123 replies to this topic

[RG Notch]

Ph30nix, on 05 April 2013 - 11:03 AM, said:

for an accurate statistical sampling you really dont need a large portion of the total population.

for example if i grab a truely random 1000 people off the streets and ask them yes or no to a question the %'s would be withing +/-3% or so as if i had asked the entire planet.

The hilarious part is you can never get a truly random sample so it's entirely meaningless.

[Rigatoni]

Lukoi, on 05 April 2013 - 11:01 AM, said:

So you want to see this out of curiousity but given that you cannot control what types of SS's are actually posted, how can you take it as representative of what's happening in game? It'll still be cherrypicked by people with agendas don't you think?

Instead of a screenshot can I just tell you the map, mechs etc?

I am thinking that it is a SS of the who is alive and dead on each team part of the match? If so it would be hard to cherry pick the data. unless you were in a massive premade. even with that though i would guess that it would not be difficult to pick out the ones that were premade and wich ones were not. the names would all be the same.

and a SS takes more time to change then just writing down the data.

[Vasces Diablo]

Ph30nix, on 05 April 2013 - 11:03 AM, said:

for an accurate statistical sampling you really dont need a large portion of the total population.

for example if i grab a truely random 1000 people off the streets and ask them yes or no to a question the %'s would be withing +/-3% or so as if i had asked the entire planet.

I'm not saying the size alone would invalidate things. The issue is that that the sample size combined with the fact that it won't be truely random could lead to some seriously skewed results. A single persons with an axe to grind would blow up a small data set.

That's all fine if its just for fun. I'll read the results and enjoy them (bit of a numbers nerd myself). The issue is that people will try and use them as proof, while it means nothing in comparison to the massive, true data set already being compiled on the back end.

[hammerreborn]

Vasces Diablo, on 05 April 2013 - 11:12 AM, said:

I'm not saying the size alone would invalidate things. The issue is that that the sample size combined with the fact that it won't be truely random could lead to some seriously skewed results. A single persons with an axe to grind would blow up a small data set.

That's all fine if its just for fun. I'll read the results and enjoy them (bit of a numbers nerd myself). The issue is that people will try and use them as proof, while it means nothing in comparison to the massive, true data set already being compiled on the back end.

We'll have to see, won't we. Lol. I'm hoping I get a lot of bites, and sticking to 10-20 results per person (though maybe I should just limit it to 10) would mean one guy with an axe to grind wouldn't have the ability to negate 10-15 other peoples results with sheer numbers.

And I don't think we'll ever get a true data set from PGI so maybe we can get a small glimpse from this.

[RG Notch]

hammerreborn, on 05 April 2013 - 11:07 AM, said:

Honor system? It's also pretty obvious if someone is trying to skew the results. If 10 people post and they have ravens at 5% overall (my last data had them there), and all of a sudden one guy has 4-6 ravens per match, clearly that dude is cherry picking and I can relatively ignore it. Not to mention the more people submit (and submit honestly), the less effect the cherry pickers have as a whole.

It'll never be as good as the data I gather on my own, but at least it'll have the ability to list thousands of mechs rather than a couple hundred.

So what if you are on the other side and simply don't post any match with multiple Ravens? Are you going to throw out every match that doesn't have 5% raven?
Whatever, I get what you are aiming for I just find it pointless as people can skew the data and thus all you have is an invalid reading of invalid data. The honor system is a terrible thing to use on the anonymous internet. GIGO is a basic truism.

[Jace Lancer]

armyof1, on 05 April 2013 - 10:36 AM, said:

I like your energy hammerreborn, but you know it's pointless right? PGI already have all the data from all matches played, and I'm pretty sure they know how often ECM tips the balance to those that have several against those that have none. But frankly they just don't care, someone high up have the hots for ECM as it is currently and won't change that unless it's over his dead body or something. I'm just trying to save you some time man, use it to play more games or whatever.

Pgi might, but they don't share it with us do they? Besides, this should be a fun little experiment. If I get enough playtime in I'll post some low elo screenshots this weekend.

Best way to screenshot.. Alt enter then print screen?

[Captain Stiffy]

This is so ********. NO DON'T MOD ME YET I'm going to explain that.

PGI _HAS_THIS_DATA_ and I can't stress that enough.

Why don't they show your ELO score to you? Why? Think about that for a minute. What possible reason could they have to hide your own rating from you?

Especially since they show you your exact stats in the forum!

THE ONLY REASON THIS CAN BE IS BECAUSE IT IS BROKEN AND THEY KNOW IT!

You don't need random data collection to prove this (because you can do it with simple logic) - and by the way, to be even remotely statistically accurate, you would need at least one to two hundred results from every single player in the game to create a real trend from which you could model the rise and fall of ELO based on who is grouped with who.

Edited by Captain Stiffy, 05 April 2013 - 11:22 AM.

[hammerreborn]

Jace Lancer, on 05 April 2013 - 11:18 AM, said:

Pgi might, but they don't share it with us do they? Besides, this should be a fun little experiment. If I get enough playtime in I'll post some low elo screenshots this weekend.

Best way to screenshot.. Alt enter then print screen?

Play the game in windowed or fullscreen windowed (fullscreen will give you a black screen) and hit print screen at the EoR match then paste into paint or whatever program you prefer to use.

Captain Stiffy, on 05 April 2013 - 11:20 AM, said:

This is so ********. NO DON'T MOD ME YET I'm going to explain that.

PGI _HAS_THIS_DATA_ and I can't stress that enough.

Why don't they show your ELO score to you? Why? Think about that for a minute. What possible reason could they have to hide your own rating from you?

Especially since they show you your exact stats in the forum!

THE ONLY REASON THIS CAN BE IS BECAUSE IT IS BROKEN AND THEY KNOW IT!

You don't need random data collection to prove this (because you can do it with simple logic) - and by the way, to be even remotely statistically accurate, you would need at least one to two hundred results from every single player in the game.

Dude, calm the hell down. This has nothign to do with elo....

[th3 fr4gil3]

or use fraps to screen cap, can play in fullscreen then.

[Ozric]

Any stats this weekend will be skewed by an increase in hero/founder mechs due to the cbills bonus.

[hammerreborn]

Ozric, on 05 April 2013 - 11:23 AM, said:

Any stats this weekend will be skewed by an increase in hero/founder mechs due to the cbills bonus.

Well aware, though it'll be interesting to see which founders/hero mechs are used the most.

[RG Notch]

Captain Stiffy, on 05 April 2013 - 11:20 AM, said:

This is so ********. NO DON'T MOD ME YET I'm going to explain that.

PGI _HAS_THIS_DATA_ and I can't stress that enough.

Why don't they show your ELO score to you? Why? Think about that for a minute. What possible reason could they have to hide your own rating from you?

Especially since they show you your exact stats in the forum!

THE ONLY REASON THIS CAN BE IS BECAUSE IT IS BROKEN AND THEY KNOW IT!

You don't need random data collection to prove this (because you can do it with simple logic) - and by the way, to be even remotely statistically accurate, you would need at least one to two hundred results from every single player in the game to create a real trend from which you could model the rise and fall of ELO based on who is grouped with who.

Or the more logical reason: They don't want people trying to manipulate it and game the system. But feel free to go all tin foilly. BTW do you wear yours tri corn style or dunce cap style and which protects better from the mind control waves?

[maXe72]

Rigatoni, on 05 April 2013 - 11:01 AM, said:

how do you take screen shots of EoR and post them to the forums? .

You can use the free Trial version of FRAPS to do screenshots even in fullscreen.
Just specify a free F-Key in its setup and press it after every match, the name of the screenshot is changed automaticly.

[Rigatoni]

hammerreborn, on 05 April 2013 - 11:07 AM, said:

Honor system? It's also pretty obvious if someone is trying to skew the results. If 10 people post and they have ravens at 5% overall (my last data had them there), and all of a sudden one guy has 4-6 ravens per match, clearly that dude is cherry picking and I can relatively ignore it. Not to mention the more people submit (and submit honestly), the less effect the cherry pickers have as a whole.

It'll never be as good as the data I gather on my own, but at least it'll have the ability to list thousands of mechs rather than a couple hundred.



First you have to play in either full windowed or windowed mode, full screen wont work.

At the EoR screen, hit print screen.

Open Paint

Paste

Save image

Upload Image to preferred upload site

Post in here using ['spoiler']['/spoiler'] tags (without the 's) surrounding the image links (on the toolbar above the posting window there is a button to link to an image.

like so?

Spoiler

[hammerreborn]

Rigatoni, on 05 April 2013 - 12:14 PM, said:

like so?

Spoiler

Perfect. And good job getting Bryan in there =P

[Rigatoni]

HA! i didnt even notice him!

[Rigatoni]

4 more

002

Spoiler

003

Spoiler

004

Spoiler

005

Spoiler

[ElLocoMarko]

Rigatoni, on 05 April 2013 - 12:14 PM, said:

like so?

Spoiler

Wow, Paul didn't do too well. Right, no name. No shame.
I launched PUG with Thomas Dziegielewski last night. We both had catapults on Alpine (mounting LRM) and got demolished.
"Like shooting marshmallows" he said!

So what is a good image posting site. No signup would be perfect.

[ElLocoMarko]

Wait, so you go through every screenshot noting each mech variant, who wins, presence of ECM... while omitting the submitter (else they get represented at a 10x rate)?

I totally made a program to do that.
It shows you a screenshot and gives you a shorthand to log the results.
You never have to take your hands off the keyboard.
The arrow keys move you forward and back through your screenshot folder and it auto-zooms to the same area you were looking at on the previous image.

It even saves results per-screenshot so if you load the same file more than once it will highlight the mechs that are logged for that file. (Requires unique filenames)
The original plan was to do optical character recognition but that didn't pan out so it turned into a data entry assistant.

I shelved it when I realized I had way more fun making it than using it.

It is really close to what you are doing. It might work right out of the box. It logs variants so ECM is covered.
If you do .NET (C# under VS2010), I'll just hand it over so you can modify as you wish.
Unfortunately it is on my laptop at home and due to stuff I'll be be offline till Sunday.

If I do publish the source I guarantee I will not be supporting it. I have a job and I use my spare time to do things like play MWO pew pew BANG fwooooosh!

Regardless, I'll try to get you 10 screenshots on Sunday.

Edited by ElLocoMarko, 05 April 2013 - 01:48 PM.

[hammerreborn]

ElLocoMarko, on 05 April 2013 - 01:45 PM, said:

Wait, so you go through every screenshot noting each mech variant, who wins, presence of ECM... while omitting the submitter (else they get represented at a 10x rate)?

I totally made a program to do that.
It shows you a screenshot and gives you a shorthand to log the results.
You never have to take your hands off the keyboard.
The arrow keys move you forward and back through your screenshot folder and it auto-zooms to the same area you were looking at on the previous image.

It even saves results per-screenshot so if you load the same file more than once it will highlight the mechs that are logged for that file. (Requires unique filenames)
The original plan was to do optical character recognition but that didn't pan out so it turned into a data entry assistant.

I shelved it when I realized I had way more fun making it than using it.

It is really close to what you are doing. It might work right out of the box. It logs variants so ECM is covered.
If you do .NET (C# under VS2010), I'll just hand it over so you can modify as you wish.
Unfortunately it is on my laptop at home and due to stuff I'll be be offline till Sunday.

If I do publish the source I guarantee I will not be supporting it. I have a job and I use my spare time to do things like play MWO pew pew BANG fwooooosh!

Regardless, I'll try to get you 10 screenshots on Sunday.

I appreciate the offer but I prefer my method. It may be a bit slower but I have it arranged so other than a few manual data entries (aside from the mechs of course) it pretty much does all the crunching on it's own

ElLocoMarko, on 05 April 2013 - 01:42 PM, said:

Wow, Paul didn't do too well. Right, no name. No shame.
I launched PUG with Thomas Dziegielewski last night. We both had catapults on Alpine (mounting LRM) and got demolished.
"Like shooting marshmallows" he said!

So what is a good image posting site. No signup would be perfect.

And that's Bryan =P