15 replies to this topic

[Fiachdubh]

Despite the amount of people stating that the servers are flooded with Ravens I was just not seeing it so I decided to take a survey since PGI are unlikely to give us the info. Not to disprove them as I do not doubht them but out of curiosity, it was merely the inspiration for the idea. For 30 matches over the weekend I took note of every chassis variant in the server (my own team at beginning and enemy at the end screen), also which map it was on.

I did have problems such as getting disconnected on 3 occasions that I remember before noting the enemy mechs and I probably made an error or two somewhere along the way, it is not meant to be a definitive announcement on mech usage just my observations.


MAPS:

Frozen City - 2 (6.67%)
Frozen City Night - 5 (16.66%)

River City - 5 (16.66%)
River City Night - 6 (20%)

Forest Colony - 3 (10%)
Forest Colony Snow - 4 (13.33%)

Caustic - 2 (6.67%)
Alpine - 3 (10%)

So no surprise about River City seemingly being on high rotation with over 36% but shockingly I only got Caustic twice, must of been real lucky this weekend.

So on to the Mechs. The overall percentages include my 30 appearances in Hunchies but also calculated the percentage of Hunchies without counting myself.

Lights: Total: 65 (14.38%)
Raven
3L - 19
2X - 5
4X - 5
Total: 29 (6.41%)

Spider
5V - 1
5D - 4
5K - 2
Total: 7 (1.55%)

Jenner
D - 4
F - 10
K - 7
Total: 21 (4.65%)

Commando
2D - 5
3A - 1
1B - 0
1D - 0
Knell - 1
Total: 8 (1.77%)


Mediums: Total: 96 (21.23%) (excl. my 30 HBKs: 66 (14.6%))
Trebuchet
7M - 7
3C - 7
5J - 1
5N - 5
7K - 7
Total: 27 (5.97%)

Hunchback
4G - 5 (3)
4P - 6
4H - 1
4J - 7 (28)
4SP - 7 (13)
Total: 26 (5.9%) (incl. me: 56 (12.16%))

Centurion
A - 3
AL - 7
D - 4
YLW - 2
Total: 16 (3.54%)

Cicada
3M - 4
2A - 4
2B - 0
3C - 0
Total: 8 (1.77%)

Heavy: Total: 143 (31.63%)
Dragon
1N - 6
1C - 6
5N - 5
Fang - 1
Flame - 1
Total: 19 (4.2%)

Catapult
C1 - 11
A1 - 17
K2 - 29
C4 - 10
Total: 67 (14.82%)

Cataphract
1X - 13
2X - 11
4X - 11
3D - 6
IlyaM - 16
Total: 57 (12.61%)

Assault: Total: 138 (30.53%)
Awesome
8Q - 2
8R - 4
8T - 2
8V - 3
9M - 4
PB - 7
Total: 22 (4.87%)

Atlas
D - 27
D-DC - 25
RS - 16
K - 5
Total: 73 (16.15%)

Stalker
5M - 11
3F - 13
3H - 7
4N - 5
5S - 7
Total: 43 (9.51%)

Total mechs: 453 Average 45.1 mechs per mach (some 7 v 8s, some I was disco before counted enemy mechs).


So there we have it. The Atlas and Catapult each appear more often than all lights combined and three variants (COM 1B, CDA 2B, CDA 3C) did not appear even once.

I apologise for the presentation it was just a quick off the cuff survey, also there is one math error (I based % off of 452 mechs, realised there were 453 while writing this). There maybe more so let me know and I will correct it.

I will let ye take from it what ye will and am expressing no opinions so please do not flame me (here or in game )

I did this over the weekend just for my personal curiosity, maybe no one cares but if enough people were interested in this type of info I may do it again, perhaps over a whole month but this time taking end game screen shots to reduce errors and make it easier on myself and then put it together in a clearer more suitable format. The forum software does not allow for the neat and orderly dissemination of information.

Enjoy.

Edited by Fiachdubh, 03 March 2013 - 03:56 PM.

[Stone Profit]

Data sample too small. Leads to flawed results. Good try tho.

[Pat Kell]

This was interesting, figured the trebuchet would have made more of an appeareance. Hopefully pgi could be talked into releasing some data along these lines...nothing major just sort of what you have but something across all matches rather then just what you see...is nice that you gathered this info tho.

[Shismar]

No surprise for me there. I have seen many different mechs all the time, not just Ravens 3L and Atlas D-DC. Sure, they are popular but I also see Catapults, Cataphracts and Stalkers all over the place. Not to mention a good shock of mediums but the Cicada (worst mech in game that's why).

The number of Dragons does surprise me. Random games fluke or maybe they just die too fast to register.

It would be great if PGI would release their numbers. No clue why they don't.

Edited by Shismar, 03 March 2013 - 03:59 PM.

[Derick Cruisaire]

Pat Kell, on 03 March 2013 - 03:53 PM, said:

figured the trebuchet would have made more of an appeareance.

I have noticed this as well. Since the release of the the Trebuchet I have not seen nearly as many as I had expected given the general fervor seen when a new Mech comes into the game.

[The Trice]

We need more players to do this with a screen shots as archive , we may even use fraps(disable showing frames and just using its screenshot option) and we just collect a 50 matches data from each player we may add also in collected info the lose and win of each mech just for fun. Your idea is nice and i am going to do it , if you are interested pm me and we will put you in the team.

[Fiachdubh]

Stone Profit, on 03 March 2013 - 03:49 PM, said:

Data sample too small. Leads to flawed results. Good try tho.

Yes it is very small, had intended on more but due to connection and ping problems that is all I could play. Plus taking records with pen and paper at start and end of each match was a huge pain, really should of just downloaded FRAPS or some such software for screenshots.

The Trice, on 03 March 2013 - 04:03 PM, said:

We need more players to do this with a screen shots as archive , we may even use fraps(disable showing frames and just using its screenshot option) and we just collect a 50 matches data from each player we may add also in collected info the lose and win of each mech just for fun. Your idea is nice and i am going to do it , if you are interested pm me and we will put you in the team.

That idea of taking screenshots and linking to them on photobucket or wherever as a source of info for people occurred to me also. Might look into it after next patch release. Maybe people could submit their screenshots for a larger sample.

Edited by Fiachdubh, 03 March 2013 - 04:09 PM.

[Geadron Kane]

1 of the 15 or so matches I was in yesterday had a team with 7 Ravens. Not seeing Ravens is a rarity.

[Murku]

Surely the composition of mech types in your matches are strongly affected by how high in the Elo ladder you sit.

I imagine they are wall to wall in the top tiers.

[Artgathan]

Stone Profit, on 03 March 2013 - 03:49 PM, said:

Data sample too small. Leads to flawed results. Good try tho.

Technically the recommended number of Data Points in a given experiment is 10 per group involved. So if I run a simple 'Experimental Medicine' vs 'Control Group' study I need 20 participants. 453 is a pretty decent number.

However I will agree that there are other factors that impact this, which include Fiach's Elo score and the time when this survey was taken (late at night, EST time, on Saturdays seems to see more cheesebuilds in my experience).

That said, thanks for taking the time to do this Fiach! The results are interesting!

[Fiachdubh]

Yes, I am clearly not at the top of the elo tree Did think of including my K/D and W/L for the matches in question to give an idea of where I would stand elo wise but I forgot to check what they are before starting (K/D 249/227 1.10 W/L 252/302 overall)

The MM did seem to be going mental all weekend so probably got a wider sample (player skill wise) than would normally get with one person when elo is working properly.

[Stone Profit]

Artgathan, on 03 March 2013 - 04:20 PM, said:

Technically the recommended number of Data Points in a given experiment is 10 per group involved. So if I run a simple 'Experimental Medicine' vs 'Control Group' study I need 20 participants. 453 is a pretty decent number.

However I will agree that there are other factors that impact this, which include Fiach's Elo score and the time when this survey was taken (late at night, EST time, on Saturdays seems to see more cheesebuilds in my experience).

That said, thanks for taking the time to do this Fiach! The results are interesting!

Recomended by whom? The only recomendation ive ever heard was "more is better". Not saying you are wrong, but the scientific method I have always been taught you need at least thousands if not more.

[Walk]

Stone Profit, on 03 March 2013 - 04:36 PM, said:

Recomended by whom? The only recomendation ive ever heard was "more is better". Not saying you are wrong, but the scientific method I have always been taught you need at least thousands if not more.

You can easily approximate a larger population via sampling. Technically, in order to get the "true" numbers, you would need every single game ever played. But obviously, this is a) impossible, and b ) incredibly impractical and could never be accomplished in real world scenarios where statistics are relevant and important. However, small samples from a larger population can lead to a very accurate approximation. For example, hundreds of thousands of games have been played. Sampling 30, 50, or a 100 of them at random will absolutely give you a reasonably accurate approximation of what the real numbers are. At this point, any deviation from the "true" values are statistically insignificant. Most textbooks state that around 30-50 samples is a good number for approximating a larger population.

More is (almost) always better, but more isn't always possible or practical.

Edited by Walk, 03 March 2013 - 04:56 PM.

[Artgathan]

Stone Profit, on 03 March 2013 - 04:36 PM, said:

Recomended by whom? The only recomendation ive ever heard was "more is better". Not saying you are wrong, but the scientific method I have always been taught you need at least thousands if not more.

My university profs. I agree that more is better though (and they would too). It depends on what you're studying. Typically clinical drug trials will have participants numbering in the thousands but the reason for that is that it's easier to achieve a significant (ie "not by chance") result with a larger sample size (and the pharmaceutical companies can afford it), which allows them to market a drug as 'being effective' even though the only reason it showed up as 'being effective' was due to the fact that they had 10,000 data points.

However in psychology trials it's not uncommon to see sample sizes of only a few dozen participants. It's all relative. More is better, but it does not mean that you can't do science without a massive sample size.

[jay35]

Murku, on 03 March 2013 - 04:11 PM, said:

Surely the composition of mech types in your matches are strongly affected by how high in the Elo ladder you sit.

I imagine they are wall to wall in the top tiers.

Since you can only imagine, then that must mean you're not there, which means it isn't affecting you, which means for an average player the concept of an all-powerful and over-represented 3L is a myth and not the normative experience for the average player. I have no idea where I am in the ELO rankings, but I'm a halfway decent pilot and I don't see any over-abundance of 3Ls either. If it is purely a myth in anything other than the very top of the ELO brackets or only really present in 8-mans, then it doesn't affect the majority of the playerbase and it's not something worthy of a nerf except perhaps in some manner that would limit them in team-based (i.e., 8-man) tournament play. But proper tournaments should have limitations to avoid cheese like an all-DDC or all-3L team trying to break the system. It would be very easy to do this with class restrictions.

Edited by jay35, 03 March 2013 - 05:05 PM.

[WANTED]

Interesting list. Just happy I am one of the few who runs the Commando 1B and sometimes I get 400 damage with that bad boy. I only have 1 MD 1 MPL and SRM 2 streak.