8 replies to this topic

[Montageus]

Now that the results of the Chassis Challenge are in, it's worth reflecting on what the results tell us about the coming quirks.

(Apologies for the tables, the editor here is... suboptimal)

EDIT: Here's a CSV: <https://dl.dropboxus...5622861/mwo.csv>

Here, I take a first cut at the results by looking at the number of points required to get 10th place. At first blush, the results seem to indicate the common wisdom that a number of the Clan mechs are considerably more powerful (although note the rank of the Nova):

Name Tons Clan Points10
Vindicator 45 0 1790
Nova 50 1 1840
Wolverine 55 0 1850
Commando 25 0 1861
Hunchback 50 0 1871
Spider 30 0 1879
Trebuchet 50 0 1893
Dragon 60 0 1893
Griffin 55 0 1908
Centurion 50 0 1916
Adder 35 1 1917
Kitfox 30 1 1918
Locust 20 0 1920
Thunderbolt 65 0 1923
Shadowhawk 55 0 1927
Victor 80 0 1928
Quickdraw 60 0 1931
Kintaro 55 0 1947
Cicada 40 0 1949
Blackjack 45 0 1971
Summoner 70 1 1985
Cataphract 70 0 2003
Jagermech 65 0 2009
Raven 35 0 2032
Orion 75 0 2032
Catapult 65 0 2038
Firestarter 35 0 2045
Highlander 90 0 2057
Jenner 35 0 2061
Awesome 80 0 2073
Stormcrow 55 1 2088
Banshee 95 0 2115
Battlemaster 85 0 2119
Atlas 100 0 2125
Maddog 60 1 2126
Stalker 85 0 2156
Warhawk 85 1 2165
Timberwolf 75 1 2183
Direwolf 100 1 2280

The difference between the top and the bottom, given the formula: (Kills × 20) + (Kill Assist × 10) + ((Damage Done - Team Damage) ÷ 15) + (Wins × 20)
means that the 10th DWF pilot (2280) gets, say, 2.5 more kills per match than the top VND pilot (1790): (2280-1790/10/20 = 2.45)

But one might expect that tons and technology play an important role here - that is, perhaps with more armor and more guns you might get more kills. A basic regression gives us

Points10 = 1778 + 3.34*Tons + 69*Clan

With an R^2 of about 0.5 (so about half the variation is explained). So for every 5 tons, you get a little less than an extra kill per 10 matches (3.34*5 = 16.7 points over 10 matches), and just for being a Clan mech, you get 3 kills and an assist over 10 matches.

So, taking into account these "expected" differences, we look at the residuals instead - that is, the difference from expected performance given tons and Clan/IS tech. If we sort by that, we get a very different picture:

Name Tons Clan Points10 Predicted Residuals
Nova 50 1 1840 2015 -175
Vindicator 45 0 1790 1929 -139
Victor 80 0 1928 2046 -118
Wolverine 55 0 1850 1962 -112
Summoner 70 1 1985 2082 -97
Dragon 60 0 1893 1979 -86
Hunchback 50 0 1871 1946 -75
Thunderbolt 65 0 1923 1996 -73
Griffin 55 0 1908 1962 -54
Trebuchet 50 0 1893 1946 -53
Quickdraw 60 0 1931 1979 -48
Adder 35 1 1917 1965 -48
Shadowhawk 55 0 1927 1962 -35
Kitfox 30 1 1918 1948 -30
Centurion 50 0 1916 1946 -30
Highlander 90 0 2057 2079 -22
Kintaro 55 0 1947 1962 -15
Cataphract 70 0 2003 2013 -10
Commando 25 0 1861 1862 -1
Spider 30 0 1879 1879 0
Orion 75 0 2032 2029 3
Atlas 100 0 2125 2113 12
Jagermech 65 0 2009 1996 13
Banshee 95 0 2115 2096 19
Awesome 80 0 2073 2046 27
Warhawk 85 1 2165 2132 33
Cicada 40 0 1949 1912 37
Blackjack 45 0 1971 1929 42
Catapult 65 0 2038 1996 42
Battlemaster 85 0 2119 2063 56
Stormcrow 55 1 2088 2032 56
Locust 20 0 1920 1845 75
Maddog 60 1 2126 2048 78
Timberwolf 75 1 2183 2098 85
Stalker 85 0 2156 2063 93
Direwolf 100 1 2280 2182 98
Raven 35 0 2032 1896 136
Firestarter 35 0 2045 1896 149
Jenner 35 0 2061 1896 165

From this we see that, relative to the predicted outcome, that the IS 35-ton lights are actually punching far above their own weight. It's not that a Nova is the "worst" mech - it's that, for its weight and tech, it's just not doing very well.

But looking at this chart, there's something funny going on - lights are doing amazingly well. So maybe there's an inherent penalty due to weight class (movement archetype, large size, etc.) If we put in variables for whether a mech is a light, the picture changes again, since the regression is now:

Points10 = 1643 + 5.2*Tons + 59*Clan +137*Light

Name Light Tons Clan Points10 Predicted Residuals
Victor 0 80 0 1928 2058 -130
Nova 0 50 1 1840 1962 -122
Adder 1 35 1 1917 2021 -104
Vindicator 0 45 0 1790 1876 -86
Summoner 0 70 1 1985 2066 -81
Wolverine 0 55 0 1850 1928 -78
Kitfox 1 30 1 1918 1995 -77
Dragon 0 60 0 1893 1954 -61
Thunderbolt 0 65 0 1923 1980 -57
Spider 1 30 0 1879 1936 -57
Highlander 0 90 0 2057 2110 -53
Commando 1 25 0 1861 1910 -49
Atlas 0 100 0 2125 2162 -37
Hunchback 0 50 0 1871 1902 -31
Quickdraw 0 60 0 1931 1954 -23
Banshee 0 95 0 2115 2136 -21
Griffin 0 55 0 1908 1928 -20
Trebuchet 0 50 0 1893 1902 -9
Cataphract 0 70 0 2003 2006 -3
Shadowhawk 0 55 0 1927 1928 -1
Orion 0 75 0 2032 2032 0
Centurion 0 50 0 1916 1902 14
Awesome 0 80 0 2073 2058 15
Kintaro 0 55 0 1947 1928 19
Warhawk 0 85 1 2165 2144 21
Jagermech 0 65 0 2009 1980 29
Battlemaster 0 85 0 2119 2084 35
Locust 1 20 0 1920 1884 36
Catapult 0 65 0 2038 1980 58
Direwolf 0 100 1 2280 2221 59
Raven 1 35 0 2032 1962 70
Stalker 0 85 0 2156 2084 72
Firestarter 1 35 0 2045 1962 83
Timberwolf 0 75 1 2183 2092 91
Blackjack 0 45 0 1971 1876 95
Cicada 0 40 0 1949 1850 99
Jenner 1 35 0 2061 1962 99
Stormcrow 0 55 1 2088 1988 100
Maddog 0 60 1 2126 2014 112

For those who are interested, there were not major differences when dummy variables were used for each weight class, and AIC/BIC criterion preferred this model over the other two (R^2 = 0.63, or 63% of variation accounted for).

What are the implications of this? Perhaps some of the mechs that appear overpowered are not punching as much above their weight as might be thought, at least when piloted by experienced pilots.

But it is clear that, in general, Clan mechs - given weight and class - have a sizable advantage over IS mechs. But perhaps more interestingly, light mechs have over double the advantage that tech brings over other weight classes (when accounting for tech and tons).

Regardless, though, it is time to give the poor Victor some kind of quirks. Here's the full chart sorted by tons so that the savvy shopper may get the best bang for their hard-earned MC:

Name Light Tons Clan Points10 Predicted Residuals
Locust 1 20 0 1920 1884 36
Commando 1 25 0 1861 1910 -49
Spider 1 30 0 1879 1936 -57
Kitfox 1 30 1 1918 1995 -77
Raven 1 35 0 2032 1962 70
Firestarter 1 35 0 2045 1962 83
Jenner 1 35 0 2061 1962 99
Adder 1 35 1 1917 2021 -104
Cicada 0 40 0 1949 1850 99
Vindicator 0 45 0 1790 1876 -86
Blackjack 0 45 0 1971 1876 95
Hunchback 0 50 0 1871 1902 -31
Trebuchet 0 50 0 1893 1902 -9
Centurion 0 50 0 1916 1902 14
Nova 0 50 1 1840 1962 -122
Wolverine 0 55 0 1850 1928 -78
Griffin 0 55 0 1908 1928 -20
Shadowhawk 0 55 0 1927 1928 -1
Kintaro 0 55 0 1947 1928 19
Stormcrow 0 55 1 2088 1988 100
Dragon 0 60 0 1893 1954 -61
Quickdraw 0 60 0 1931 1954 -23
Maddog 0 60 1 2126 2014 112
Thunderbolt 0 65 0 1923 1980 -57
Jagermech 0 65 0 2009 1980 29
Catapult 0 65 0 2038 1980 58
Cataphract 0 70 0 2003 2006 -3
Summoner 0 70 1 1985 2066 -81
Orion 0 75 0 2032 2032 0
Timberwolf 0 75 1 2183 2092 91
Victor 0 80 0 1928 2058 -130
Awesome 0 80 0 2073 2058 15
Battlemaster 0 85 0 2119 2084 35
Stalker 0 85 0 2156 2084 72
Warhawk 0 85 1 2165 2144 21
Highlander 0 90 0 2057 2110 -53
Banshee 0 95 0 2115 2136 -21
Atlas 0 100 0 2125 2162 -37
Direwolf 0 100 1 2280 2221 59

Edited by Montageus, 28 October 2014 - 05:06 PM.

[DocBach]

You'd do better to like, make an excel spreadsheet to format your data so it's easier to read. I saw the first blob of numbers and it made my eyes hurt, so I quit reading it. I'm sorry to be so blunt after the obvious time you took to make it.

[Bartholomew bartholomew]

I find i amusing that in you first balance list that spiders actually zero out and are therefore the only balanced mech ingame.

Nice work putting this together, thanks.

Some of the unexpected deviance can be explained by various nerfs. There was a nerf to clan Mlas aimed at the wolves. And that alone was what destroyed the nova's viability for example.

But most nerfs were not put in for true balance, but instead as a cave in to player preference.

[Montageus]

Yeah, sorry about the tables...

Here's a CSV for those who want to roll their own (and get readable tables):
<https://dl.dropboxus...5622861/mwo.csv>

[Artgathan]

There's already a dearth of topics on this subject - was it necessary to create another?

Another point: the numbers are meaningless. Consider that if we take these numbers as indicative of some king of 'relative balance', that would mean that the Dragon, Shadowhawk and Locust (all of these scored within 30 points of eachother) all have the same level of 'usefulness', and the Locust is a 'better mech' than the Kit Fox or Spider.

[Lily from animove]

The issue is undertsanding what makes mechs viable and what not.

its speed to dodge damage.
its hiboxes to spread damage
its hitboxes that create a time to exposure defining the ability to negate enemy fire
it's Hardpoints and locations affecting the above.

A dead mech can't deliver damage, no matter how epic his wepaon laodout is.

These are the reasons for a mech being good or not. Its rather class independend. thats why IS lights can score that high. because they are fast and small to be untouchable. Yet they have with their internal DHS enough cooling to sustain some proper damage.

most mediums suck, because they are generally too big, and Nova is so bad because it violates all rules of important criteria. its quite slow for a medium, and has assault size, while not having good armor. You will always hit a Nova where you aim at.
Nova vs SCR is the perfect example why "CLANS" is not a comparison criteria. Clans being clan has nothing to be with being good or not. Its purely the mech itself defining this. Thats unaffected form the IS vs clan discussion. SCR, TBR and DWRF can just rely on some epic amounts of hardpoints being distributed very nice (at least well high placed Torsi hardpoints). And then SCR and TBR are also blessed with good agility for their class and some nice hitboxes giving them survivability.

Pilot a Nova pilot a SCR. both 2 ERPPC, or both ith laser vomit builds. You will see why the Nova sucks. Nova can't outrun threats and can not use some very own tactics. SCR can use the epic torso mounted hardpoints. Which are a huge advantage when making sure to expose a minimum amount of mech. Together with its good speed, this even expoentially increases that fact.

Tonnge is meaningless, clan/IS is meaningless. A mech is a mech by said above features. That whats makes him viable in a FPS like shooter. can't see me, can't hit me. can I be gone within 1 second and dodge that AC 20 and ppcs or can I not? Can I tank the damage or not?

The Nova is a pure no on all this and thats why it sucks. The vindicator sucks because of this as well, it is a easy target to hit where I want my shots to hit, yet its rather restricted in speed making this hitting more easy.

Imagine now Nova and SCR would change purely only their models, I geuss they would be rather equal now, because the SCR chassis at the 50t Nova model would have its advantage by high weapon hardpoints and super torso twist abilities. Yet the Novamodel at the 55t Stormcrow mech even if at disadvantage at mentioned criteris would at leats be more nimble as an exchange.

Think about what in a fps shooter is important. reflect that to MWO and you already know why the things turned out as they did.

DWF, TBR and sCR, imagine now they would have the same hardpoints but were IS mechs, so having the heavier wepaons.They would still be top tier mechs. because its not their "Clan affiliation" making them good, its the hardpoints and engine size allowing you to make a very optimised mech.
In preperation to the upcoming Nova I bought a Phract and equipped it with the fastest XL available and slapped 2 ERLL and 2 LL an a ppc on it. Just to get a feel about how piloting a Nova would be. (low slung arms, XL engine, rather similar speed). Gave me a lot Nova feeling and made hopping over to a Nova a familiar feeling. Yet the Nova sucks simply by being a papermech in atlas wideness. It can not dodge damage, it can not absorb damage and has to expose a LOT.
Take any other mech that is top tier and give it Nova characteristics and it will sololy for this drop some tiers. Unless this mech would be able to go 120kph. Because Even an Atlas has some higher mounted torso hardpoints.

So I can nearly not agree with anything you wrote. Give an atlas a Direwold hardpoint distribution and it would do epic as well, dualgauss + srms and Mlas. just epic. Not a clanfeature. Purely a chassis feature that 2 gauss fit into a DW. Imagine a atlas holding maybe 3 lbx 10, would be devasting. King crab is coming, going to be a beast, yet probably lacking enough E-hardpoints to create a IS Direwolf.

Why is a jenner better than a commando? well time to exposure. Little jenner has some nice hardpoint locations and of course some more tonnage. But imagine a Commando would have 2E in a high mounted shoulder it would totally change the outcome of what commandos suddenly can do. Hill peeking and cornerpeeking is what the lights do. And you win if you do it with minimum time of exposure. That requires the right geometry and hardpoints as well as speed.
Why can the locusts by being lighter than commandos exceed them? Because it has exactly this, its basically a smaller jenner, high mountpoints. fast hit and run. Thats why. Same is cicada vs Nova. change cicada and Novas model, and cicada gets uselss, and Nova gets quite a good mech.

In MWO the mech geometry, speed and hardpoints decide the usefullness of a mech. Nothing else, because IS mechs are basically empty hulls in all the other aspects. Engine size is exchangeble, ES, FF as well, JJ's as well. Yet they perform so differently. Not linear to their tonnage or speed. IS mechs differ only in tonnage, geometry and Hardpoints and some in engine capabilities. The true defining factors of a mechs usefulness.

The Tournament shows this clearly. Yet we can not simply change mech models and hardpoint locations, because by lore and what Battletech is, defines what we will get. Imagine a summoner would have some CT E hardpoints mounted rather high, it would suddenly perform a lot better. Still engine and structure is the same as well as the geometry. The Adder did fairly well, and I told people already if they put their weapons in the torso mountpoints and use that as an advantage they will have a very well doing clan light, even if rather slow. Because it will now be a valid mech able to minimise time to exposure.

It's an endless story about that probably repeating itself if you wanna go and compare all the mechs, but it is what truly makes a mech well in a FPS game. On TT Mech geometry never cared, they just made some cool looking designes. Here it is defining 60% of a mechs performance If I should rate it. Skilled pilots know that and use that for their advantage. The average adam in the mid elo probably not but he feels this when his mech does a lot more worse than another by dying faster and more often.

Edit: some typos, there were a lot, and probably still be many xD

Edited by Lily from animove, 30 October 2014 - 06:16 AM.

[The Boz]

I'd just like to point out that hard data makes me hard.

[Kjudoon]

Why cool artwork does not equal good. Sometimes what makes it look good is really quite bad for performance.

[Angel of Annihilation]

Lily from animove, on 28 October 2014 - 06:03 PM, said:

The issue is undertsanding what makes mechs viable and what not.

its speed to dodge damage.
its hiboxes to spread damage
its hitboxes that create a time to exposure defining the ability to negate enemy fire
it's Hardpoints and locations affecting the above.

A dead mech can't deliver damage, no matter how epic his wepaon laodout is.

These are the reasons for a mech being good or not. Its rather class independend. thats why IS lights can score that high. because they are fast and small to be untouhable. yet they have with their internal DHS enough cooling to sustain some proper damage.

most medums suck, because they are generally too big, and Nova is so bad because it violates all rules of important criteria. its quite slow for a medium, and has assault size, while not having good armor. You will always hit a Nova where you aim at.
Nova vs SCR is the perfect example why "CLANS" is not a comparison criteria. Clans being clan has nothing to be with being good or not. Its purely the mech itself definign this. Thats unaffected form the IS vs clan discussion. SCR, TBR and DWRF can just rely on some epic amounts of hardpoints being distributet very nice (at least well high placed Torsi hardpoints). And then SCR and TBR are also blessed with good agility for their class and soem nice hitboxes giving them survivabikity.

Pilot a Nova pilot a SCR. both 2 ERPPC, or both ith laser vomit builds. You will see why the nova sucks. Nova can't outrun threats and can not use some very own tactics. SCR can use the epic torso mounted hardpoints. Which are a huge advantage when making sure to expose a minimum amount of mech. Together with its good speed, this even potentially increases that fact.

Tonnge is meaningless, clan/IS is menaignless. a mech is a mech by said above features. that whats makes him viable in a FPS like shooter. can't see me, can't hit me. can i be gone within 1 second and dodge that AC 20 and ppcs or can I not? Can I tank the damage or not?

The Nova is a pure no on all this and thats why it sucks. The vindicator sucks because of this as well, it is a easy target to hit where I want my shots to hit, yet its rather restricted in speed making this hitting more easy.

Imagine now Nova and SCR would change purely only their models, I geuss they would be rather equal now, because the SCR chassis at the 50t Nova model would have its advantage by high weapon hardpoints and super torso twist abilities. Yet the Novamodel at the 55t Stormcrow mech even if at disadvantage at mentioned criteris would at leats be more nimble as an exchange.

Think about what in a fps shooter is important. reflect that to MWO and you already know why the things turned out as they did.

DWF, TBR and sCR, imagien now they would have the same hardpoints but were IS emchs, so havign the heavier wepaons.They ould still be top tier mechs. because its not their "Clan affiliation" making them good, its the hardpoints and engine size allowing you to make a very optimised mech.
In preperation to the upcoming Nova I bought a Phract and equipped it with the fastest XL available and slapped 2 ERLL and 2 LL an a ppc on it. Just to get a feel about how piloting a Nova would be. (low slung arms, XL engine, rather similar speed). Gave me a lot Nova feeling and made hopping over to a Nova a familiar feeling. Yet the Nova sucks simply by being a papermech in atlas wideness. It can not dodge damage, it can not absorb damage and has to expose a LOT.
take any other mech that is top tier and give it Nova characteristics and it will sololy for this drop some tiers. unless this mech would be bale to go 120kph. Because Even an Atlas has some higher mounted torso hardpoints.

So I can nearly not agree with anything you wrote. Give an atlas a Direwold hardpoint distribution and it would do epic as well, dualgauss + srms and Mlas. just epic. Not a clanfeature. Purely a chassis feature that 2 gauss fit into a DW. imagine a atlas holding maybe 3 lbx 10 would be devasting. King crab is coming, going to be a beast, yet probably lacking enough E-hardpoints to create a IS Direwolf.

why is a jenner better than a commando? well time to exposure. Little jenner has some nice hardpoint locations adn of course soem more tonnage. But imagine a Commando would have 2E in a high mounted shoulder it would totally change the outcome of what commandos suddenly can do. Hill peekign and cornerpeeking is what the lights do. And you win if you do it with minimum time of exposure. That requires the right geometry and hardpoints as well as speed.
Why cna the locusts by beign lighter than commandos exceed them? Because it has exactly this, its basically a smaller jenner, high mountpoints. fast hit and run. Thats why. same is cicada vs Nova. chaneg cicada and Novas model, and cicada gets uselss, and Nova gets quite a good mech.

In MWO the mech geometry, speed and hardpoints decide the usefullness of a mech. Nothing else, because IS mechs are basically empty hulls in all the other aspects. Engine size is exchangeble, ES, FF as well, JJ's as well. Yet they perform so differently. Not linear to their tonnage or speed. IS mechs differ only in tonnage, geometry and Hardpoints and some in engine capabilities. The true defining factors of a mechs usefulness.

The Tournament shows this clearly. Yet we can not simply change mech models and hardpoint locations, because by lore and what battletech is defines what we will get. Imagien a summoner would have some CT E hardpoints mounted rather high, it would suddenly perform a lot better. Still engine and structure is the same as well as the geometry. The Adder did fairly well, and I told people already if they put their weapons in the torso mountpoints and use that as an advantage they will have a very well doing clan light, even if rather slow. Because it will now be a valid mech able to minimise time to exposure.

It's an endless story about that probably repeating itself if you wanna go and compare all the mechs, but it is what truly makes a mech well in a FPS game. On TT Mech geometry never cared, they just made some cool looking designes. Here it is defining 60% of a mechs performance If I should rate it. Skilled pilots know that and use that for their advantage. The average adam in the mid elo probably not but he feels this when his mech does a lot more worse than another by dying faster and more often.

Your 100% right on the money with this post. Shape and size of a mech is probably the most important factor on how good it will be followed by hard point configuration at a close second. Engine and movement characteristics come third. Followed by weight and available free tonnage and then finally technology which with the recent nerfs isn't all that different between Clans and IS.