136 replies to this topic

[AC]

What was your confidence interval for these statistics?

[Alistair Winter]

AC, on 01 October 2013 - 11:23 AM, said:

What was your confidence interval for these statistics?

[FupDup]

AC, on 01 October 2013 - 11:23 AM, said:

What was your confidence interval for these statistics?

Oh god, don't remind me of Stats class...

[Trauglodyte]

AC, on 01 October 2013 - 11:23 AM, said:

What was your confidence interval for these statistics?

You can't really bring confidence into the equation. As it stands, these are all hand entered stats based on small sample sizes based on a number of variations that include:

• Player skill
  
• Tech level
  
• Number of efficiencies
  
• Voice chat
  
• Teamwork and premades
  
• Weapon loadouts

What Alistair and the rest of us are doing is simply presenting the base stats to describe the current game environment. PGI has the much better info. We're simply finger painting right now.

Edited by Trauglodyte, 01 October 2013 - 12:23 PM.

[Trauglodyte]

Now, to continue my stats from earlier, I'm seeing the following:

• Matches averaged out with the winning team weighing in at 786.5 tons versus 786 tons for the losing team (well done match maker).
  
• The winning team was heavier in only 9 out of the 20 games tabbed.
  
• The greatest weight difference was 185 tons (the heavier team won) but this was due to the losing team dropping with only 11 players. The smallest difference was 5 tons.
  
• The weight class distribution looks like this: 78 Lights, 88 Mediums, 167 Heavies, and 145 Assaults.
  
• The Atlas was the most popular mech with 72 of them played over the 20 games while the Trebuchet was the least played with only 4 in 20 games played.
  
• The Atlas had 44 more uses than the next most popular Assault mech (Stalker at 28) while the Light, Medium, and Heavies only had 6, 6, and 4 more uses than the next most popular mech in the respective weight classes.

[Trauglodyte]

Oh, and one more thing, while I cannot confirm that ECM was present on all of the chassis involved, I went back through my screen shots and there were:

• 7 CMD-2Ds
  
• 25 SPD-5Ds
  
• 6 RVN-3Ls
  
• 6 CDA-3Ms
  
• 23 AS7-D-DCs

Totals at 67 ECM capable mechs out of the 478 (14%) mechs in my sample.

[The Boz]

What was the ECMable mechs win %?

[Trauglodyte]

The Boz, on 01 October 2013 - 01:11 PM, said:

What was the ECMable mechs win %?

7 of the 20 (35%) games the winning team had more ECM capable mechs than the losers
7 of the 20 (35%) games the winning team had less or no ECM capable mechs than the losers
6 of the 20 (30%) games the winning team had and even amount of ECM capable mechs

So, in my very very small sample, it doesn't look like the number of ECM mechs was that impactful. Though, it should be noted that I don't know how many of the 67 mechs capable of ECM carried it, how good they were if they had it, or the correlation between the number of ECM capable mechs, the weight differences in those matches, or how many Lights, Mediums, Heavies, or Assaults were present. And that doesn't even bring into the equation skill, premades, voice chat, efficiencies, etc.

[New Day]

Alistair Winter, on 25 September 2013 - 11:05 AM, said:

My base stats here and my mech stats for all mechs with 50+ matches here.

I'm usually pug'ing, or playing with a friend of mine. All these 20 matches were pug drops.

EDIT: Of all my most played mechs, my YLW has the best KDR and second best W/L ratio. What's up with that?

I'm better in my 8q than you in any other mech, hmmm....

[Alistair Winter]

NamesAreStupid, on 01 October 2013 - 01:44 PM, said:

I'm better in my 8q than you in any other mech, hmmm....

Uhm, good for you? I'm not sure what your point is.

Corpsecandle asked me if I would be comfortable sharing my stats, because it was relevant to the discussion. I assume he asked me if I was comfortable because some people don't like showing their stats, in case some troll starts comparing e-peens. Needless to say, I don't give a flying fiddle if some stranger on the internet (e.g. you) is better at this computer game than I am, so I gave all the stats I thought might be relevant.

You're a good player, NamesAreStupid. I salute you for playing a hundred matches in the AWS-8Q. If you're mostly pugging, like me, then those stats are impressive. Reward yourself with a cookie and a cup of tea, and enjoy a short break in front of your fireplace while you reflect on your success and gently pat yourself on the shoulder, crying a few proud tears over all you've accomplished.

[Shadey99]

Some early numbers from my Spider runs:

Most Common Lights (adjusted out of my bias):
* Jenners @ 34.69% of all lights and 7.17% of all mechs (JR7-F the most common variant 22.45%/4.64% on it's own)
* Spiders @ 26.53% of all lights and 5.49% of all mechs (SDR-5D the most common variant)
* Ravens @ 12.24% of all lights and 2.53% of all mechs (RVN-3L the most common variant)
* Commandos @ 2.04% of all lights and 0.42% of all mechs (COM-1B the only variant)
Made up 20.68% of all mechs in my games (including my SDR-5D)

Most Common Mediums:
* Hunchback @ 26.19% of all mediums and 4.64% of all mechs (HBK-4SP the most common variant)
* Centurion @ 23.81% of all mediums and 4.22% of all mechs (CN9-A the most common variant)
* Blackjack @ 21.43% of all mediums and 3.8% of all mechs (BJ-1 the most common variant)
* Trebuchet @ 14.29% of all mediums and 2.53% of all mechs (TBT-7M the most common variant)
* Cicada @ 11.9% of all mediums and 2.11% of all mechs (CDA-3M/2A tie)
* Kintaro @ 2.38% of all mediums and 0.42% of all mechs (KTO-19 the only variant)
Made up 17.72% of all mechs in my games

Most Common Heavies:
* Jagermech @ 38.64% of all heavies and 14.35% of all mechs (Firebrand)
* Cataphract @ 17.05% of all heavies and 6.33% of all mechs (CTF-4X)
* Catapult @ 17.05% of all heavies and 6.33% of all mechs (CPLT-K2)
* Orion @ 13.64% of all heavies and 5.06% of all mechs (ON1-V)
* Dragon @ 12.5% of all heavies and 4.64% of all mechs (DRG-5N)
* Quickdraw @ 1.14% of all heavies and 0.42% of all mechs (QKD-5K)
Made up 37.11% of all mechs in my games

Most Common Assaults:
* Atlas @ 51.72% of all assaults and 12.66% of all mechs (AS7-RS, nearly all champions)
* Stalker @ 13.79% of all assaults and 3.38% of all mechs (STK-5M)
* Highlander @ 13.79% of all assaults and 3.38% of all mechs (HGN-733C)
* Victor @ 13.79% of all assaults and 3.38% of all mechs (VTR-9B)
* Awesome @ 6.9% of all assaults and 1.69% of all mechs (AWS-8Q)
Made up 24.49% of all mechs in my games

I haven't finished the faction data or the entries for kills/assists/damage.

As of this update all the early stats are complete.

Btw this distribution shows Heavies as the most popular weight class currently, followed by assaults, then lights, and finally mediums. Which is the classic trend.

Edited by Shadey99, 02 October 2013 - 10:10 AM.

[AC]

Trauglodyte, on 01 October 2013 - 11:48 AM, said:

You can't really bring confidence into the equation. As it stands, these are all hand entered stats based on small sample sizes based on a number of variations that include:

• Player skill
  
• Tech level
  
• Number of efficiencies
  
• Voice chat
  
• Teamwork and premades
  
• Weapon loadouts

What Alistair and the rest of us are doing is simply presenting the base stats to describe the current game environment. PGI has the much better info. We're simply finger painting right now.

haha, it was a joke, but it got quite the reaction.


More seriously though, since you answered me seriously.... You are 100% correct. If PGI presented us with the server stats, I am sure we as a community have enough skilled people in the Statistics area to run legitimate statistical tests to help them balance the game based on what the community as a whole is doing in matches.

I can pretty much guess that PGI doesn't have the skill to run these tests themselves, but nor should we expect them to. They are programmers and developers and artists. I really wish they would trust the community more and allow us to help them. It would go a long way to build some of the trust that was lost recently, not to mention.... IT IS FREE HELP! lol

Edited by AC, 01 October 2013 - 04:40 PM.

[AC]

Trauglodyte, on 01 October 2013 - 12:05 PM, said:

Now, to continue my stats from earlier, I'm seeing the following:

• Matches averaged out with the winning team weighing in at 786.5 tons versus 786 tons for the losing team (well done match maker).
  
• The winning team was heavier in only 9 out of the 20 games tabbed.
  
• The greatest weight difference was 185 tons (the heavier team won) but this was due to the losing team dropping with only 11 players. The smallest difference was 5 tons.
  
• The weight class distribution looks like this: 78 Lights, 88 Mediums, 167 Heavies, and 145 Assaults.
  
• The Atlas was the most popular mech with 72 of them played over the 20 games while the Trebuchet was the least played with only 4 in 20 games played.
  
• The Atlas had 44 more uses than the next most popular Assault mech (Stalker at 28) while the Light, Medium, and Heavies only had 6, 6, and 4 more uses than the next most popular mech in the respective weight classes.

The Treb part doesn't supprise me. When PGI buffed the medium class, a fully leveled treb is really hard to use because the arms move so fast that you lose the fine metering required of the arm energy weapons. Ie... IT IS DAMN HARD TO AIM! I wish I could remove the pilot unlocks for this mech to slow down the arms. I tried reducing my mouse sensativity, but that also made the required mouse area on my desk to get full torso swing about the size of a couch cushion at the point where the arms felt about right.

I don't think PGI tested the medium mech 'buffs' at all. Some of them are not so much buffs as hinderances, in the case of the treb.

[shintakie]

Whoa, you actually saw a Commando in a game?

I don't think I've even seen a Commando since the last time they were a trial mech choice.

[aniviron]

Trauglodyte, on 01 October 2013 - 01:30 PM, said:

7 of the 20 (35%) games the winning team had more ECM capable mechs than the losers
7 of the 20 (35%) games the winning team had less or no ECM capable mechs than the losers
6 of the 20 (30%) games the winning team had and even amount of ECM capable mechs

So, in my very very small sample, it doesn't look like the number of ECM mechs was that impactful. Though, it should be noted that I don't know how many of the 67 mechs capable of ECM carried it, how good they were if they had it, or the correlation between the number of ECM capable mechs, the weight differences in those matches, or how many Lights, Mediums, Heavies, or Assaults were present. And that doesn't even bring into the equation skill, premades, voice chat, efficiencies, etc.

I know this thread is about numbers instead of gut feelings, but just throwing this out there; I can count on one hand the number of times I have seen ECM-capable chassis without ECM. It might be a decent working hypothesis to assume that any mech capable of carrying ECM will do so.

Alistair Winter, on 01 October 2013 - 02:02 PM, said:

Uhm, good for you? I'm not sure what your point is.

Corpsecandle asked me if I would be comfortable sharing my stats, because it was relevant to the discussion. I assume he asked me if I was comfortable because some people don't like showing their stats, in case some troll starts comparing e-peens. Needless to say, I don't give a flying fiddle if some stranger on the internet (e.g. you) is better at this computer game than I am, so I gave all the stats I thought might be relevant.

You're a good player, NamesAreStupid. I salute you for playing a hundred matches in the AWS-8Q. If you're mostly pugging, like me, then those stats are impressive. Reward yourself with a cookie and a cup of tea, and enjoy a short break in front of your fireplace while you reflect on your success and gently pat yourself on the shoulder, crying a few proud tears over all you've accomplished.

I have 494 games in an 8Q; can I have a gallon of tea and a whole plateful of cookies? My tears wouldn't be proud even though my stats are better, but I will definitely be crying.

[Alistair Winter]

Attention

While I really appreciate the effort of everyone who's contributed to this thread, I am obliged to point out that we're missing out on the most interesting potential created by all these statistics. Unless you're sharing the actual numbers behind the statistics, the actual spreadsheets, then it's much harder - in some cases impossible - to combine all our numbers for some sort of meta-analysis where we can estimate the average for hundreds of matches.

The only way to get a good idea of what's typical for this game, is to compare all of our numbers. So I really hope more of you will take the time to upload a screenshot of the spreadsheet with the actual numbers behind your statistics.

aniviron, on 01 October 2013 - 05:14 PM, said:

I have 494 games in an 8Q; can I have a gallon of tea and a whole plateful of cookies? My tears wouldn't be proud even though my stats are better, but I will definitely be crying.

Well, they're clearly not going to do anything to balance the AWS-8Q, because they want it to be a "natural progression" from Awesome to Atlas. So I suppose we can all look forward to more crying. They should just rename it the AWS-8QQ
(see what I did there?)

[RF Greywolf]

Spreadsheets!!!! Sorry it took so long.


Standard Stats

Spoiler

Mech Chassis

Spoiler

I am not a spreadsheet guy so sorry if it's kinda messy.

Edited by RF Greywolf, 02 October 2013 - 04:37 AM.

[Shadey99]

Alistair Winter, on 01 October 2013 - 11:53 PM, said:

While I really appreciate the effort of everyone who's contributed to this thread, I am obliged to point out that we're missing out on the most interesting potential created by all these statistics. Unless you're sharing the actual numbers behind the statistics, the actual spreadsheets, then it's much harder - in some cases impossible - to combine all our numbers for some sort of meta-analysis where we can estimate the average for hundreds of matches.

It's just a long process with all the stats I'm working on, so this is going to take awhile to process. I'm estimating Friday at the earliest before I'll have the spreadsheets themselves for download.... If I didn't have to work most of the day it would go faster.

Edited by Shadey99, 02 October 2013 - 10:11 AM.

[Agent 0 Fortune]

I finally got around to uplaoding the screenshots and adding the data to a spreadsheet (Excel 2003 compatible). I have spoiled the screenshots to save space.

Highlights:

Match 1
2 fewer heavy mechs on the losing team. 5 trial mechs on the losing team, it is almost like someone* had it in for this team from the start.

Spoiler

Match 2
Giving up 100 tons put this team behind the 8-ball, and with half the team sand bagging, this was a non-starter.

Spoiler

Match 3
58% Elo accuracy (meaning only 7 of 12 mechs per side matched up). I guess this is what you would call a handicap match, where a much better team is only given 11 players (and a disconnect).

Spoiler

Match 4
Only a 75% (9 mechs per side), then the losing team outdamages the winning team by 35% more damage per pilot (577 damage total). Sadly this was the closes, hardest fought match of the entire set. But it is a good example of why the team with the heaviest mechs wins. They can soak, up more damage.

Spoiler

Match 5
This is one of those matches were I would love to know the Elo of the particpants, because statistically it looks close, but it ended up being a runaway, maybe just the wrong mechs for the wrong map.

Spoiler

Match 6
The perfect match, the stars aligned, the matchmaker got its $#@! together for an unprecidented 92% weight-class balance. And as exciting, close, and hard-fought as this was, it would have been even closer if the losing team wouldn't have had a disconnected Atlas.

Spoiler

Match 7
This is two great matches in a row! I would like to point out that all of these matches were played consecutively from 11am - 1pm PDT Saturday, so I can' think of a reason why some of these matches are so well balanced, while others are not.

Spoiler

Match 8
Another good match-up. I think losing team probably had a few pilots under the average Elo, but despite the final tally, the matchmaker did a pretty good job.

Spoiler

Match 9
The matchmaker tried but this was pretty much a blowout. I this particular match I recognized the top 4 pillots on the winning team and know they are from the same Merc Corp, so this can likely be chalked up to Elo and the matchmaker being unable to effectively balance pre-made teams in with PUGs.

Spoiler

Match 10
And the matchmaker $#&! itslef again. 58% weight class match (7 mechs per team), only 11 mechs on 1 team, 170 ton weight discrepancy. This is the kind of carp that happens when the match maker does not have firm Elo boundaries.

Spoiler

Edited by Agent 0 Fortune, 03 October 2013 - 02:19 PM.

[WVAnonymous]

Anyone have any luck with screen capture and OCR, an actual piece of code, or is everyone still typing numbers...?