28 replies to this topic

[MrJeffers]

Raggedyman, on 02 August 2015 - 08:58 AM, said:

In TT they set the numbers, got back an incomplete set of data by asking player their anecdotal opinion based on about 0.001% (if that) of all the games of TT ever played (including lord knows how many on homebrew rules or misunderstood actual rules) and then made another pitch at it every few years to gague the impact via sales figures and fanmail.
That is somewhat different to the ability of a video game to record every single game in as much detail as wanted.
Please note that I am also saying that it gives you the ability to revise the points multiple times until you get it right, I would expect and hope to see more revisions than in the TT version, to begin with I would hope to see them weekly if not daily.

And it is really that simple, as what I have described is basically how MMO's (and other game styles) calibrate weapon costs/powers/points/whatever

You give a weapon a bunch of stats and a rough points/cost/power/tier/whateveryourmarkeriscalled value
You run it in game a couple thousand times (simulation, QA department, or player base)
You see how the results marry up to what you thought would happen
You move the stats or the points
You run it again (and again and again and again)


Simple does not necessarily mean "easy" or "not a ballache"



You score it by assigning it a rough value, then seeing how it works out when you look at the big data.
And then you revise it until you get it being correct about 95% of the time.




The reason that people have complex discussions about how well things are or aren't working is because they have a very limited set of data and are trying to extract very complicated conclusions from it. They also try and do this from the presumption that their data set (their "personal experience") is accurately recorded, has had all variables accounted for, and is an accurate average of what is happening through out the game. Often they also do it from the premise that if their experience or analysis is questioned or discounted/ignored/not acted on as accurate average of all other experiences then they have been slighted and their honour must be defended.

If they had the data from the whole of the game it would be possible to point and go "well over 2000 rounds thats got a pointcost of 3 but a victory result of 4, so it needs a change as they should be equal" (as an example).

The TT battle value system is based on mathematical models (Example here Weapon Battle value Calculator) not some randomly picked numbers that are then play tested by "0.001%" or created by people who don't understand the system.

Everything in TT BattleTech can be equally measured based on hit/damage charts and percentages. It's fairly easy to do there because most weapons do a fixed amount of damage and weapons that don't deal fixed damage have averages that are used based on the dice percentages.
Hit locations are randomly determined with weighting to certain locations being higher to spread damage around so 16 tons of armor is always more survivable than 10 tons of armor, and so on.

None of this is true in MWO. We get to aim for specific locations, so things like an XL engine in an Awesome is a death sentence in MWO when it makes sense in TT. Most weapons in MWO are variable damage and the damage they do is based on the player's skill not the luck of the dice. High or low mount points change the value of any given weapons, as does the number, etc. Even with thousands of game data sets it wouldn't be easy to create a TT like BV system for MWO that has any great accuracy.

And you would have to be able to separate player's skills from the results so that the FS9-S in the hands of an expert player isn't treated the same as the non-basic'ed meta FS9-S build in the hands of a new/low skill player otherwise there is zero value in using it for any matchmaking balance.

Creating an MWO value system that uses an objective model to rate a mech, then gives it quirks according to that rating is far easier to do since it would be rating the potential options of the mech and not being used as a match making balance metric.

Edited by MrJeffers, 02 August 2015 - 02:06 PM.

[LookUpGaming]

Raggedyman, on 02 August 2015 - 12:30 AM, said:

It's surprisingly easy to do. You start with some core assumptions (small lasers are worth 1pt, each bit of armor is worth 1pt) and scale up from there ( weapon X does twice the damage, its worth 2pts, weapon y does 2x dmg and 2xrng so its worth 4pts etc). You then use the masses of stats feedback from game performance to fine tune the values based on real usage by real players.

The hard part is picking the points difference you'll allow between teams (the closer you want the Balance the longer players will have to wait for their drop) and being willing to weather the storm of individual moans (if you have 99% of 10000 matches be a 50/50 victory chance hundreds of people will get highly vocal that they have lost three in a row).

I disagree. You can't just assign a linear value based off damage. You have dozens of other metrics that come into play. Damage per Heat, Range, DPS, is it Pin Point Front Loaded Damage. How fast is the mech the weapon systems are mounted to, how fast can it redeploy them to another area where it is needed. How much does it matter?

ECM would have to have a different value based on the mech it is attached too, ECM on a locust is 'worth' a different value than ECM on a Cauldron Bourn or Atlas.

Porting over BV from tabletop to MWo is a horrible idea. Bad Bad Bad,

[Davegt27]

Mors Draco, on 02 August 2015 - 11:19 AM, said:

Tweeted Russ earlier today to clear up confusion that pervades the forums about Battle Value.

I knew from listening to the Town Hall meeting that the MWO BV had nothing to do with Battletech TT BV

It was more of "for a lack of a better term" type statement

I played a lot of TT/bookcase games and most function in a similar manor

I was wondering how TT battletech works

I see now that BV is only one aspect of the TT game I should have asked
How the game functions in its basic game mechanic

For example Tactical Air War TT the game turn is 2.5 seconds hex is 600 ft (I think)
Stuff like that

[EgoSlayer]

Davegt27, on 02 August 2015 - 07:53 PM, said:

I knew from listening to the Town Hall meeting that the MWO BV had nothing to do with Battletech TT BV

It was more of "for a lack of a better term" type statement

I played a lot of TT/bookcase games and most function in a similar manor

I was wondering how TT battletech works

I see now that BV is only one aspect of the TT game I should have asked
How the game functions in its basic game mechanic

For example Tactical Air War TT the game turn is 2.5 seconds hex is 600 ft (I think)
Stuff like that

TT Battle Value is a unit limit that is used to balance equipment relative to each other. E.g. you play matches with some set total BV, say 10,000 on each side.

The IS Urban Mech UM-R60 has a BV2.0 value of 504.
The Clan TimberWolf Prime has a BV 2.0 value of 2,737.

So for example the IS player could field 19 Urban mechs with a Total BV of 9576 vs 3 TimberWolves with a total BV of 8211 and 2 Myst Linx (BV 871 each) for a total of 9953. Or something there about - the key being trying to get your unit numbers to add up to as close to 10,000 as possible.
That, in theory is a balanced match. In reality it isn't, just because of the sheer numbers of Urbies they will probably win as long as it isn't an open range map. Obviously you can mix and match any types of mechs to get the right balance that you are looking for as long as they total 10,000 points or less.

[Nightshade24]

xe N on, on 02 August 2015 - 01:56 AM, said:

I really hope, the "BV" system accounts in hitboxes, size and hard point location as well. Otherwise it will be quite useless.

They said the "MW: O" BV value will consider hardpoint locations.


But the problem with hitboxes is that it depends on the player.

I think mechs like the Catapult has the best hitbox in game due to large CT and small ST's but hate the hitboxes of a stalker because its' the oppasite- small CT and large ST.

While other people are the complete oppasite, they love the stalkers hitbox and hate the catapults.


How do you assign a BV to what say... half the game think is the best and completely OP and the other side thinks the other side is and vice versa... Unlike weapons or what have you, there is no deffinitive answer.

In the end- hitboxes is impossible to put a BV system, if PGI do it they will do it to there likings and tastes and experience.

that's the difference between making the Nova a T1 mech in BV's eyes for the hitbox and then say a Timberwolf is a T4 for hitboxes...

Also how will you arrange hitboxes for mecsh that change drastically with weapons on or off? timberwolf and arctic cheetah with LRM's?

[Raggedyman]

MrJeffers, on 02 August 2015 - 02:05 PM, said:

The TT battle value system is based on mathematical models (Example here Weapon Battle value Calculator) not some randomly picked numbers that are then play tested by "0.001%" or created by people who don't understand the system.

I didn't say that.
What I said was the TT developers set the numbers and then got feedback on how well they worked from a very small portion of the actual number of games played with those numbers, including games that would have been played with the rules incorrectly applied.

I was trying to show how data collection, both in depth and accuracy, for a table top game is different to that in a single source video game.

MrJeffers, on 02 August 2015 - 02:05 PM, said:

And you would have to be able to separate player's skills from the results so that the FS9-S in the hands of an expert player isn't treated the same as the non-basic'ed meta FS9-S build in the hands of a new/low skill player otherwise there is zero value in using it for any matchmaking balance.

Except that ELO exists and time played is an already recorded stat.

The data is there

MrJeffers, on 02 August 2015 - 02:05 PM, said:

Creating an MWO value system that uses an objective model to rate a mech, then gives it quirks according to that rating is far easier to do since it would be rating the potential options of the mech and not being used as a match making balance metric.

How is what I have described not an objective model?
What I'm saying is "you can look at the data and use that to balance things"
Your objections so far as 'but what about this variable?' and my answer is 'it's in the data so it can be looked at'
TBH all the quirks given so far have had a "well, we [PGI] looked at the data" statement at some point so we know some form of this is already happening.

Will it give a 100% accurate result?
No, and I've never said it would. In fact I've repeatedly said it wouldn't and it can't. Because the last 5% is a nightmare and the final 1% is madness inducing.

LookUpGaming, on 02 August 2015 - 04:11 PM, said:

I disagree. You can't just assign a linear value based off damage. You have dozens of other metrics that come into play. Damage per Heat, Range, DPS, is it Pin Point Front Loaded Damage. How fast is the mech the weapon systems are mounted to, how fast can it redeploy them to another area where it is needed. How much does it matter?

ECM would have to have a different value based on the mech it is attached too, ECM on a locust is 'worth' a different value than ECM on a Cauldron Bourn or Atlas.

Porting over BV from tabletop to MWo is a horrible idea. Bad Bad Bad,

A TT port would be horrible, as this is not the TT game so it wouldn't work and it would be a waste of good data.

My intention was not to suggest using straight linear values based off of damage alone, I was simply using the small laser as a starting point on how to model things as a beginging point. You can complicate and complicate it further and further as invariably needed, especially as a computer will be doing most of the heavy lifting.

Deathlike, on 02 August 2015 - 12:31 PM, said:

So, let's say you do assign a value to something. The question then becomes, what does this value actually mean?

Understandably some mechs will have greater BV than others because they are better. That's probably not that complicated. HOWEVER, between "meddling variants", how does number "describe the mech" then?

For instance, the Maddog has enough potential firepower through tonnage to be on the level of the Timberwolf. Besides the obvious tonnage differential, how does one accurately portray the "effectiveness" of the low mounted hardpoints of a Maddog vs a JJ-capable Timberwolf that has some higher mounts - such as that the Maddog is only as effective on even ground whereas the Timberwolf doesn't suffer from that as greatly?

Even then, how does one compare the Maddog compared to a Summoner? Summoner has limited diversity. The Summoner can have a semi-Maddog loadout, but it is more durable than a Maddog at accomplishing the same thing. How does that translate into numbers?

It's easier to kinda see "through individual numbers" how it can translate.. both the Maddog and Dragon are not durable mechs. The Quickdraw isn't too far behind. However, it's very difficult to combine all these attributes into "just one number". It honestly wouldn't really work.

On the other hand, PGI gives misleading firepower values in their own mechlab... as an MG and Flamer "on PGI's firepower stat listing", they are way higher than their actual value in combat. This in itself skews actual reality and practicality of the weapons involved.

My point is that I can see putting numbers to "each type of aspect" in a mech... like speed, or firepower or durability.. etc. Putting it all in one number tells you little to nothing contextually to A NEW PLAYER, let alone people that are even "trying" to understand the numbers what those values actually mean. That's why it JUST WON'T WORK, no matter how hard you try to make out the numbers.

"what does this value actually mean" is a good question. The best answer I can give to the method I'm suggesting could be used in MWO (and is used in a lot of other games) is that it would mean that in the majority of incidences if two sides with the same points value and equal experience meet in battle then it would get 50/50 odds as to who would win. Which is not to say that luck would decide the outcome, just that before things start if you had the equipment lists you wouldn't be able to call the fight.

As for value to A New Player: my understanding was that like with ELO this was going to be a background/"hidden" stat, so I wouldn't expect them to know about it to begin with.


TL : DR
*The data is there, most of it is already being collected by the game.
*Perfect balance/accuracy is impossible, especially for a game in flux. However that doesn't stop 95%+ accuracy
*Outliers will happen, as will individual cases. However it's about the big numbers and thousands of games, not "that one time when...."
*There is way less variation in this game than people think, or that they think than can be computed.
*If a system is dynamic/operates with a strong feedback system then improvements can be made and accuracy can be improved. Starting points do not denote end points, unless things are cast needlessly in stone.

Edited by Raggedyman, 04 August 2015 - 04:23 AM.

[MrJeffers]

Raggedyman, on 03 August 2015 - 04:04 AM, said:

Except that ELO exists and time played is an already recorded stat.

The data is there



How is what I have described not an objective model?
What I'm saying is "you can look at the data and use that to balance things"
Your objections so far as 'but what about this variable?' and my answer is 'it's in the data so it can be looked at'
TBH all the quirks given so far have had a "well, we [PGI] looked at the data" statement at some point so we know some form of this is already happening.

Elo doesn't separate equipment from skill it only tracks wins or losses and that is a aggregate of skill and equipment. The only way to do that is have ranked matches were fixed equipment types were used so that skills could be separated from the equipment.

The difference between what you are proposing and what is really happening is that PGI is doing objective modeling on the capability of the mech using a set of requirements/specifications (e.g. number of high mounted energy hardpoints) and then adjusting/normalizing the mechs against each other's relative capability. The way it was done previously was a completely subjective model of player/PGI tier rankings and "Role" without any specifications for what make up the requirements for being in a tier.

What you are proposing is measuring the the capability of the mech based on the skills of who is using it, which can't be identified by the data at hand because they are one measure (Skill+equipment = elo rating based on win/lose), and doesn't result in fair balancing because player skills directly affect how good or bad a mech plays. The data isn't there as you seem to believe. Elo doesn't identify why or how you won or lost, if it was because you were outplayed or out meta mech'ed it's the same result a loss. The data isn't there.

Edited by MrJeffers, 03 August 2015 - 08:16 AM.

[Deathlike]

MrJeffers, on 03 August 2015 - 07:38 AM, said:

Elo doesn't separate equipment from skill it only tracks wins or losses and that is a aggregate of skill and equipment. The only way to do that is have ranked matches were fixed equipment types were used so that skills could be separated from the equipment.

The difference between what you are proposing and what is really happening is that PGI is doing objective modeling on the capability of the mech using a set of requirements/specifications (e.g. number of high mounted energy hardpoints) and then adjusting/normalizing the mechs against each other's relative capability. The way it was done previously was a completely subjective model of player/PGI tier rankings and "Role" without and specifications for what make up the requirements for being in a tier.

What you are proposing is measuring the the capability of the mech based on the skills of who is using it, which can't be identified by the data at hand because they are one measure (Skill+equipment = elo rating based on win/lose), and doesn't result in fair balancing because player skills directly affect how good or bad a mech plays. The data isn't there as you seem to believe. Elo doesn't identify why or how you won or lost, if it was because you were outplayed or out meta mech'ed it's the same result a loss. The data isn't there.

Yea, he really doesn't understand how Elo functions.

Elo doesn't care about your build.

Elo doesn't care about how you win or lose. Iit only cares if you win/lose to an opfor that has a higher/lower average Elo).

There are 4 Elos, one for each weight class.
For instance, a Locust/Mist Lynx is considered the same Elo-wise as a Firestarter/Arctic Cheetah... none of which are not actually comparable directly.

Just because PGI has "telemetry" doesn't mean they know what they are doing with it.

Edited by Deathlike, 03 August 2015 - 07:47 AM.

[RedDragon]

BV as used in TT won't work with MWo because of how they butchered all the stats and the underlying game mechanics.

But that does not mean a BV system won't work in MWO, quite the contrary. PGI has the data from thousands of matches, they could easily assign a baseline of BV values to mechs and equipment. It doesn't have to be perfect at all, because - and that is the best thing about it - you could adjust it on he fly. A certain variant becomes meta and everyone is using it? Add some more BV to it and people can't field too many of them. Another variant doesn't see any use? Lower its BV so people use it to make room for the big guns on the team.