.
Personally i don’t consider your example to have a sufficient sample size per M44 to have a conclusive outcome.
But that’s beside the point, like i stated later on, i used WR % as a broad example (which has been largely ignored, oh well, my fault for cutting corners in the initial post), and later reiterated that more statistics should be take into consideration for a more precise evaluation of performance.
Of course vehicle duplicates in different nations should have the same BR anyways.
You are extremely overestimating sample size and time needed, and it’s not like we are getting BR changes every week or two, it takes Gaijin several months (sometimes much much longer) to just look and excel sheets and change BRs of obviously broken vehicles, which would take much less time if someone at Gaijin actually launched their own game once in a blue moon