136 replies to this topic

[The Boz]

Are you able to gather intel on the correlation between disconnects and losing?
If your method was limited to print-screening the end match screen, disconnected players killed would not be visible...

[Alistair Winter]

The Boz, on 26 September 2013 - 06:24 AM, said:

Are you able to gather intel on the correlation between disconnects and losing?
If your method was limited to print-screening the end match screen, disconnected players killed would not be visible...

Yep, true. To be fair, there's a lot of variables I can't pick up on. For example, players that are AFK or people who disconnect because they're about to die and ragequit. So apply pinches of salt where necessary.

[RF Greywolf]

The numbers are still kinda unexpected, at least for me. There is a slight lean to the heavier team but not as much as I would have expected. Even better is that the amount of games ending in cap are negligible when you look at your spreadsheet (even though the Capwarrior whiners will still whine I guess). I am still kinda saddened to see that mediums are not more often used (not really helping I guess, mainly a heavy pilot here). I am curious if the trend is similar for other players of different brackets. The data is interesting enough that I might have to do something similar...

[ARamaLamaDingDong]

These numbers look alot like the MM is doing a far better job than many in this community are willing to acknowledge.

Edited by ARamaLamaDingDong, 26 September 2013 - 07:23 AM.

[Alistair Winter]

ARamaLamaDingDong, on 26 September 2013 - 06:49 AM, said:

These numbers look alot like the MM is doing a far better job than many in this community are willing to aknowledge.

The matches are certainly more even than I expected. Would not have guessed an average of 5 kills on the losing team. Feels more like an average of 2 kills, sometimes.

[Billygoat]

To give some numbers from a different player's perspective:

Here's my results from 20 matches played this afternoon (my time). All matches are Assault PUG matches played at what I would assume is "average" Elo. I played a medium class Blackjack for all matches (18 in a BJ-1 and 2 in a BJ-3, when my BJ-1 got stuck in a dead match for 15 minutes after a server fart, both have very similar setups).

My spreadsheet is not as nice as Alistair Winters', but it should be functional enough.

Spoiler

Aping some of the interesting results from the OP:

• Distribution of weight class was 21% Assault, 43% Heavy, 21% Medium and 15% Light. Lights were the least played class by a fair margin.

• The winning team made an average of 10.2 kills whilst the losing team made 4.9.

• Players on the winning side did an average of 252.1 damage and the losing teams players did 194.8 damage.

• The winning team had an average of 7.75 heavy and assaults mechs while the losing team had 7.35.

• In 50% of matches the winning team had more heavy & assault mechs. The losing team had more heavies & assaults in 35% of matches. In 15% of matches, the combined total of heavies and assaults was even.

• Interestingly, however, in 55% of matches the losing team actually had more Assault mechs than the winning team, whilst the winning team tended to have more heavies. In only 20% of matches did the winning team have more assaults.

• 6 out of 20 or 30% of matches ended in a cap.

• 60% of matches had at least 1 disconnect or empty slot (that is, the match started with 1 or more mechs simply missing from the match).

A lot of things very similar, but a few interesting differences. Might be just luck, or things might be a bit different if you're in a higher bracket than I am, which is definitely possible.

[King Picollo]

What's interesting is the average damage between winning and losing isn't that much.

[Trauglodyte]

Numbers seem pretty sound, Allistair. The only problem is the continued overweighting towards the heavier mechs. But, that is fixable with weight limitations.

[scJazz]

Excellent work Alistair!

Since the 12 vs 12 Patch I have collected data on every single match I've played. Unfortunately, I haven't been able to data mine all of it. I wanted to complete another project first. Namely, what are the effects of heat on each map. I finished my preliminary report on that last night. You can find that report here.

http://mwomercs.com/...s-heat-and-you/

Here is some data from 100 matches played during the period August 28th 2013 thru September 11th 2013. All matches were played Solo Pugging with Mode Any. Matches were played with a Catapult, Cataphract, Centurion, or Quickdraw. It is my unconfirmable opinion that I am a Mid ELO player.

Average Match Time: 7.9 min
Match Distribution: 55 Assault vs 45 Conquest
Average Blue/Red Team Weight: 65.12 tons, 65.34 tons (my own mech weight is isolated and removed)
Average Team Weight Imbalance: 55.65 tons with a Standard Deviation of 42.33 tons
Win/Loss Ration of Sample: 1.38:1
KDR of Sample: 1.54:1
Assault Matches Ending in Cap: 12 (ONLY 6 evaluated as bad captures though, if the Capping team had only 1 or 2 mechs it counted as a good victory or when Team A started a Cap and Team B piled on and won it)
Assault Matches Bad Ending Percentage: 11%
Number of Matches with at least 1 DC, AFK: 52*
After DC Hotfix Patch: 8 of 16 had a DC/AFK
Average Playing Time: 6.43 min (time until I died or the match ended which ever comes first)
Average Conquest Playing Time: 6.57 min (time until I died or the match ended which ever comes first)
Average Conquest CB: 71192.5
Average Conquest XP: 477
Average Assault Playing Time: 6.31 min (time until I died or the match ended which ever comes first)
Average Assault CB: 81574.6
Average Assault XP: 642.1

*Part of the sample occurs on either side of the DC Hotfix Patch. However the numbers are somewhat similar so I've included them here.

Now that I've finished the whole Maps and Heat thing I'll be working on more of my data. Unfortunately, to do it I'm going to have to write a couple AHK Macros, an application to simplify data mining the screen captures, and torture myself with really bad OCR software trying to capture san-serif fonts on a muddled background.

Hope you enjoy!

[PEEFsmash]

Joseph Mallan, on 26 September 2013 - 03:25 AM, said:

LOL I would not say "only the best" are over 60%.

Depends on how many people you want to put as "the best." I thought about a better modifier there, but I couldn't think of one. In this case, I am saying that "the best" is "top 20%"....which is a very generous best =]

[stjobe]

Trauglodyte, on 26 September 2013 - 07:06 AM, said:

Numbers seem pretty sound, Allistair. The only problem is the continued overweighting towards the heavier mechs. But, that is fixable with weight limitations.

Or by not having only four light chassis available.

There's 11 heavy and assault chassis (6 heavy, 5 assault), and 10 light/medium chassis (4 light, 6 medium).

If every weight class was equally popular, the percentages would look like this:

• Assaults 23.81%
  
• Heavies 28.57%
  
• Assaults+Heavies 52.38%
  
• Mediums 28.57%
  
• Lights 19.05%
  
• Mediums+Lights 47.62%

From this we can confirm what we all knew; Assaults and especially Heavies are over-represented, and Lights and especially Mediums are under-represented to the same degree (even though lights are the least played weight-class, they also have the least number of chassis available).

[RF Greywolf]

stjobe, on 26 September 2013 - 07:32 AM, said:

%From this we can confirm what we all knew; Assaults and especially Heavies are over-represented, and Lights and especially Mediums are under-represented to the same degree (even though lights are the least played weight-class, they also have the least number of chassis available).

Hopefully that will get fixed when the start drop limits. Mediums are, by canon, the most common mechs to see on the battlefield. Heavies and assaults were reserved to higher ranking officers and commanders. I do agree they need to introduce some more light chassis into the game.

[scJazz]

Here is the data from those 100 matches referred to earlier. This represents just a partial mining. I have well over 200 matches of very clean data captures.

https://docs.google....VqUGROV1E#gid=0

[DaZur]

Joseph Mallan, on 26 September 2013 - 03:25 AM, said:

LOL I would not say "only the best" are over 60%.

I've managed to PUG up to .71% before being sent crashing back down to .65% with a long run of losses.

At .71% I'm pretty sure the folks I was playing with / against had their mice biomechanically grafted to their hands and spoke perfect binary... So I'm not really disappointed with the drop.

​Right now for me .65% seems to be my "Elo sweet spot"... as matches are competitive, most players are fairly competent and I'm not being opened up like a can of SPAM inside of 2 min...

[tynothy drake]

One little thing that will skew the cap/non-cap victories are fights where early game you get that 2-4 lights that start a cap right away forcing the "defender" to break battleline to fight it off. Most of the time I see these it ends in steamroll as the 'attacker" then pushes over them.
I know that there is not good way to track this, but as I saw stated before, "pinch of salt" is added often when statistics are involved.

[Alistair Winter]

tynothy drake, on 26 September 2013 - 08:05 AM, said:

One little thing that will skew the cap/non-cap victories are fights where early game you get that 2-4 lights that start a cap right away forcing the "defender" to break battleline to fight it off. Most of the time I see these it ends in steamroll as the 'attacker" then pushes over them.
I know that there is not good way to track this, but as I saw stated before, "pinch of salt" is added often when statistics are involved.

Well, that's not so much something to be taken with a pinch of salt as a question of what you want the statistics to show. If you want to register every time someone tries to "ghost cap" a bit to draw the enemy back to base, so the rest of the team can attack them as they retreat or split up, then what about registering every time one team forces the other to reposition merely by getting close to their base, without actually being able to cap? Or what about all the times one team sends a whole lance to chase a single Spider way out in left field, a kilometer away from base, just to be sure the Spider doesn't cap?

[Joseph Mallan]

PEEFsmash, on 26 September 2013 - 07:25 AM, said:

Depends on how many people you want to put as "the best." I thought about a better modifier there, but I couldn't think of one. In this case, I am saying that "the best" is "top 20%"....which is a very generous best =]

... Wouldn't the top 20% have an above 80% win rate? To me, IIRC 90% was an A and in Collage 95%... So the top 3-5% would be my expectation of "The Best". Maybe I have higher standards.

stjobe, on 26 September 2013 - 07:32 AM, said:

Or by not having only four light chassis available.

There's 11 heavy and assault chassis (6 heavy, 5 assault), and 10 light/medium chassis (4 light, 6 medium).

If every weight class was equally popular, the percentages would look like this:

• Assaults 23.81%
  
• Heavies 28.57%
  
• Assaults+Heavies 52.38%
  
• Mediums 28.57%
  
• Lights 19.05%
  
• Mediums+Lights 47.62%

From this we can confirm what we all knew; Assaults and especially Heavies are over-represented, and Lights and especially Mediums are under-represented to the same degree (even though lights are the least played weight-class, they also have the least number of chassis available).

Now look at the factions represented. Steiners run more assaults and Davions more heavies than other classes.

[stjobe]

Joseph Mallan, on 26 September 2013 - 08:16 AM, said:

Now look at the factions represented. Steiners run more assaults and Davions more heavies than other classes.

In BattleTech, yes. In MWO though? Not quite so sure a fictional adherence to 'mech weight translates directly to an international audience of real-time FPS players.

[Joseph Mallan]

Guess we'll have to see if an international T game can translate to the International PS crowd then. As Joe I have 2 Assault a Heavy and a Light (Sara !!!)

My Davion has a Heavy and a Assault. I will have another heavy soon.

My Kuritan has 2 Mediums(Terb-K and a Cent), a Heavy (Catapult-K2), and Heavy Metal

My Clanner(soon to be) has Mediums and Lights... I am not good with Lights! Even Ravens!

[Shadey99]

This weekend I'll keep some stats... Faction... Weight Class... Mech... Variant... Over multiple types of my own mechs...

I've done 100+ matches on a weekend before...