A Dynamic Self-Balancing Battlevalue System
#101
Posted 19 May 2014 - 12:21 AM
First exclude all drops in mechs that do not yet have all Elite pilot skills mastered (tons of players grab basic and elite skills on mech variants they do not want to use that are bad, just to be able to master another variant and use it). This will also reduce the affect new players have on determining battle values before they have yet learned what is actually good in practice. This should also prevent trial mechs from throwing the values off.
Secondly as others have mentioned, real money only mechs like Hero, Founders, Champion should be dealt with in a separate pool. If you have only 10%(random number) of the player base having bought a hero mech, and 90% having bought a similar non hero variant, even if the hero one is better the non-hero could have 2-5 times the usage easily.
__________________________________________________________________
Also this may run into an issue, since solo drops ignore battle value, and most drops are solo there is potential to skew values in some important cases. For instance lets say in the case of the 2xPPC, 2xAC/5 meta that 35% of solo players use the Victor-9B, and 5% of solo players use the Victor-9S. Solo players are ignoring battle value (which is probably require fort his to work right) so all the group players just start using the Victor-9S but since there are fewer group than solo drops its battle value will hardly rise. Also these same players could keep an exact duplicate Victor-9B build and use it for solo drops thus artificially keeping the other variant lower in value. (maybe make drops in groups be counted as more valuable for determining battle value to offset this?)
Another issue is having lots of variants/mechs that can run near exactly the same build (Victor-9S, Victor-9B, Victor-DS(hero), Highlander-733c, Cataphract-3D all tend to run the exact same 2PPC + 2 AC/5 build and play near identically right now). For arguments sake lets say Jump 2PPC + 2AC/5 is twenty times as popular as 8xSPL 150+ kph light (Firestarter-9A, Firestarter-9K) Your battle value system though would only show it as being eight times as popular since it is spread across 5 variants instead of 2.
I am not sure how to reconcile these issues properly but I do think your idea has merit and hope you or another can point out a good solution to them.
#102
Posted 21 May 2014 - 09:50 AM
Roland, on 05 March 2014 - 10:27 AM, said:
If this system was applied as a compensation modifier for underutilized mechs. I think we would see a wider range of mechs and not just those optimized for the current meta and or best k/d ratios. Unfortunately compensation is tied directly with mech utility by design. since the player community tends to min max, mech frequency will trend.
A 10x rewards for cb and xp might be suficant to play a locust or Awesome. The bonus could be refined all the way down to variant. model a is used 80% model b,c,d the remaining 20% there must be a reason and i think that is directly tied with combat performance.
A sufficiently large reward for taking a mech that deliberately places me at a combat disadvantage lets the player base choose what in portent to them... money vs. min maxing the meta. jack up the reward and mech type frequency will flat line.
otherwise out of 200 mechs only the 10-12 perceived best will be played regularly.
you can see this happen with weapon usage in some FPS only 4-5 out of how many hundreds of weapons are used.
#103
Posted 25 June 2014 - 02:52 PM
For many players, especially the best ones, they really have no need of additional earnings. Many players are sitting on literally billions of cbills, with basically nothing to spend them on because they already own every mech.
This is why I decided to focus the limitations imposed by mech usage on mech usage... because no matter who you are, if you're playing as a group (as the most effective players are), you will be limited in the mechs you can bring.
It ends up serving a few multiple purposes in this regard. It helps balanced highly organized teams matched up against smaller groups, and also increases the variety of mechs seen (with the bonus of having the less utilized ones put in the hands of players playing in organized groups, thus resulting in them being more effective than if played alone).
This thread popped into my mind again, with the recent introduction of the clans.
A system like this can help equalize the different mechs across the tech line, without requiring that all of the clan mechs be made exactly equal to IS mechs. With such a system, clan mechs could be better than IS mechs, and end up just having a significantly higher Battlevalue. Indeed, this was kind of the point of Battlevalue originally.
#104
Posted 22 July 2014 - 04:32 AM
I think its a very good idea.
It should not be based on useage but on chance of winning the match imo. Thats a simple single number. The whole thing is in the end about predicting winning chances and based on that build team compositions with equal winning chances. So winning chance seems to be the right parameter.
Instead of using BV as a function of each individual mech you could instead use BV(mech_type,component_type) for a much finer representation of the BV of each single build of each chassis. So i.e. a medium laser on a hunchback has a BV that adds up together with all other components on the hunchback to ints final BV.
Ofc this might run into problems of dependencies between the components. I.e. the 9th medium laser on a hunchback with no heatsinks might be less efficient then the 4th medium laser on a hunchback with sufficient heatsinks. So in the whole component space of an individual mech (describing what and how many components it has) there will be interdependencies among them. This space could be clustered in order to assign a BV to "similar hunchback builds". So there would be a cluster of hunchbacks with 4 medium lasers and 18 DHS that have 150 to 160 armor i.e. . the resolution of this clustering could be determined based on the data. There are some approaches that can be taken here i think. There are for sure tradeoffs here between model complexity and required accuracy. Even basic implementations would be only off by so little that a human wouldnt notice anyway.
is this overkill ? i dont think so. Game balances is a major quality aspect of a competitive game and just a little bit of statistics wouldnt be that hard to add.
Edited by MadTulip, 22 July 2014 - 04:36 AM.
#105
Posted 04 August 2014 - 11:04 PM
First, setting a hard cap on this type of BV over a group, such that someone in a group is required to use a far-from-optimal variant/build in order for their teammate to use an optimal variant/build, is critically fun-breaking and unacceptable. There has been plenty of debate in the replies to this over the similarity between this proposed system and ELO, and I think Roland has explained quite effectively that this proposed system is very different from ELO. I'd suggest that we could incorporate more similarity with ELO to solve the limit problem.
To highlight why a hard limit on BV for a group is unacceptable (even if that value somehow moves with the market, following average BV or other market trend), let's consider the case of a team that brings all very far-from-optimal variants/builds (and I'll explain why I also say "builds" in a moment). Ignoring ELO for a moment (maybe by assuming it's equal among all pilots of interest for this scenario), how do we match a team with a total BV far below the limit against other teams? Are all teams with a BV below the limit considered equally "valid," and all teams with BV above the limit considered equally "invalid"? This would seem to substantially reduce the usefulness of BV as a matching tool.
Also, is it possible to set a BV limit on a solo player? Under the proposal here, the answer would seem to be no, since the variant itself is the only element that has a BV, and setting a limit lower than the BV of the most expensive variant would make that variant unusable. If we cannot set a BV limit on a solo player, how do we justify setting a BV limit on a group? How can we match that BV limited group against an opposing team of BV unlimited solo players?
I'd propose instead that BV be used as a distributed factor in matchmaking just like ELO is. The Matchmaker should match high BV teams against other high BV teams, just like it matches high ELO teams against high ELO teams (in principle). Conversely, it would match low BV teams against other low BV teams. We would probably need to combine the two values into a final Matchmaking Factor, but that can be done easily enough. The formula for final Matchmaking Factor would be either a linear combination of ELO and BV or a product of the two, with each factor scaled appropriately based on average ELO and average BV or some other reference point for each value. It would be a fascinating exercise to mine the data to figure out how strongly ELO and BV correlate (based on past data, before implementing this system) and to figure out how strongly each one correlates with victory. If the two are strongly correlated with each other in past data, and both strongly correlated with victory, we should probably make our Matchmaking Factor a product of (appropriately scaled) ELO and BV. I am not proposing that the correlation between BV and victory be tracked going forward; I think Roland's usage-based system is sufficient (and maybe better) for future management of BV once implemented.
I think the discussion of moving similar Weapon Sets to different variants highlighted that we need a factor in BV based on the usage of a particular combination of weapons. If a large number of builds are using that combination of weapons, any mech using it needs to get a "bonus" to its BV. An example would be 2x PPC. Many dominant builds of the last year+ have included 2x PPC in their Weapon Set. This combination needs a "bonus" to BV associated with it. If this system can be implemented, this kind of bonus value for weapon combinations would be a straightforward addition. This is the way we manage the well-known fact that combinations of weapons are at least as big a problem in stagnation of optimal builds and tactics as any features of the bare mech variants.
Since the original writing of this proposal, I think quite a few people involved in community-run leagues have become aware that certain hero mechs (the Dragon Slayer and Ember) currently have an edge in highly-optimized competitive matches. Roland acknowledged in one of the earlier replies that the more restricted access to MC-only variants might skew the system, and that that was an acceptable flaw. I disagree. Usage includes two factors: ownership of the variant, and selection of the variant for use in actual matches. If people that own many mechs are using one or two variants almost exclusively, and not using any of the other mechs they own, I think we need to seriously consider an additional modification to BV calculation to account for that. I think we also need a modification if few people own a variant, but all of them frequently use it. This "relative usage" factor would take some additional thought to calculate correctly, but without it, mechs owned by only a limited subset of the player base might be undervalued. They are being used less frequently, not because they aren't useful, but because external factors prevent people from owning them.
This might give us a way to eliminate the "lagging behind the market" problem, by the way. If, all of a sudden, everyone that owns variant ABC-1D goes from never using it to using that variant frequently, our BV calculation needs to take that into account, even if a very small fraction of the player base owns ABC-1Ds because they previously thought they were terrible. This "surge in relative usage" factor probably needs to be quite a bit smaller than the overall range of BV, but it could be a handy tool to minimize lag if we can figure out the right way to calculate "relative usage" compared with the ownership rate of a variant.
I want to begin and end with compliments on the effort put into this concept. Roland has given this a lot of thought, and I hope someone on the Dev team is paying at least a little attention.
#106
Posted 05 August 2014 - 09:21 AM
#107
Posted 07 August 2014 - 09:47 PM
#108
Posted 13 August 2014 - 11:51 AM
#109
Posted 29 August 2014 - 10:36 PM
Once they segregate clan from is in CW, the easiest implementation would be to start with a 12 man is team and match up enough BV out of clanners to be even. the clans then get a good fight and the IS then get a solid shot at victory.
Best part? little to no nerfs needed.
#110
Posted 30 August 2014 - 04:43 PM
#111
Posted 31 August 2014 - 09:04 PM
Hoffenstein, on 30 August 2014 - 04:43 PM, said:
I'm afraid you may be failing to grasp the critical aspect of the system I proposed, as it's certainly not what Warthunder does.
Warthunder uses a BR system, which is certainly akin to the BV system proposed here. However, the key difference is how those values are determined.
For Warthunder, the developers attempt to manually evaluate planes' utility values, and assign appropriate BR's to each. The problem with such a system is that it's very difficult to correctly establish what the "correct" BR's are, especially given that they will likely fluctuate over time.
The system that I am proposing here bypasses all of that, by automating the establishment of the values, based on statistical analysis of usage statistics.
I realize that the first few posts are quite long, but it may be worth reading through them to get a better idea of what is actually being proposed here.
#112
Posted 08 December 2014 - 09:50 AM
#113
Posted 18 May 2015 - 10:08 AM
You choose what faction you want to log in IS or Clan.
and then there should be a combined lobby for IS vs IS (12vs12,8vs8), Clan vs IS (10vs12, 5vs8), Clan vs Clan (10vs10, 5vs5)
Also the player can start matches and should be able to set the weight class, so only lights vs lights, or heavies vs heavies., everything with a BV cap. where IS has a slight BV advantage of 500 to 2000BV or so.
#114
Posted 18 May 2015 - 12:27 PM
http://mwomercs.com/...w-pgi-saved-cw/
Thanks for making this thread and laying the groundwork. I've been wanting to see a dynamic usage-based valuation system in a BattleTech online game for about 15 years now, ever since Multiplayer BattleTech Solaris had only a couple viable mechs. I'd be really interested to hear if you have anything to add to the thread above.
I also wanted to pose a question here: what do you think about altering the base frequency of each variant? The way you've approached this, every variant is equally weighted, as if an equal number of each were produced. I'd think this is rather unattractive for any future CW logistical framework, if not also for game balance in general. After all, the "mech market" would be skewed by the simple accident of how many variants have been released at any given time. Do you think it'd be unreasonable for PGI to set a sort of "base production" value for each variant (or even just weight classes as a whole), a varied initial level of supply from which market demand drives the eventual price?
Edited by Freebrewer Bmore, 18 May 2015 - 12:29 PM.
#115
Posted 19 May 2015 - 07:34 AM
Roland, on 04 March 2014 - 02:31 PM, said:
What could POSSIBLY go wrong??
Biggest Problem: an organized team splinters into groups of 3 & sync-drops into the queue the instant the queue has a multiple of 12. They either all get in or 3/4 of them sync-drop with pug's and/or 2-3 man group fielding very power drop decks. Or they just "pug" drop while all on the same TS.
Problem with any solution: count number of the same unit tags & send error popup saying "think again, reduce your dropdecks"? Fine, lets make a sh**load of 3-man units named "US01", "US02", etc, or each unit only has 1 person in it total, the rest are aligned but technically unit-less pugs.
The other issue is this relies on the current state of weapon/quirk balance. Lets compare the WVR-6K & 6R. In most cases the 6K will do more damage (imo) because it has less time exposed. However on insanely hot maps with large structures (the sun/hpg CW map), the 6R may end up with an advantage.
It also depends on the skill of the driver (both piloting & building). Someone may suck hardcore with ballistic (eyeballing trajectory, ammo efficiency, etc) so someone may do better in a HBK-4H than a DRG-1N, which is usually NOT the case. Someone may also just suck at building the meta into a metamech. I saw a dual-AC2 Huginn last night, it was one sad panda.
Honestly the problem is some people care to find the metamechs (literally, but can also be a euphemism for skill) while others don't. If they're just lazy (no research and/or no practice), they reap what they sow. Do you want to drive a Pretty Baby because the paint is identical to the paint of their tabeltop Dragon? Well you get kicks from being nostalgic while I get kicks from other mechs into piles of rubble. So we both "win", just to different degrees.
#116
Posted 21 May 2015 - 05:19 AM
Twitteritis - Twi-tter-i-tis (noun, singular): The addiction to Twitter and the ability to only comprehend sentiences up to 140 characters long.
1. Poor Russ, he would never order anything except appetizers when he went out to eat, his twitteritis prevented him from getting past the first part of the menu.
2. Russ's twitteritis was getting worse, he would panic if he couldn't find TL;DR notes at the end of a thread.
3. The reason everything keeps changing in MWO is because Russ's twitteritis prevents him from reading past the first sentence on the change log.
Edited by CarnageINC, 21 May 2015 - 05:23 AM.
3 user(s) are reading this topic
0 members, 3 guests, 0 anonymous users