Data Compiling (Need your score Screenshots!)

[Zoraster]

Well, because I'm bored with too much time on my hands, I'm going to ask a favor from everyone: please get a screen capture of the score sheet in the game. Screen captures need to follow the following rules:

1. Please only take the screen cap of the scores BEFORE the round ends. I do NOT want end-game scores.

2. Take the screen capture with between 1 and 10 tickets left for one side.

3. Please write a short summary of when (time) and where (map) the screen capture was taken.

4. Try to make sure you're not reposting the score of a game that's already taken place.

5. Please don't JUST take screen captures of you doing well. Also, please only send screen caps that you take from this point forward (no old screen caps).

The purpose of this is to compare stats and tickets. Specifically, I'm going to try and figure out how many kills are "wasted" kills (kills that don't translate into a ticket reduction), how many tickets might be gone due to ticket bleed, how many more kills a winning side has on average than a losing one (and if there are many instances where bleed overcompensates for being out killed).

I'd like to get about 40 screen captures from different games before I compile the data, so I'll need all of your help. I hope that's not too much to ask!

[Lorax74]

you really must be bored...

[Guardianx11x]

Good man. Can you SS the score w/o fraps?

[sc1ence]

Time and map are going to be critical variables in the analysis.

why not end of game scores?

why not past games?

The sample is going to be flawed for sure due to people posting screens from games they did good in (social desirability) or some other subject related variable. But this can be overcome by eliminating the poster's score from the analasis.

is this info available in server logs?

[Zoraster]

good questions, sc1ence.

Why not end of game scores? They don't have the aggregate kills on each side, and it doesn't show how many tickets are left. Also, it only shows the top 20 players. Really though, as much as I'm bored, I don't want to have to spend my time adding.

why not past games? Because past games will probably be, if anything, more skewed to blowout games since people will generally take screen shots when they're doing really well. I assume this will happen to some degree anyway, but I think it'll be less if people are trying to get screen shots from each game they play.

This isn't really meant to be a scientific study. There are too many changing variables to make it truly reliable. But it may give us a rough idea of what's going on.

I may take out the poster's score from the mix, but I think i probably won't. I'm interested in aggregate numbers, and i'm interested in comparing it to tickets, so leaving out a data point (an important data point, no less) will put the numbers off rather than put a more reliable control on it.

Time and map will almost certainly be the most important variables, and if I get enough responses, I'll try and work out something with those. But if the response is only a smattering, it'll be much harder to have a sample size large enough to make even my pseudo-statistical conclusions.

[Zoraster]

Roughly 3:30 PM EDT in Cerbere.

[Crux]

Originally Posted by Zoraster

I may take out the poster's score from the mix, but I think i probably won't.

Good call. Just because the screenshot taker did well in a round to prompt him/her taking the screenshot does not invalidate the data at all. Even if he/she didn't do well, someone else would. We're not going to exclude their stats, so...

[Lorax74]

Bear in mind that I have no experience with the BattleRecorder system as of yet...

Can this information be gleamed from the BattleRecorder files?

[Crux]

Originally Posted by Lorax74

Bear in mind that I have no experience with the BattleRecorder system as of yet...

Can this information be gleamed from the BattleRecorder files?

Not to my knowledge.

[Zoraster]

let's take the above graphic. Some conclusions I can make about this game (with one very large assumption*):

There was 138 tickets of bleed on PAC (330 [starting tickets]- 192 [PAC deaths])

There were 42 revives on the PAC side (234 [EU kills] - 192 [PAC deaths])

There were 27 revives on the EU side (141 [PAC kills] - 114 [EU deaths])

*This being that no player leaves while playing. Because of this assumption, the numbers will only ever be a very rough idea of what actually happens in the game.

[Zoraster]

Originally Posted by Guardianx11x

Good man. Can you SS the score w/o fraps?

yes you can, but you'll have to paste to paint or something and then save the file from there. Basically, when you press the print screen button, it just copies the screen to your clipboard. If you do this, save it as a JPEG or GIF or something rather than a bmp. Those things are huge.

[Guardianx11x]

Originally Posted by sc1ence

Time and map are going to be critical variables in the analysis.

why not end of game scores?

why not past games?

The sample is going to be flawed for sure due to people posting screens from games they did good in (social desirability) or some other subject related variable. But this can be overcome by eliminating the poster's score from the analasis.

is this info available in server logs?

I wasn't going to get in to this, but Sc1ence got the ball rolling so...

While the social desirability factor is there, I dont believe eliminating the posters score would just invalidate your data. These games are not won and loss by individuals, but by a team, and because your scope is wider I think you should be alright at this level. But, I do this he was barking up the right tree.

Map obviously plays a huge factor, I'd expect your results to vary greatly with this alone. I mean, Verdun and Carebear....come on.

Time I don't think is a big deal.

The number of players must always be the same for all samples (I can't stress this enough). 20vs20 player matches are not comparable to 10vs10.

Obviouly you shouldn't accept games that end in a lock-out. If you do, they need to be treated as sperate events, maybe you could see if there is a correlation there...like whenever "Guardian is on team A with Sc1ence team B is locked out" or other obvious things etc.

Like Sc1ence, I don't see why you'd want to use pre-end game ticket counts. If you were shooting for a "snapshot" so you could develope a statistic for the losing team...it doesnt help you. Considering you know the number of tickets the team started with determining the number lost, is no problem and you arent tossing out data. Also, using the end-game ticket results will allow you to include games that end in a near tie! The inclusion of these games is VERY important to your analysis.

EDIT: Ok, if you are doing it to save time then thats fine. I definatly understand your feelings on this one. Try to make sure its on that last ticket though, just yesterday I played a game that ended in 1-0. But I admit, they are not common and can be overlooked if need be.

Make sure you you get a mean and St.Dev for the ticket deficit. Just so when you look at the results,so you know without personal bias if you are dealing with an apple or an orange.

Server logs seem like a great idea if it is possible, assuming you can tell if a team is locked out or not.

Sorry...I'll probably rant more on this thred.

[Zoraster]

Map obviously plays a huge factor, I'd expect your results to vary greatly with this alone. I mean, Verdun and Carebear....come on.

Sure, absolutely it's a big deal. As I've said, if I can get enough data points, then I'll compare only apples to apples. If I only get 20 people sending in screen shots, the best I can really hope for is splitting it into Assault maps and Head-to-head maps.

Time I don't think is a big deal.

This is one of the reasons we do studies like this. It may not be a big deal. It may be. This should give us a good idea. The other reason I want people to include the time they post is that it makes it easier for other people to know if their screen cap has already been posted.

The number of players must always be the same for all samples (I can't stress this enough). 20vs20 player matches are not comparable to 10vs10.

See above. It may matter, it may not. The more data I get, the more of these kinds of things I can look at.

Obviouly you shouldn't accept games that end in a lock-out. If you do, they need to be treated as sperate events, maybe you could see if there is a correlation there...like whenever "Guardian is on team A with Sc1ence team B is locked out" or other obvious things etc.

I'm not sure what you mean by "locked out." Anyway, I don't have some sort of super data compiler that I'm going to sit and enter individuals who are playing. If this is a project you wish to embark on, by all means do it.

Like Sc1ence, I don't see why you'd want to use pre-end game ticket counts. If you were shooting for a "snapshot" so you could develope a statistic for the losing team...it doesnt help you. Considering you know the number of tickets the team started with determining the number lost, is no problem and you arent tossing out data. Also, using the end-game ticket results will allow you to include games that end in a near tie! The inclusion of these games is VERY important to your analysis.

First, I'm not using it because I have no desire to spend 20 minutes per screen shot adding up kills, deaths, etc. The in-game score sheet provides these things. Second, it doesn't include all players, only the top 20. Third, although ties will be harder to determine, it's actually not that important in the end. The type of data I'll glean from this will be nearly the same. Just because one team lost a single ticket before the other isn't important in the end.

Make sure you you get a mean and St.Dev for the ticket deficit. Just so when you look at the results,so you know without personal bias if you are dealing with an apple or an orange.

Personal bias? I don't really have any ulterior motive for doing this. Still, standard deviation will be included in any statistical analysis I do.

Server logs seem like a great idea if it is possible, assuming you can tell if a team is locked out or not.

Again, locked out? Server logs might be okay except for a few things: 1) I don't have access to these logs. 2) It won't provide data as easily as the screen captures.

[Guardianx11x]

I'm not sure what you mean by "locked out."

What I mean by "locked out" is that a team can lose after all the flags are taken, and all the players are killed.

First, I'm not using it because I have no desire to spend 20 minutes per screen shot adding up kills, deaths, etc. Second, it doesn't include all players, only the top 20. Third, although ties will be harder to determine, it's actually not that important in the end. The type of data I'll glean from this will be nearly the same. Just because one team lost a single ticket before the other isn't important in the end.

I added an EDIT for the note about the time the score is taken. Second, it is a very poor assumption that the type of data you glean from this will be nearly the same. Say one team revivies more while the other kills more. This type of data allows you to pit them against one another on (or roughly on) the same plain. Re-think this please.

Anyway, I don't have some sort of super data compiler that I'm going to sit and enter individuals who are playing.

I was obviously kidding about the correlation involving individual players, however, it could be done with all the variables you are considering anyways.

Personal bias? I don't really have any ulterior motive for doing this. Still, standard deviation will be included in any statistical analysis I do.

Ulterior motives? What do they have to do with personal bias?

1) I don't have access to these logs. 2) It won't provide data as easily as the screen captures.

1) Didn't know you couldn't ask 2) Maybe, I've never seen them myself...have you?
I (like Sc1ence) immediaty thought this could be a great way to get a large unbias sample size.

Sure, absolutely it's a big deal. As I've said, if I can get enough data points, then I'll compare only apples to apples. If I only get 20 people sending in screen shots, the best I can really hope for it splitting it into Assault maps and Head-to-head maps.

I think if you compare Verdun with FoB, you are wasting your time.

[Zoraster]

ah yes. these should probably be thrown out or considered separate. good point.