115 replies to this topic

[Roland]

Introduction

I've all but given up playing MWO, being disheartened that PGI has decided to renege on their promise to allow groups between 5-11 players to play again. However, as some of this decision hinges on perceived complexity with matchmaking in such cases, I have decided to present a solution to a large number of the problems that PGI is currently encountering with matchmaking and balancing within the game.

This suggestion hinges on an idea I presented long ago, where mechs are balanced based on a dynamic Battle Value system. However, in order to address the fact that Battletech's BV system is designed to balance a boardgame, I present a system which is instead derived from actual mech usage statistics. To put it most simply, mechs which are used most by the population of the game end up being the most expensive from a matchmaking perspective.

Now, when I originally presented this, I did so fairly vaguely. What I've done here this time is present what is effectively a complete implementation of the system. This system could be quite easily coded in any language, hooked up to the database storing mech usage stats, and incorporated into the game.

Summary
PGI has struggled with a way to establish a mechanism for balancing the equipment that two teams bring to the field. Without restrictions, the variety of mechs dropped tends to coalesce around a fairly small subset of all available variants in the game. In a game with 120 different variants, a large chunk of those mechs are considered "non-competitive" by much of the game's population. In many cases, it's not because these mechs are flat out bad, but rather because they are simply slightly less efficient on the battlefield than other mechs. Thus, piloting them is still an enjoyable experience, but some incentive must be given to encourage pilots to voluntarily take "sub-optimal" mechs.

At one point, the idea of tonnage balancing was investigated. However, this balancing scheme falls apart, because it can easily be sidestepped by dropping lances of top-tier assault mechs grouped with top-tier light mechs, which is already a highly efficient drop configuration. The issue that is exposed by attempts at tonnage matching is that mech utility is not directly correlated to tonnage. Certain mechs, like the awesome, are effectively inferior to other mechs of the same tonnage, like the Victor.

In Battletech, the notion of Battle Value, or BV, was developed as a means by which to address such issues (especially once clan tech arrived, and clan mechs had clear advantages for a given tonnage). However, BV itself can have certain critical issues.

Primarily, the weakness of BV stems from its reliance on pre-specified values that attempt to estimate the combat utility of a mech or piece of equipment. The failure of such a system derives from the fact that in a complex combat environment consisting of thousands of individual users, it's difficult for a human to effectively summarize something as ethereal as "combat utility". Pre-defining BV for mechs and equipment ends up having the same problems as price-fixing by a governmental organization in a market-based economy. The market, or player-base in this application, will naturally identify the mechs which are "incorrectly" priced below their true utility value, and gravitate towards them. The emergent intelligence of the "hive-mind" consisting of the whole player population will always be better at identifying the true value of any mech. It will always work against any directed balancing attempts by developers, to identify the most unbalanced aspects of the game and exploit them.

Thus, the solution is not to fight that emergent intelligence. Instead, the solution is to leverage the player-base's intelligence to automatically perform balancing for us. That is what this solution attempts to do.

The beauty of this system is that it distills mech efficiency and utility to its most easily evaluated statistic... whether or not people actually use a mech. All of the complexities which ultimately comprise a mech end up being automatically accounted for, without the need of us directly measuring (or even identifying) them. What's more, by looking at usage, the system will automatically adapt itself to changes in the playerbase's actions. As new metagame features emerge, resulting in new mech designs and drop configurations, the system will automatically incorporate these changes into its evaluations of BV, and the system will automatically balance itself.

Here we effectively create a market-based system, where a mech's BV is derived directly from its usage stats (which are already being tracked by PGI). We calculate an average number of drops which constitute a "basline usage rate", which is the number of times we would expect to see a mech to be used in game if all mechs were exactly equal. (With the current 119 variants, the baseline usage rate would constitute the total number of mechs dropped over a time period, say the past month, divided by 119) Then, mechs which are dropped more than this number have their BV increased, while mechs dropped less than this number have their BV decreased.

Thus, BV becomes a representation of how useful the overall game's population feels a particular mech variant is.

Once this BV is established, we then apply it to team composition, in such a way as to encourage usage of those "less useful" variants. For solo pilots, we impose no restrictions, as they have no way to balance their usage. For grouped teams, we impose increasingly strict restrictions based on the number of team members. A very small team of only 2 players would suffer almost no restrictions at all. However a full team of 12 players would need to balance high-utility mechs with low-utility mechs.

What this means is that this system not only provides a mechanism by which to balance the mechs used by each team in a way that cannot be achieved through simple tonnage balancing, but it also provides a means to balance the intrinsic advantages that pre-organized teams have over random solo puggers. This system will, effectively, allow less organized players to drive better mechs.

So, in summary, this thread presents a system which will achieve the following goals:
1) Increase the variety of mechs used in game
2) Balance the mechs used by each team
3) Provide a balancing mechanism that accounts for pre-organized groups of players, allowing them to be matched against groups of different sizes as well as solo pilots, while minimizing the inherent advantage that comes with player organization
4) Create a system which will dynamically adjust to changes in player behavior over time.

What this does NOT attempt to do:
1) This is not a matchmaking system. Instead, this merely provides a value for individual mech variants, which could then be incorporated INTO a matchmaking system, as one component of the matchmaking process.

Below, I will present a more detailed explanation of how such a system would actually be implemented. I will also attempt to present a justification for the belief that this system will indeed be able to provide an effective balancing value while only analyzing the drop rates of mech variants, while not needing to examine more detailed play statistics.

MORE TO COME AS I GET AROUND TO IT

Edited by Roland, 04 March 2014 - 02:31 PM.

[Roland]

BattleValue - What it is and what it is not

Put simply, BattleValue is simply a single numeric measurement of a mech's "utility". I put that word in quotes, because each person's perception of what it means to be a useful mech differs. But therein lies the strength of a market-based, or usage-driven system. No single person needs to define utility for everyone. Each of us gets to decide on our own how useful a given mech variant is.

Here, BattleValue can be regarded as a "price" of sorts. That is, ultimately, after we explain how it is derived by the system, we will go into a more detaield explanation of how it can be used within a larger system to limit mech usage at play time. Here, however, I will merely explain briefly that for groups of players dropping together, the group as a whole can be allocated a maximum BV for their whole team. Thus, they are essentially being given a pool of BV, which can be throught of as "money" which can then be "spent" on mechs that they want to drop.

In this way, BV becomes the price of a given mech variant. Choosing to put a pilot into that mech in your team effectively constitutes "purchasing" that mech with your team's allocated BV.

The way in which this system works, is that it tracks all of those "purchases". The more a mech is used, the higher its "price" (it's BV) goes up in the market. Likewise, mechs that no one uses will get cheaper in terms of their BV. Eventually, the really good mechs will become so expensive in terms of BV, that it will actually be a meaningful decision to use them... it won't just be a no-brainer. Likewise, eventually, very under-used mechs will eventually be so cheap that they will effectively be worth using (generally, because using them will conserve BV, which will then in turn enable another player in your team to use one of the really expensive mechs).

An important note here is that BV, while a term that originated in TableTop Battletech, is NOT the same thing as that table-top system. The numbers used in the TT game were meant to balance a totally different game, and thus will not apply here. Their "prices" would be wrong. If you attempted to put them into this game, you would run into the same problem that any market-based system runs into when prices are fixed at an incorrect point... While it may have an impact in that mech usage would shift from its current usage to a different one, it would merely settle into another stagnant usage pattern. Instead of players simply focusing their usage on the highest utility mechs, as they do now, players would instead just focus on the mechs whose BV was most below what the ACTUAL utility in game was.

Even in a case where you manually defined new BV's for mechs, you would still likely have an unbalanced system, because no single person is really equipped to evaluate the "true" utility of mechs. Just like no person in the real world is capable of setting prices in any complex real-world economy. The system is so complex, and involves so many individual actors with various opinions of worth, that it's impossible for any human (or even small group of humans) to oversee such a balancing act. They will always make evaluations which presuppose certain ideas about how the game is going to be played, and then the player base will strive to break that system. And it will always succeed, and find the most powerful, least balanced configuration available. Because that playerbase is a huge hive-mind consisting of every single player.

What we're doing here, with this notion of BattleValue, is that we are giving up on the goal of trying to beat the players when it comes to figuring out what the strongest configuration is. We're not going to try an defeat the hive-mind.

We are going to USE that hive-mind, by using its own actions to determine BattleValue.

So here, BattleValue is going to be the ACTUAL value of a given mech variant. Not determined by me, or you, or PGI, or any single person or group of people. It will be the actual value, as perceived by the overall community of Mechwarrior Online. And this value will be based upon the one action that matters in this regard; whether or not people use that mech.

In that single statistic lies everything you need to know about the utility of a mech. All of the complex details about what it can carry, or what it is carrying, or whatever... That all gets boiled away. Because ultimately, it's all represented in whether that variant gets used. Variants get picked BECAUSE of what they carry. BECAUSE of what they can do. You don't need to analyze any of that, because it all gets handled intrinsically behind the scenes when you measure the simple usage stats.

And what is great about this, is that as the player-base attempts to work around the current BV's at any given point in time, the system will automatically adapt to those changes. As certain mechs get more expensive, others may come into favor. But as those get used more, then they too will get more expensive. The originally used ones may even get a bit cheaper. Eventually, all variants will find themselves in an equilibrium.

As new mechs come into the fray, or new weapons show up, or weapons themselves are balanced? All of those things will result in changes in the usage patterns of the players. And thus, all of those changes will be automatically accounted for. The system will automatically adjust itself to prevent imbalance. It will continually exert a force on players to encourage usage of different mechs in the game.

Detailed Implementation

In order to implement such a system, we will define a few variables which we will then use throughout the process.

BV = BattleValue. A number representing the total perceived utility of a mech.

BV_adj = The adjusted battlevalue after taking into account the usage of a given chassis.

BV_base = The original value used to represent a mech's value, prior to any inclusion of drop stats. Essentially, this just provides a representative value which will guide the resulting values after accounting for usage. I used a value of 1000. It really could be anything, and doesn't matter. If you use really huge values, then all BV's will be really huge. If you use a small value, then all BV's will be tiny. I don't think I'd go smaller than 1000, because then you may need to actually care about the decimal places and not just be able to round to the nearest integer.

M = the set of all mech variants in the game. Note, this includes each VARIANT of a chassis, as they can effectively be regarded as separate mechs for all intents and purposes.
M_f = the number of drops (or frequency) for one particular mech, over a period of time.
Tw = The time window over which mech drops will be analyzed for the purpose of calculating BV. This time period could be any length, although I would probably start with a month.

Tc = The time between running the calculation to adjust prices. This will be a much shorter period of time. perhaps as short as every hour or so.

BDA = Baseline Drop Average. This is the number of times a mech would be dropped in game, if all mechs were dropped in game at an exactly equal rate. That is, if every variant in game was used exactly the same amount as every other.


Of, so there we've got a few basic ideas. First, let's work at establishing the BDA. This calculation is pretty simple, in that it's just the total number of mechs dropped during the period of analysis, or Tw, divded by the number of variants in game. So...

BDA = (Total # of Individual Player Drops during the Tw window) / (Set of variants in M)

As an example, let's imagine we have a 30 day month where every day a thousand games are played every day. Each game has 24 people (and frankly, you can even include failed drops here, since basically you only need to account for the number of times a player ATTEMPTED to drop in a given chassis).
24 players per game * 1000 games * 30 days = 720k drops
BDA = 720,000/119 (119 variants in game currently, not including the new Banshee) = ~6050

So, if every mech was perceived as perfectly balanced by the player community, and dropped exactly the same number of times as every other variant, then every mech would have been dropped approximately 6050 times. This number is of course notional, and the value itself doesn't really matter. Although the system wil work better with larger numbers of drops, it shouldn't really suffer any problem unless the number of drops over period T_w is so small that tons of mechs never get dropped, and the ones that do only drop a few times. Any reasonable size T_w (say, a month) would prevent this though.

Once we calculate that, we calculate the standard deviation for the population of drop frequencies. (Basically any statistics program or spreadsheet will do this calculation for you. Plug in the set of drops for each mech.) We'll call this Sigma.

Ok, so now we want to calculate how many standard deviations each mech's drop frequency M_f during period of time Tw represents. We'll call this M_sig.

M_sig = M_f / Sigma

Now we take that number, and need to use it to generate a new Battlevalue (note the ' though, as this isn't quite the final value we'll use).

BV_adj' = BV_base + ((Abs(M_sig) * M_sig) * BV_base)

All we are doing here is effectively squaring the mech's number of standard deviations (using an absolute value to preserve the sign of mechs with a negative sigma value), and then multiplying that by the base battlevalue. This will give us a modifier, either positive or negative, that relates to the number of standard deviations the mech's drop frequency has strayed from the average.
Squaring M_sig isn't strictly necessary here. I did it so that it would dampen the difference of mechs less than one standard deviation from the average, while accelerating the difference that are further out. The effect of this is that mechs which have very high or very low usage compared to the average, will end up having significantly higher or lower prices, relatively. Thus, it creates an additional disincentive, or incentive to use them respectively.

Now we have a value, but a bunch of them will be negative. This isn't strictly a problem, but will cause some weirdness, so we re-level everything to bring it all back into positive territory.
Let MinBV' = the minimum value in the set of BV_adj' we just calculated.
BV_adj = BV_adj' - MinBV'

At this point, you have a BV for that mech, which represents its utility relative to other mechs, for the period of T_w , where the lowest BV is at zero.

As a test case, I generated a bunch of drop frequencies for all of the mechs, and then ran this process on them. The following graph shows the result of this process. You can see how the drop frequency for each variant (The Red Line) results in relative battle values for those chassis (The Blue Bars).

For reference, I've also included a line showing the average drop frequency (The Orange Line), to provide a reference for which mechs were dropping higher or lower than average. The Green Line shows the BV which will be allocated to each pilot during team composition (Explained in detail below, so don't worry about this too much right now). The only reason I include the green line here is so that you can see which mechs are expensive, vs. which ones are "bargains".



I left the specific variant names off of the chart, since they don't actually matter here.. since the drop data was made up. The actual results aren't totally terrible, as I seeded the data using Kaos Mavrck's mech tier list, so it kind of seems reasonable, but ultimately the specific values of the graph would only be meaningful if they corresponded to actual usage statistics. Of course, if PGI wants to provide me with some subset of data, I'd love to use it to test my system. There are likely tweaks which I'd make once I observed a real run.

Ok, now that you have this system for calculating BV, you need to run it. That's where Tc comes into play. Every period of time, Tc, you run the above calculation over the window Tw. Appropriate values could be something like Tc = 30minutes, and Tw could be a month.

This will create a sliding window, where prices will tend to adjust fairly gradually. That is, you won't have some sudden activity that slams the price of a chassis up or down real suddenly.

To follow, I'll present how this can be used to actually limit mech usage in the actual game.

Edited by Roland, 05 March 2014 - 03:57 PM.

[Roland]

Ok, since folks are asking questions (which is good), I'll try and finish something up here just so folks have some idea of how this can be applied. This is going to need some revising, but at least folks won't think this is meant to be used for Elo rating or something.

Also, bear in mind, it's totally POSSIBLE to use the aforementioned BV calculation as part of a matchmaking algorithm. Really, you could use it for all kinds of things. It's just a way of measuring the relative strength of different variants.

However, here, I'm going to present one use... largely from the perspective of a player who participated in competitive league play in MW4, where our matches were limited by tonnage.

Application of BV in Game for Team Composition


Now that we have a BV for individual mechs, what can we do with it?

The primary use I have in mind for it, is to address the problem that PGI had when they investigated implementation of tonnage limits.

The problem with tonnage limits for a team, is that tonnage isn't really correlated with performance. Certainly, there are a lot of really great mechs that are ALSO very heavy, and the combination tends to be overpowering.. but the tonnage alone doesn't equate to quality. Many light mechs are clearly superior performers compared to some mechs more than twice their weight.

The BV I've presented here, however, deals with that problem. If you limit a team's overall BV, you are actually doing something meaningful. Unlike a tonnage limit, they can't simply sidestep it by choosing to use a combination of heavy and light mechs which are all very good.

For instance, say you said, "Ok, for 2 mechs I'm going to give a tonnage limit of 60 tons per mech." Seems reasonable.. It's at the bottom of the heavy chassis weight class. Not super heavy, but with flexibility to drop what you want.

But in that weight, you could bring a victor and a Raven 3L, two very strong chassis. You really haven't limited much at all.

However, with BV, you CAN'T do that. Because both the victor AND the Raven, despite being totally different mechs that do totally different things, are both extremely common and widely used, and thus both have high battle values. In order to use a popular mech like the Victor, your partner will need to bring a mech that is much more rarely used.. although not necessarily light. He could bring any "bad" mech.. maybe a locust, but maybe something like an Awesome.. Or a quickdraw.. or even a raven, but not the 3L...or any of the 119 variants that we very rarely ever see on the field.

So here is how we do that.

Before we start, a few things.. There is one thing here which tends to go against the spirit of BV, but I think it's important, in that I wanted to preserve the ability for a solo player to drop in literally any mech he wants, no matter what. The result is that solo players are largely unaffected by BV, and as a result, could skew the drop frequencies.. Although I believe that the calculations I laid out above will stand up to this.

Ok, so the key thing here is that this system is mainly designed to facilitate formation of TEAMS of pilots. Meaning, prior to you hitting the launch button, your team has X BV to spend on all of its mechs. Thus, the reason to bring the "cheap" mechs is generally so that another team member can bring the "expensive" mechs.

Each team gets an amount of BV allocated per pilot. We call this the Per Pilot BV Allocation, or PPBV.
Now, the PPBV will fluctuate over time, every time the BV's are calculated for mechs. The PPBV is equal to the average BV of all of the available mechs.

For the spread of BV's shown in that graph I posted earlier, which was taken over a simulated 19,181,222 individual mech drops, the PPBV ended up being 2084.

This means that each pilot on a team is given that many points to spend on his mech. Now, in reality though, the points from all of the pilots in the team are put together into one big pool, and then checked against all of the selected mechs when the group lead hits launch. Remember, we're talking about a team you formed on your own here, before launching.. Not the random folks you are going to be matched up with if you have less than 12 people.



On that graph, which I've included here again for easy viewing, the green bar represents the PPBV. As you can see, due to the way BV is calculated, most mechs are going to be fairly close to the PPBV. It's variants that are used WAY more than others which tend to have those really high BV's, and mechs that no one uses have the really low BV's.

What will happen here, is that you may want to bring one of those super awesome mechs, like a HGN-733C or something. Well, suppose it's the most used mech in the game. That means it's gonna have a really high BV.

In order for your team to put you in that mech, someone ELSE is going to have to take a mech which is equally UNpopular. Not necessarily bad, of course.. but one that is not commonly fielded. Or, perhaps, two different players could both bring mechs which were moderately unpopular. Or any other combination, as long as the BV adds up and doesn't go over the team's cap.

What's important to keep in mind here, is that cheap does not necessarily mean bad.. Cheap means that a mech is merely not perceived as useful by the community. There could be numerous reasons for that. Perhaps it's great, but there's another mech variant which is just a TINY bit better. A lot of the stalkers, for instance, may fall into this category... they all have extremely similar loadout capabilities, but with a few minor differences (in the case of the 3F, a fairly significant torso twist boost). The point here though is that being "forced" to bring a lower BV mech isn't the same as being forced to bring a bad mech. But it means that you will have to perhaps be more creative.. You won't be able to just bring the same mech everyone always brings.

But the upside is that neither with your opponents.

Additionally, if you are creative and can find some new way of building an effective mech using a previously unused chassis, then you gain an advantage... because you will be extracting more utility out of that chassis than the community at large perceives it to be capable of. That "savings" of BV can then be used by your teammates to bring other more traditionally useful mechs.

Eventually, that build may "get out into the wild" and start to become popular.. and that will cause its price to go up... And eventually the price will get to where it's in equilibrium with its utility.. and then it starts over, and you go find some other new diamond in the rough.


Now, in terms of the team composition, the overall BV is a combination of each pilots PPBV. However, I've added a few extra tweaks to this, which I think achieves an interesting effect.

First of all, as I mentioned previously, the whole notion of BV here is all about being able to "trade" BV with your teammates. One guy takes a cheap mech in order to enable another teammate to bring an expensive one.

But for a solo pilot, there's no one to trade with. So we can't simply limit him to the PPBV, because then he'd never be able to drop in an above average mech. And that's garbage for him.

On his own, a solo pilot should be able to drop in ANY mech he wants, no matter what.

In order to facilitate this, we set the Solo player's PPBV, or PPBV_s to be equal to the highest BV of any mech available at the time.

For a full team of 12, the PPBV is just the straight up PPBV. Their limitations are most strict. They are going to be the most organized groups, and they are going to have the most flexibility in terms of being able to trade BV between pilots and adjust things, so they get the fewest points to play with. Although, they still have enough points for every pilot to drop in an "average" mech, and as you can see in the graph I provided, that's going to tend to be the majority of mechs in the game.

Then for groups of different sizes, we calculate what we call the Per Player Team Mod, or PPTM.
PPTM = (PPBV_s- PPBV) / 12

(At this point, I'd just like to point out that this forum sofware really hurts my brain, as every time I use the subscript, the next time I hit backspace it goes back to the subscript, and deletes a bunch of stuff, and enrages me. Apparently this is a known bug with this forum software, and may be fixed in a current version. Please fix it. Please. Anyway.. Ahem.)

That is, the PPTM is simply the Solo players PPBV, minus the base PPBV that we're applying to pilots on a 12 man team, divided by 12.

Then, we calculate the PPBV for pilots of team sizes of 2-11, by taking the solo player's PPBV, and subtracting the PPTM from it for every player on the team.

These values are from the BV's generated in the data above. Now, every time the BV's are recalculated, these numbers will fluctuate, and are automatically regenerated. But these specific numbers just happen to fit the set of prices that were generated by that particular dataset.

PPBV_S = 7289
PPBV₂ = 6422
PPBV₃ = 5988
PPBV₄ = 5554
PPBV₅ = 5120
PPBV₆ = 4686
PPBV₇ = 4253
PPBV₈ = 3819
PPBV₉ = 3385
PPBV₁₀ = 2951
PPBV₁₁ = 2517
PPBV₁₂ = 2084

As we can see, there are some interesting implications of this system.

First, as I mentioned, a solo player can drop in whatever mech he wants, always. This could potentially have a slightly inflationary effect on the prices, although since they are being calculated based on a base value and not as a delta of prior values, I do not believe this will actually break the system.

One thing that this system enables though, is a mechanism which balances teams based on their size.

That part is really important there, because it's some source of frustration for tons of players, both solo and group players alike.

In this system, while solo players can drive whatever they want, as players become more organized, they become more and more limited in what they can drive. Never so limited as to make it impossible to find a mech.. because there are so many variants in the game to choose from.

For small groups, the limits are very slight. With two players, you will be able to drive ALMOST anything in the entire game. The only combination that may be prevented, would be if both players wanted to drive the most widely used mech in the entire game, at the same time.

As group size increases, the limitations become gradually more apparent.

What this means, is that you could implement an unrestricted queue where teams of any size can participate, and there would actually be some counteracting force to help balance the larger teams. If they were going up against other large teams, then both teams will have the same restrictions imposed on their mech selection.. If they go up against small teams, or solo players, those guys will have had the option to bring "better" mechs. Now, as I mentioned previously, BV is merely the overall population's view of a mech's utility.. so an organized team could very likely identify mechs which the community at large hasn't really found yet, and use that to minimize the BV penalty they are suffering here. Indeed, organized teams tend to be quite good at doing that kind of metagame analysis (at least the good teams are).

Potential Applications for Clans and Mercenary Contracts

Above, I lay out some ways to apply this system to the current gameplay environment.

Additionally, at some point in the future with more complex Community Warfare elements, this BV system could facilitate a few different gameplay elements within that larger world.

For Clans, this BV system would enable a bidding system. Clans, when dropping against a target, could bid a specific BV for their team in order to secure rights for a drop. Having BV's which match actual in-game utility is required for such a system.

Clans, due to their inherent technological advantage, may need to be represented as a completely separate "market" of mechs. That is, clan mechs' usage statistics would only be compared against other clan mechs' usage statistics, while IS mechs are only compared against IS mechs. This could enable you to set up two separate price scales for each. If you apply a higher base value for the clan mechs, you would effectively result in a situation where their BV was inherently higher for any given usage frequency.

Likewise, for mercenary units, a BV system like this could potentially be worked into a bidding system used for contracts. While not fitting into the lore quite as perfectly as it does with the clan batchals, it would still provide a functional basis for a system where different units are able to "do more with less" to potentially prove themselves over other units. This could potentially be reflected in a lower overhead rate, and thus higher profits for mercenaries who use cheaper mechs. This, in turn, could potentially enable them to take lower paying contracts while still making the same profit, making them more competitive to potential employers.

Edited by Roland, 07 March 2014 - 10:58 AM.

[Domenoth]

Roland, on 04 March 2014 - 02:31 PM, said:

This system could be quite easily coded in any language, hooked up to the database storing mech usage stats, and incorporated into the game.

I believe the validity of this statement is nearly impossible to determine based solely on the knowledge available to us (the players). You may be right, and you may be dead wrong. I believe you have done your best to describe your idea in a way that would make it easy to implement but without intimate knowledge of interconnections of the system, your guess is as good as mine.

Now for feedback.

I am concerned that you're premise rests solely on usage. If a group the size of the goons decided to all play the locust 1V for x months straight just to throw off your numbers, what would stop them? Throw in the fact that that's exactly the kind of thing they think is funny and my concern level rises considerably.

Even throwing in win/loss tied to usage could be gamed.

Now we have to build some form of "intelligent usage statistics" (weed out the trolls if you will) and now magically I fear we are well out of the easy-to-implement-realm.

I'm not meaning this to sound like I think you are wrong. All scenarios I've tried to come up with on my own time have been similar to yours (dynamic BV of some sort) but I've yet to come up with something simple and unlikely to be affected by malicious users.

Good luck with your efforts. As sad as it makes me, and you too I gather, I think something as basic as 3-3-3-3 might be the absolute best solution when factoring in how long it will take to implement and deploy (sure a better solution might be available given 2 years of dev time but I'd rather see something that improves matches in a month if it's available).

I respect your right to vote with your wallet, but if the game can be made fun-enough now and that allows it to survive for 2 more years, that opens up the chance for bigger and better things. My recommendation is try things out with an open mind. If games get better with 3-3-3-3, great! If not, constructive feedback why 3-3-3-3 isn't better and wallet-votes are likely the best means to steer PGI in the right direction. That's currently my plan of action.

[PawPaWuFF]

great post roland! i love everything about it, couldnt have said it better myself

[no one]

Roland, on 04 March 2014 - 02:31 PM, said:

The beauty of this system is that it distills mech efficiency and utility to its most easily evaluated statistic... whether or not people actually use a mech.

The problem with that is that such a system falls down in the face of external constraints on what 'Mechs are more prevalent or available for use. Hero 'Mechs, or any 'Mechs that are expensive to build or master are going to have artificially low values.

I believe you're on the right track, but you haven't properly touched on what to balance by. Personally I'd say that if each player has an Elo by which they are matched, then each 'Mech chassis could have an Elo modifier.

If the game kept track of your ability with each 'Mech, it could balance teams with the Elo of the pilot, adjusted by some factor that represents that pilot's individual effectiveness in the 'Mech they're driving. Eg. - You're a 2000 elo player, so you're matched with other players in the 2k elo range, but you rate a 2500 adjusted elo pilot when you're in an atlas and a 1000 adjusted elo player in a locust. So OpFor's team then needs to have a pilot with a similar level of adjusted skill. Some guy that's really bad with his dragon might rank with some other guy's skillful commando, but tonnage mismatches would be less prevalent.

Along with your personal skill rating, have each 'Mech start with an elo multiplier that reflects the average player's performance in that chassis. Eg the average pilot might be half as effective in a locust, so the game starts you with a .5 * (your base elo) for that chassis and then adjusts with your personal performance rating from there.

Edited by no one, 04 March 2014 - 07:09 PM.

[Roland]

Domenoth, on 04 March 2014 - 06:05 PM, said:

I believe the validity of this statement is nearly impossible to determine based solely on the knowledge available to us (the players). You may be right, and you may be dead wrong. I believe you have done your best to describe your idea in a way that would make it easy to implement but without intimate knowledge of interconnections of the system, your guess is as good as mine.

Well, I haven't finished describing the implementation yet. I'll finish it up soon, and then you'll be able to see exactly how to implement it.

It only depends upon usage stats, which we know they track. The math isn't that hard.

Quote

The problem with that is that such a system falls down in the face of external constraints on what 'Mechs are more prevalent or available for use. Hero 'Mechs, or any 'Mechs that are expensive to build or master are going to have artificially low values.

THIS is true, in that the usage of those hero mechs is indeed lowered "artificially" by virtue of the fact they are harder to acquire. However, that being said, I don't believe that those few mechs would suddenly become unbalancing to the overall system

Honestly, the the Muromets is fairly good, but not by any means a dominant mech in the game... The Misery, again, is pretty strong, but still not nearly as dominant as other mechs have been in the past.

Quote

I believe you're on the right track, but you haven't properly touched on what to balance by. Personally I'd say that if each player has an Elo by which they are matched, then each 'Mech chassis could have an Elo modifier.

Again, it's not a matchmaking system. It's merely a metric by which to evaluate relative utility for different mechs, in a way that can then be used within some kind of matchmaking system or team restriction.

To put this in the context of what PGI has been doing... This doesn't deal with matchmaking, but rather this provides a better, more dynamic system for achieving what PGI is trying to achieve with their 3/3/3/3 limits.

What makes this system better, is that it'll actually provide incentive to use the under-utilized mechs within each weight class... so you'll actually have a reason to drop a locust, or an Awesome, for instance.

[FupDup]

While a dynamic BV would obviously be better than a static one like TT had, I've always been a fan of just balancing out the weapons/gear/etc instead of having external point systems. I.e., instead of having the LB 10-X have a super-duper ultra low BV due to it sucking so hard, we would buff it to be an actually effective weapon (i.e. increase damage per pellet).

[WildeKarde]

Could be interesting to use but BV needs to account for several factors to balance:

• Mech & Equipment
  
• Pilot Skill in mech / class of mechs for example.
  
• A variable based on build / weapon groups. A single AC20 is decent but twin AC20's should be higher than 2 x AC20 value for example.
  
• A value based on being in a group / depending on group size.

[Procyon Alpha]

I don't know if I missed it but the battle value on G-d like systems such as ECM needs to be increased like 100 fold because the TT version of ECM was kind of silly.

[Varent]

I like the idea (and have said so on the balance forums) But to properly impliment this you also need to track what weapons are being used on those mechs. Otherwise you will just be trading in one broken system for another.

Different mechs wielding different weapons have totally different uses and abilities. Raising the match maker value of a mech just based on the chasis type could be bad if they are using a totally non meta (just for fun build). For example. Also While many may be using a mech for jump sniping or long range bombardment, Some players may buy a mech for overall versatility and use it in a number of rolls where they like playing. Raising a players elo basically for them being frugal can be a bad idea.

[Roland]

Quote

I like the idea (and have said so on the balance forums) But to properly impliment this you also need to track what weapons are being used on those mechs. Otherwise you will just be trading in one broken system for another.

Really, you don't. The variant itself encapsulates its hardpoints, which effectively encapsulates the weapons it can bring.

You don't need to deal with weapons individually, because the usage of a given mech implies the usage of certain weapons and equipment.

For instance, ECM is really useful.... But you don't need to explicitly state that, because it'll show up in the usage stats. Mechs which can carry ECM get used more than comparable mechs which can't. That's why the DDC is used more than other Atlases, and why the 5D is used more than other Spiders.

The same basically goes for all other weapons. The weapons stats are effectively covered by virtue of the fact that mechs which can use good weapons are used more often then mechs which can't. This can actually be seen empirically at various points in MWO's development history.. For instance, as an extreme example, back when SRM's were crazy strong, what was the impact on mech selection? Mechs which had a ton of missile hardpoints were extremely common. Mechs like the A1 were seen in every match.... But then when SRM's were nerfed, what happened? The usage rates of various mechs changed dramatically. This is because mech usage is so heavily related to weapons, to the extent that you don't need to separately consider equipment.

Indeed, not specifically caring about the equipment actually has advantages, in that the system will automatically deal with more complex situations, such as combinations of pieces of equipment.

As an example of this, consider how utterly dominant the Raven 3L was compared to other mechs at earlier stages in the game's development. I can hypothesize that a big part of this was due to the COMBINATION of ECM mixed with Missile hardpoints.. It let it carry streaks and crush other mechs, while the hit detection issues at the time made it very tough. Of course, there may be other reasons I'm not considering.. And again, therein lies the beauty of the system. I don't need to know why the Raven was so strong... All I know is that it WAS strong, represented by the fact that you saw tons of them. That one statistic neatly wraps up all the other more nuanced aspects. It leverages the emergent intelligent of the entire playerbase to make a measurement of worth that is beyond the ability for any single person.

The only thing that not accounting for individual equipment BV results in, is that you won't really get any benefit to bringing a really high quality chassis loaded up with garbage weapons. So, if you bring a Highlander, but load it up with NARC's and machine guns, the BV is going to be identical to someone who actually runs a good highlander build.. But I'm really not interested in that. I'm willing to accept that.

Likewise, while this system will provide some balancing for weapon TYPES against each other (that is, missiles vs. ballistics vs. energy), it will have less impact on balancing weapons within a particular type. This is mainly because the hardpoint system we have now really only limits the number of weapons of a type that a mech can carry, and thus only this information is really contained within the mech's "identity" to the system.

This system could potentially account for ballistics being really strong, for instance, by increasing the price of mechs that have ballistic slots... If missiles were really weak, the CPLT-A1 would likely become cheaper. But it won't really account for things like the LBX-10 being really weak, because most mechs which can carry an LBX10 could easily swap it with another ballistic which wasn't weak.

The exception to that will be that certain mechs are uniquely capable of carrying certain pieces of equipment, and if those pieces of equipment are very powerful, then this could show up in the usage stats... For instance, if the AC20 was really awesome, then a mech with a ballistic HP in the arm, but also a lower arm actuator, may show up as cheaper than a mech with no lower arm actuator, since only the latter could actually carry that powerful weapon.

Of course, this isn't to say that a usage based dynamic system couldn't be constructed that takes into account individual equipment, but I'm not tackling that right now... it would add a huge amount of complexity, and I suspect it would give you very little additional precision compared to simply looking at the variant usage.

[no one]

Roland, on 05 March 2014 - 10:27 AM, said:

. . . it's not a matchmaking system. It's merely a metric by which to evaluate relative utility for different mechs, in a way that can then be used within some kind of matchmaking system or team restriction.

To put this in the context of what PGI has been doing... This doesn't deal with matchmaking, but rather this provides a better, more dynamic system for achieving what PGI is trying to achieve with their 3/3/3/3 limits.

What makes this system better, is that it'll actually provide incentive to use the under-utilized mechs within each weight class... so you'll actually have a reason to drop a locust, or an Awesome, for instance.

I understand that. The only thing I'm really saying is that usefulness is not always the major impetus for how 'under-utilized' a 'Mech is. That doesn't mean it's a useless statistic for balancing within weight classes, but I think there are better.

I'm saying the best indicator of how useful a given 'Mech is would be how effective that 'Mech is in combat. Averaged Elo or match scores would, I think, be good ways to track this and assign a meaningful value to it.

Here let me give you another example of where the 'usefulness by usage' indicator gets wobbly -

Take the Locust and the Atlas. (I know, I know. Different weight classes, but part of your premise is that weight classes are a bit arbitrary.) Specifically, take the AS7-K and the LCT-3S. Both of those are normally regarded as the 'worst variants' of their respective Chassis.

Locust-3S
1 - Required variant for mastery of the chassis.
2 - Cheap.
3 - Dies quickly, leading to faster game turnover.
4 - Low damage and survivability. Low XP gain per match, more matches are required to master it.

Atlas 7-K
1 - Not required for chassis mastery.
2 - Hella expensive.
3 - Dies slowly, leading to slower game turnover.
4 - Decent damage and good survivability, makes XP at a respectable rate, takes fewer matches to master.

Now which of those do you think gets used more often, and by how much?

- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -

Now let's look at a simplified balance by effectiveness in use. For effectiveness here we'll use the average match score across all users. My numbers are indicative of my personal experience, but I'd love a 'word of god' on this.

Locust-3S
Average match score - 22.5
10,000 matches

Atlas 7-K
Average match score - 70.2
5,000 matches

Average player score : 38.4
Median player score : 46.35

Effect multiplier by average :
Locust-3S : 0.5859
Atlas 7-K : 1.828

Effect multiplier by median :
Locust-3S : 0.48544
Atlas 7-K : 1.5146

So even if you're not making a match maker with it directly, I don't think 'use' is the greatest indicator of 'utility'. Sorry, I know this is a bit sloppy as far as arguments go, but I simply don't have the statistics needed to make it bear the burden of proof.

- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -

As far as trying to balance by equipment, that has a lot of hitches (as you say) because a piece of equipment's value will change based on any number of factors. An erPPC in the arm of a DHS spider is more useful than an erPPC in the side torso of a SHS Stalker. Six large lasers will not always be six times as effective, etc. and so on.

Edited by no one, 05 March 2014 - 04:06 PM.

[Varent]

the problem is it really really doesnt.

Let me give a good example.

http://mwo.smurfy-ne...29349b62d5139f5 - Altered with new rules in mind.(4jj)

http://mwo.smurfy-ne...40787ff7a110487

Same mech, very different uses. Depending on the player could have very different elo. Also The second just ebcause how its made probly isnt super viable. cant zombie one of its sides and has a very mixed engagement range.

Also someone may (as ive said before) be cheap and choose to have one mech for different roles. The second mech listed can be an ok brawler in a pinch.

I dont see why your against using weapons as well. It simply makes sense to incorporate them...?

I like your system. I approve of it and am all for it. I just feel you need to add in that last detail.

[Roland]

Varent, on 05 March 2014 - 03:08 PM, said:

the problem is it really really doesnt.

Let me give a good example.

http://mwo.smurfy-ne...29349b62d5139f5 - Altered with new rules in mind.(4jj)

http://mwo.smurfy-ne...40787ff7a110487

Same mech, very different uses. Depending on the player could have very different elo. Also The second just ebcause how its made probly isnt super viable. cant zombie one of its sides and has a very mixed engagement range.

Also someone may (as ive said before) be cheap and choose to have one mech for different roles. The second mech listed can be an ok brawler in a pinch.

I dont see why your against using weapons as well. It simply makes sense to incorporate them...?

I like your system. I approve of it and am all for it. I just feel you need to add in that last detail.

Again, it doesn't matter though to the system.
Your use of the mech doesn't matter. The pilot doesn't matter.

All this is evaluating is how useful the mech variant is, overall, in the eyes of the total player base. If you have two separate uses for that variant? Great! The system does not care at all. Either way, when you drop it, the system says, "Hey, someone thinks this variant is good.. Make note and use it to adjust the value next time." Effectively, your dropping in that mech, regardless of what you plan on doing with it, equates to a "purchase" of that mech at the given BV, and will influence the BV in the next cycle accordingly.

If you loaded it up with garbage weapons and stripped all the armor off? Doesn't matter. The system doesn't care. Just like if you go buy a ferarri, you're gonna pay the same amount even if you don't know how to drive stick and just ruin the transmission before you get off the lot.

Adding in weapon statistics could potentially result in a system where you'd have different BV's for specific loadouts, but it's really not necessary... because from the perspective of generating variety in what we see on the field, seeing different variants will achieve the desired results.

As for why I'm not interested in doing it, it would complicate the system and make it much harder to deal with and implement, and I do not believe that it would actually achieve any additional benefits in terms of the final results.

This system will work without it, so there's no reason to include it. If you can simplify a system and still achieve the desired goals, then you should always do that.

I understand your instinct is that you need information about those weapons, but I'm telling you that you don't. Because all of those things will come out in the wash when you're comparing usage statistics.

Anyway, I added in some information on how to implement this, as well as a visual aide to help folks understand the final result.

The next piece will be how to mesh this with the grouping system to improve variety in the actual game (that is, now that we have a BV, what can we do with it?)

Edited by Roland, 05 March 2014 - 03:54 PM.

[Varent]

I dont think it would complicate the system as much as you think it would. Its simple addition. And since it would be coded into the game itself the math would be done probly before the match even starts and simply saved with the mech itself and your overall stats and the server stats.

There is no reason not to impliment weapons with this system.

Also for those that are frugal this system could be a detriment. Lets say for example the people that only own the original 4 mech bays and (as I listed earlier) want to simply use those mechs in different variations so they can continue to play a free to play game. They would be punished for using a mech outside of the meta in a different variation and dropped into a higher elo simply because they are using X chasis. If the system accounted for the weapons this could be avoided.

[Roland]

Again, this system isn't going to affect your elo. You won't get dropped into a higher elo for using better or worse mechs.

I'll get into this in more detail, but for now try to think of bv not as an analog to elo rating, but rather as a replacement for tonnage in a limiting system.

Remember when pgi talked about wanting to do tonnage limits, but found that it didn't handle cases like bringing all lights and assaults?

This system solves THAT problem. Instead of saying, ”your team must bring a certain number of these specific mech classes", we will say, "your teams can only bring a total BV of X".

It will facilitate the variety that players wanted when they asked for tonnage limits, but will enable dramatically more flexibility on the part of players and teams.

[Roland]

no one, on 05 March 2014 - 02:57 PM, said:

I'm saying the best indicator of how useful a given 'Mech is would be how effective that 'Mech is in combat.

Ah, but what I would suggest is that the mechs that are most effective in combat, will end up being used the most.
People choose to drop in the most effective mechs, which increases their frequency. That's why when some particular mech or weapon is overpowered, you see it all over the place, and then folks complain about it on the forums.

That's how this works... Because utility in combat directly translates into usage frequency.

Quote

Take the Locust and the Atlas. (I know, I know. Different weight classes, but part of your premise is that weight classes are a bit arbitrary.) Specifically, take the AS7-K and the LCT-3S. Both of those are normally regarded as the 'worst variants' of their respective Chassis.


Now which of those do you think gets used more often, and by how much?

If I had to guess, I would guess that both of those mechs gets used extremely infrequently compared to other mechs in the game. You rarely see either on the field.

So, the result in this system is that as long as there are other, better options availible for a similar price, BOTH those mechs would have a low BV price.

Now, what would likely happen is that when other assault mechs became more expensive, and pilots started to look for other more "economical" options (from a BV point of view), that they would then tend to move to the Atlas K more than they would the Locust.. So, what would then happen is that the Atlas K's price would go up, while the locust's price would stay low (at least for a while).

But of course, it's hard to really predict exactly what would happen.. which is why you use a market system instead of trying to decide on what the "real" value is yourself. And that value will show up in the frequency which players choose to drop in that mech.

[GalaxyBluestar]

interesting proposal.

the TL;DR crowd; elo matching for builds, thus creating a value index on metas and rubbish and matchmaker seeks to create two equal value teams.

results would be metas only vs same amount of metas and rubbish only faces rubbish. this value index is determined by usage so the well known jager bomb started out with a low index but as it's usefullness became founded amoungest the community it's index woulde rise till ac40 boats would only appear amoungest the poptarts etc etc.

this is all i can theory craft from what's garnered here. Roland could you use your system with the pilot elo included and give us 3-5 example matches so we can see moreclearly the style of balance your system would enable MWO to have? i could be misinterpreting a lot.

could top pilots go into farming noobs if they took a trial out for a spin or just used an build which isn't used as an frequent meta but still potant?

Edited by GalaxyBluestar, 05 March 2014 - 04:48 PM.

[Roland]

GalaxyBluestar, on 05 March 2014 - 04:46 PM, said:

interesting proposal.

the TL;DR crowd; elo matching for builds, thus creating a value index on metas and rubbish and matchmaker seeks to create two equal value teams.

No, this isn't really it.
As I pointed out to Varent, try not to think of this like Elo rating.

Instead, try to think of this as being used in place of a tonnage limit.

Instead of a system where the game says, "Your team can only have 500 tons", you'll have a system that says, "Your team can only have 5000 BV." or whatever.

And the reason it's better than tonnage is because certain mechs, like the Awesome, are generally not used much at all.. Not because they're truly terrible, or can't be used effectively, but rather because other mechs can simply do what they do better.

A system like this would make the awesome cheap (at least for a while), so that someone may choose to take it to enable a teammate to drop in a higher value mech (like, a victor, for instance).

I'll start to work on writing up how this will all work.. I already got the placeholder up there for it. Seems like writing it up would eliminate a lot of confusion.

Quote

results would be metas only vs same amount of metas and rubbish only faces rubbish.

Ah.. no... What the result will be, is that you WON'T be able to make a team consisting only of the most used mechs.
In order to bring some super top tier mech like a highlander, you may have to take some ultra-low tier mech like a locust, in order to balance out the BV and come in under the team's cap.

Quote

this is all i can theory craft from what's garnered here. Roland could you use your system with the pilot elo included and give us 3-5 example matches so we can see moreclearly the style of balance your system would enable MWO to have? i could be misinterpreting a lot.

Again, this really isn't a matchmaking system. However, I think that once I lay out how it's used (and how it's clearly different from something like Elo rating), things will be more clear.