28 replies to this topic

[Nightbird]

GentleMouse, on 31 August 2021 - 11:35 PM, said:

I love the analysis, though I'm not sure it was wise to split your mechs up. I know it can seem narcissistic but your own personal performance can have pretty drastic effects on the outcome of a game. For instance with the CTF-0XP I have a 1.27 win/loss ratio across 93 games, and with the CTF-3D only 1.03 across 118 games (ECM master race).

Statistically speaking if you're playing a mech lighter than the mean (about 65 tonnes) then you're more likely to end up on the lighter team than if you were playing a mech heavier than the mean. This means that if your most practiced, best performing mechs are above average tonnage while your worst performing ones are below average tonnage then your data will, whilst looking fine, be quite skewed. Equally this could work in the other direction.

This is true. I wouldn't have been able to do an analysis this way as data I collect would probably not show a statistically significant effect for tonnage. This is because an observer participating in the experiment can alter the outcome.

MW Waldorf Statler, on 31 August 2021 - 08:07 PM, said:

in the Random Multiplayer from MW4 all the Kids drives Assaults with unlimited Ammo and heat and only one Team wins ;-)

in T5-3 the Assaults most the first dying Teammates, with nothing great influence to the Match,im seeing matchs who the complete Assault Lance is diying in Seconds after the Matchstart while run in the Murderball or eliminated from Light Wolfpacks.

This is why stats is important, anecdotes and personal memory can be biased and is considered the worst possible source of information for stats.

[1453 R]

Observer bias is something to watch for, but by the same token there's not many ways for someone who isn't employed by PGI to generate this kind of data other than to cut it from their own matches. Just because the dataset isn't flawless doesn't mean it should be discarded. Perfect is the enemy of good, and this flawed data set can still provide a lot of useful insight.

The data as given seems to back up my admittedly entirely, unscientific-ay-eff gut feeling that truly top-heavy teams win more. Not because of the Terrible Matchmaker or Ohh-Pee ******** Assault 'Mechs or any of that junk, but simply because a team with 5+ assault 'Mechs against a team consisting mostly of leaner, lighter 'Mechs can out-bulk its enemies to death. It takes more time, more ammo, more firepower to bring down top-heavy super-durable Sphere assault 'Mechs; after a while a team on the receiving end of a significant tonnage imbalance is simply worn down and less able to kill yet another assault 'Mech. One or two assault 'Mechs' fatness doesn't make a significant difference; five or six assault 'Mechs' fatness very much can.

We've all been in those matches in our medium or light 'Mechs where we go on a tear, score a relative ****-ton of damage, and still end up dying and losing against the fifth or sixth enemy assault 'Mech simply because the enemy team had that damage to give to you. They could sustain you doing all that damage and still have resources left after you'd exhausted all of yours fighting like crazy, and if everybody else on your team doesn't simultaneously have an amazing On Point game right alongside you, too much weight advantage becomes a difficult hurdle to overcome.

[dubstep albatross]

GentleMouse, on 31 August 2021 - 11:35 PM, said:

I love the analysis, though I'm not sure it was wise to split your mechs up. I know it can seem narcissistic but your own personal performance can have pretty drastic effects on the outcome of a game. For instance with the CTF-0XP I have a 1.27 win/loss ratio across 93 games, and with the CTF-3D only 1.03 across 118 games (ECM master race).

Statistically speaking if you're playing a mech lighter than the mean (about 65 tonnes) then you're more likely to end up on the lighter team than if you were playing a mech heavier than the mean. This means that if your most practiced, best performing mechs are above average tonnage while your worst performing ones are below average tonnage then your data will, whilst looking fine, be quite skewed. Equally this could work in the other direction.

Who are you replying to here? I think it is to me, but I am not sure. If so, can you clarify what you meant by your first statement?

[dubstep albatross]

1453 R, on 01 September 2021 - 07:35 AM, said:

The data as given seems to back up my admittedly entirely, unscientific-ay-eff gut feeling that truly top-heavy teams win more. Not because of the Terrible Matchmaker or Ohh-Pee ******** Assault 'Mechs or any of that junk, but simply because a team with 5+ assault 'Mechs against a team consisting mostly of leaner, lighter 'Mechs can out-bulk its enemies to death. It takes more time, more ammo, more firepower to bring down top-heavy super-durable Sphere assault 'Mechs; after a while a team on the receiving end of a significant tonnage imbalance is simply worn down and less able to kill yet another assault 'Mech. One or two assault 'Mechs' fatness doesn't make a significant difference; five or six assault 'Mechs' fatness very much can.

I would just like to point out that in my analysis tonnage difference was relative to the other team. The heavier team does not have to be overloaded with assaults, necessarily, or even heavier than "average". Team composition with respect to distribution of tonnage/weight class is something I am working on at the moment. A motivating question for me is, "when a team is heavier than the other team, what makes that team heavier?". For example, did the lighter team have a four man in stealth Fleas? Mean mech tonnage is about 65 tons (median 65 tons), so a four man in stealth Fleas brings the team weight down 180 tons from the mean team tonnage of about 779. As my analysis shows, heavier teams in that scenario tend to win more.

I will agree with your underlying point, though: significant disparity in weight class composition can create the effect you observe and intuitively sense. It's really about the potential firepower (maximum possible modified by effective application of firepower) versus the raw amount of armor (modified by the ability to spread damage among locations and among teammates).

Everyone knows a good Flea pilot can murder, almost without taking a scratch, an assault pilot who is in a vulnerable position irrespective of how that pilot got into that position (bad choices by assault pilot, good choices by Flea pilot, or both, and maybe even the random nature of a running battle). The real question is in the course of a match, across a team of pilots, how often, on aggregate, does this situation occur? I would very much consider the scenario I described as an outlier that is both generally attributable to unusual high skill by the Flea pilot and unusual low skill by the assault pilot, and attributable to favorable battlefield conditions.

[Capt Deadpool]

dubstep albatross, on 01 September 2021 - 09:28 AM, said:

Everyone knows a good Flea pilot can murder, almost without taking a scratch, an assault pilot who is in a vulnerable position irrespective of how that pilot got into that position (bad choices by assault pilot, good choices by Flea pilot, or both, and maybe even the random nature of a running battle). The real question is in the course of a match, across a team of pilots, how often, on aggregate, does this situation occur? I would very much consider the scenario I described as an outlier that is both generally attributable to unusual high skill by the Flea pilot and unusual low skill by the assault pilot, and attributable to favorable battlefield conditions.

I can tell you with certainty that the frequency of such encounters has diminished significantly in recent months, as the weapon buffs and new PPFLD velocities have made the battlefield more dangerous for lights. Combined with the fact that new Polar, Canyon, and Manifold have dramatically reduced nascaring--at least at higher tier levels--as people are finally starting to understand the how to hold and defend advantageous positions, and this behavior is actually spreading to other maps.

Overall, weapon buffs and new player behavior patterns/tactics effectively have buffed the assault class (holding defensive positions both allows assaults to utilize their new shiny toys better while also preventing them from being in a position where they can be attacked by their natural counter--lights) class, while both changes are nerfs to lights.

Note I am not saying less nascaar is bad for the game by any means, only that it means there are now less targets for lights to take advantage of, and lights were essentially a novelty class even before these changes were made, making them even moreso now.

[MW Waldorf Statler]

statistics was the ground thats the the Countrys with the better Tech and Firepower thinking to win against Countrys with lower Tech and Firepower...the History is full of this Fails.

Edited by MW Waldorf Statler, 01 September 2021 - 08:19 PM.

[Nightbird]

MW Waldorf Statler, on 01 September 2021 - 08:18 PM, said:

statistics was the ground thats the the Countrys with the better Tech and Firepower thinking to win against Countrys with lower Tech and Firepower...the History is full of this Fails.

I assure you no statisticians participate in war planning lol.

[Kiiyor]

Nightbird, on 01 September 2021 - 08:54 PM, said:

I assure you no statisticians participate in war planning lol.

With all the arguing, they'd be dead loooooong before the enemy, after they tried to refute some general's plans by insisting that he wasn't taking standard deviations into account.

[Nightbird]

Kiiyor, on 02 September 2021 - 01:41 AM, said:

With all the arguing, they'd be dead loooooong before the enemy, after they tried to refute some general's plans by insisting that he wasn't taking standard deviations into account.

The only roles statisticians have in war might be tallying the aftermath. We don't talk to generals at all. There's absolutely nothing wrong with experts performing their specialty, no plumbers pretending to be heart surgeons here nosiree.

Edited by Nightbird, 02 September 2021 - 05:38 AM.