Was bored, so a look at Cam statistics?

[harry]

The front page looks very good, I like the 3 boxes layout, and 'Today's most searched' -- the 'stalker count'.

In the drop down list, it would look neater if you use Title case to capitalise them all (or all lower case), so there's some uniformity. That will leave just a few multi-word countries that would still need individual correction -- I'm not sure how U.K or U.S.A (sic) would come through without a full stop at the end.

Meanderings on graphs: I can see the countries and continents are still problematic and imagine the trouble scaling must cause. Perhaps set a minimum number of models to count for a country to be graphed in summaries? Or filter out models based on the scale (e.g. leave those <1000 out of a full range, only plot them on scales up to 2000), so there'd be 2 scales 0-maximum, and 0-2000) -- but it's difficult to know if that would be workable without seeing the data.

Scatter charts can be very good for plotting large numbers of data points, and overcoming scale problems (x, y, z = x-axis, y-axis, bubble size). If a user can select what x, y & z are, and select the models/countries it would become quite interactive.

At the individual level, the models ranked + or - x positions above and below a particular individual (global or in country) are probably the most relevant ones. Who's going up or down just above or below them? That might make it worthwhile for more models to add the names of those around them, improving the sampling, which is still an issue.

Spark lines beside model names in key lists might be good because they ignore absolute scale. They're conceived to fit inline with text, the scale is the range between Max and Min, and change over time plotted within that range). They could help break up the text with graphics on some of the key pages, though they might clutter up the longer lists.

 

[Zoomer]

The front page looks very good, I like the 3 boxes layout, and 'Today's most searched' -- the 'stalker count'.

In the drop down list, it would look neater if you use Title case to capitalise them all (or all lower case), so there's some uniformity. That will leave just a few multi-word countries that would still need individual correction -- I'm not sure how U.K or U.S.A (sic) would come through without a full stop at the end.

Meanderings on graphs: I can see the countries and continents are still problematic and imagine the trouble scaling must cause. Perhaps set a minimum number of models to count for a country to be graphed in summaries? Or filter out models based on the scale (e.g. leave those <1000 out of a full range, only plot them on scales up to 2000), so there'd be 2 scales 0-maximum, and 0-2000) -- but it's difficult to know if that would be workable without seeing the data.

Scatter charts can be very good for plotting large numbers of data points, and overcoming scale problems (x, y, z = x-axis, y-axis, bubble size). If a user can select what x, y & z are, and select the models/countries it would become quite interactive.

At the individual level, the models ranked + or - x positions above and below a particular individual (global or in country) are probably the most relevant ones. Who's going up or down just above or below them? That might make it worthwhile for more models to add the names of those around them, improving the sampling, which is still an issue.

Spark lines beside model names in key lists might be good because they ignore absolute scale. They're conceived to fit inline with text, the scale is the range between Max and Min, and change over time plotted within that range). They could help break up the text with graphics on some of the key pages, though they might clutter up the longer lists.

Cheers Harry

I just ran a simple query to Capitalise the first letter of the Country name - so there should be a far more standardised appearance. For models names I'm using ucwords (php function, turns the first letter in each word to uppercase) - which I meant to use on the countries - but in all honesty, it's simpler just to change them in the DB instead.
The reason for lower case was crawling the pages. I should have made it ignore the case to be honest, but I didn't. May alter that later (especially since queries are not case sensitive, so it's all to do with the java crawling - and it will make no noticeable difference to the crawl).

You are welcome to see the exact data if you wish - I can cobble together some php scripts to output the tables (or some rows from them as models = 7000 entries, and history = 270,000 entries).
"Present" info about a country is no issue, however, the "history" about countries is more difficult. I do have a query prepared that works out the average at the end of each day, and shoves it into a new table (country_id, date, average), I just haven't run it for a while (it's trivial, takes seconds. I'll start running it again ). This would be used to display how a countries average varies over time (the original intent of it), and also then how a continents average varies over time.

 

[AspenRae]

This is amazing! I can't believe I just found out about this site! Wow!!

 

[Zoomer]

Small update, less of interest to the models most likely - but just for keeping people "in the loop".

Created a feedback page. Got a bit carried away, incorporating posting of a few emotes and also automatically linking to a models "page" on the site if her name is correctly entered between square brackets (such as [ambercutie] would create a link to her page...even if she changes her name).
There's a 2 minute flood delay (based upon IP) to prevent spam.

The idea is simply to allow comments/feedback without revealing email addresses. I also removed a model's history as she requested it (she's left MFC and moving on in life, doesn't want any form of linkage behind) and therefore built in a means so that if a model requests herself to be removed via the contact form (and verification via MFC to be sure it's her), even if she is on MFC daily and with a high camscore, she will not be re-added.

More geek info - I have now multi-threaded the crawler, improving the crawl rate (as the number of models nearly tripled since the start, updates were taking 3-4 hours). To minimise page requests I've left crawling pages as one thread, and updating the database using a second thread. However, there is the potential to have as many threads as desired doing each job - meaning the site could be crawled far quicker. Having 2 threads crawling resulted in a 80% increase in speed, whilst the single update thread could keep up. I won't be using more than 1 thread for each simply to avoid sending too many requests.
As the DB grows I may have to resort to trying methods pre-mentioned by FifthElephant to get details much simpler and quicker than at present (although that may result in a 2 per day update). We'll see

This is amazing! I can't believe I just found out about this site! Wow!!

Thank you

 

[MyCamgirl]

You are doing well :thumbleft:

The Jave Applet for the model status fails to run here. If you want I can PM you the console verbose.

As the DB grows I may have to resort to trying methods pre-mentioned by FifthElephant to get details much simpler and quicker than at present (although that may result in a 2 per day update). We'll see

If it is what I think, then I already implemented it. Models status, camscores and nickname changes are checked every minute. Of course, this gets info only on models that are online. So, once a day it crawls all the girls that didn't come online in the last 7 days.

I'm still testing locally to be sure the code is working right and not inserting wrong data. But when I put it online, I wouldn't mind to provide you some kind of API to gather camscores. The DB has now exactly 21,410 models and retrieving the current camscore of them all takes only 1 second and about 400KB. If you only want to update your DB once a day, this wouldn't hurt my server. Unless, of course, you wanna keep playing and developing your crawler. I know how fun it is :-D

 

[Zoomer]

You are doing well :thumbleft:

The Jave Applet for the model status fails to run here. If you want I can PM you the console verbose.

As the DB grows I may have to resort to trying methods pre-mentioned by FifthElephant to get details much simpler and quicker than at present (although that may result in a 2 per day update). We'll see

If it is what I think, then I already implemented it. Models status, camscores and nickname changes are checked every minute. Of course, this gets info only on models that are online. So, once a day it crawls all the girls that didn't come online in the last 7 days.

I'm still testing locally to be sure the code is working right and not inserting wrong data. But when I put it online, I wouldn't mind to provide you some kind of API to gather camscores. The DB has now exactly 21,410 models and retrieving the current camscore of them all takes only 1 second and about 400KB. If you only want to update your DB once a day, this wouldn't hurt my server. Unless, of course, you wanna keep playing and developing your crawler. I know how fun it is :-D

Yes please - would love to The Applet output as well...
btw - MFC never replied to an email, so I didn't ask again - working on the principle that they got, and read, a very lengthy explanatory email and did not respond.
I took that to mean that they're turning an unofficial blind eye to it at present - they know and said nothing. If I kept pressing they may reveal that they could never be "okay" with it, and therefore forcing them to acknowledge would result in being told to quit it :whistle:

Ergo checking multiple times per day using the method Fifth Elephant exampled would be far better!

 

[MyCamgirl]

Zoomer, forget about the Applet error. It was a bad Java installation on my side. It works fine now.

Give me a couple of days to run some tests, import the current data to production server and create a basic API. I will compare my data with yours to see if it has some discrepancy, how much it is and why. By the way, I'm talking about camscores.

 

[Zoomer]

Zoomer - I had no idea you were responsible for that site until I read this thread. Funny enough, someone led me to that site who doesn't frequent the forum, so it looks like its getting out there pretty quickly! Awesome job!

Thank you CammiStar

In the past month it's featured in the results of 450,000 Google searches for various model names at MFC, with an average search result position of about 8 (so first page for a lot of searches) and a click through ratio of around 7% average (approaching 30,000 clicks this month) - which isn't bad considering the only link I made to the site is from here. All from just one little link

Looking outside of Google search to include the overall picture - the site is getting 1500 visitors per day viewing around 4000 pages - although I imagine perhaps 200-300 people account for the majority of page views - the diversity in what they look at is great!

Now I've got to think about what to do next. The country stats are the next graphical feature, but other site features are possible too Make it more useful - for visitors and models alike (there is potential ).

 

[Zoomer]

Hmm, I need a different past-time

Slight update. Now includes model's ratings (from MFC, including vote count). However, new page shows the adjusted rating based on a simple linear interpolation. I'll explain in a sec
Also included a "recommend" feature - there is no side bar menu to show "most recommended" - let a few recommendations build up first! It uses a fusion of cookies and IP address to prevent repeated submissions - but you will find yourself able to repeat-recommend
Currently if your IP is logged (last recommender's IP logged, not everyone's) then it should display (if you cleared your cookies) the button for pressing. If you do - it should show a red box saying you've voted.
However, bug I cannot track down. IE9 chokes on the attempting jquery function "show()" for an element to display this, so IE users won't see that (jquery checks if you've got IE, and avoids this function). Since I'm relatively new to jQuery, please say if you get odd behaviour etc. Chrome and Firefox appear to work fine.

Now, as for the amended ratings. Simply ordering ratings by highest means you can see models with rating of 5 exactly. They've had 50 votes, 80 votes, 1 vote... How can it possibly be fair for them to be "top" compared to someone with 10000 votes and a 4.90 average? It isn't. Traditionally could use a Bayesian method to calculate the average, but I couldn't be arsed as I don't know enough about it...


The average rating of models is 4.73, and AspenRae has the highest number of votes (15000 or so).

So how to make it fair - for everyone? The more votes you have, the more "confident" we can be about your rating. Taking Aspen's highest vote count it's a simple calculation I'm not using "proper" notation because I don't know how good everyone's Math is - you may all laugh at my explanation, or correct it - but i'd rather that than lose people because I stuck a load of symbols in

Max = maximum vote number (15747 or something, from Aspen - but for simplicity I'll probably use 15000 from here down).
Avg = average model rating (simply the mean, so sum up all models who have more than 0 votes for rating - to miss out unrated ones skewing the result).
Ri = the rating of model i.
Vi = votes for model i.
RiAdj = the adjusted rating of model i.
the i is the index, the "ith" element of a collection (the collection being all the models!)...where ith element(member) may be Ambercutie, or Crazysysy etc. When i is Amber, Ri = Ambers rating, Vi = the number of votes Amber got. So, anyway, that's what i represents - a current model of the collection of all models. So, onwards

RiAdj = (Vi/Max) * Ri + (1 - (Vi/Max)) * Avg

It's usually written 1/Max * Vi, but more concise this way. Think of the two extremes - when Vi = 0 (as will Ri if you like), and then Aspen as our reference point (assuming her huge number of votes gives an honest evaluation!). So when Vi = 0, and Ri = 0 - the left side is 0. Added to this is 1-0 (so 1) * Average. The average is 4.73 as I said.
So someone with no votes/ratings is given an average rating of 4.73.

Now the other extreme - Aspen.
Her Vi = Max, so Max/Max * Ri means the left is 1 * her rating - or just her rating The right side is 1-1 * Avg, which is 0. So Aspen keeps her rating - about 4.92 I think.

Now take a low rating high votes example... you'd assume that a shit load of votes at rating 2 means the model is preeeetty bad. How does this work out?
Well, take someone with 10000 votes and a rating of 2.
10000/15000 = 0.666*2 = 1.3333 for the left side. Add to this the right - which is 1 - (10000/15000) * 4.73 = 0.3333 * 4.73 = 1.57
They get a rating of 2.87 (unless my math is wrong somewhere).
You may think "hang on, but that's higher than what they had!" - yes, it is But if they had had MORE low ratings, say 15000, then you'd be pretty much equally confident that they really are as bad as their votes say. Right now she has 2/3rds of the votes Aspen says, so you can say the "confidence" isn't great

By now people probably wonder why bother - easy, example results Format will be shite...sorry, but essentially this is what it looks like when just ordering girls by rating
Format is:
NAME --- VOTES --- RATING

angel_bites 	54 	5
hottyjoy 	28 	5
comeonkaitlyn 	42 	5
sexygurrl1 	39 	5
keisha4you 	18 	5
dietagirl 	57 	5
bailey_21 	59 	5
angelfaith 	10 	5
sexybunny70 	14 	5
angelbabe9 	35 	5
naomirussell 	3 	5
godessnadia 	31 	5
sexy_dolll 	47 	5
carlynine 	58 	5
babydoll06 	29 	5
xsosexysohotx 	23 	5
sugarbabedoll 	17 	5
charmingkatia 	34 	5
daddysdiamond 	28 	5
saphirababe 	116 	5

Now to order the same amount by their adjusted camscore...

MODEL --- VOTES --- ADJUSTED RATING

aspenrae 	15431 	4.9071
ambercutie 	14548 	4.8719
lovelykittie 	14120 	4.8617
justnutts 	11182 	4.8513
ruby_renegade 	9798 	4.850
joscelynrae 	10746 	4.8474
violet420 	10630 	4.8426
tata_latina 	10345 	4.8386
lovely_cath 	12563 	4.8312
hollyhanna 	11172 	4.8196
naughtysarah 	6666 	4.8188
ginnypotter 	7732 	4.8174
goddesszoey 	7867 	4.8159
xnatashapinkx 	8264 	4.8122
jenny_squirt 	8806 	4.8091
aleahjasmine 	7748 	4.8057
angelasommers 	6552 	4.8056
ayoungdolly 	6969 	4.8049
kittywilde 	6853 	4.8037
angel_nicolle 	5661 	4.7985

Much better I think :handgestures-salute: :thumbleft:

That's all

p.s for fullness, here is the same table as above but this time it is this
MODEL --- VOTES --- RATING --- ADJUSTED RATING

aspenrae 	15431 	4.9071 	4.9071
ambercutie 	14548 	4.8811 	4.8719
lovelykittie 	14120 	4.8748 	4.8617
justnutts 	11182 	4.9008 	4.8513
ruby_renegade 	9798 	4.924 	4.8500
joscelynrae 	10746 	4.9025 	4.8474
violet420 	10630 	4.8975 	4.8426
tata_latina 	10345 	4.8963 	4.8386
lovely_cath 	12563 	4.8564 	4.8312
hollyhanna 	11172 	4.8571 	4.8196
naughtysarah 	6666 	4.9472 	4.8188
ginnypotter 	7732 	4.9132 	4.8174
goddesszoey 	7867 	4.907 	4.8159
xnatashapinkx 	8264 	4.8911 	4.8122
jenny_squirt 	8806 	4.8753 	4.8091
aleahjasmine 	7748 	4.8895 	4.8057
angelasommers 	6552 	4.9199 	4.8056
ayoungdolly 	6969 	4.9066 	4.8049
kittywilde 	6853 	4.907 	4.8037
angel_nicolle 	5661 	4.9318 	4.7985

p.s. I promise that soon I will produce the country related statistics

 

[harry]

Cheers Harry

I just ran a simple query to Capitalise the first letter of the Country name - so there should be a far more standardised appearance. ...
The reason for lower case was crawling the pages. ...

You're welcome, it looks a lot nicer now.

You are welcome to see the exact data if you wish - I can cobble together some php scripts to output the tables (or some rows from them as models = 7000 entries, and history = 270,000 entries).
"Present" info about a country is no issue, however, the "history" about countries is more difficult. I do have a query prepared that works out the average at the end of each day, and shoves it into a new table (country_id, date, average), I just haven't run it for a while (it's trivial, takes seconds. I'll start running it again ). This would be used to display how a countries average varies over time (the original intent of it), and also then how a continents average varies over time.

I'd love to see it, you've seen how I've looked at what might be derived from the data on display, so the background tables would be invaluable to be able to lookup values.

I'm most interested in the aggregate data (countries, regions) and how it might be presented visually. As we discussed above, the issue is really about bias introduced by sampling the model population from the Offline Model list (only catching those with camscore >1000).

I thought I might have a partial solution by gathering names from the Model list, removing the bias of the Offline model list. That could be used to add Region as variable to the database -- which would overcome the inaccuracies of Country, providing some validation for it.

From what's said above, MyCamGirl may well have solved sampling bias issue by gathering the entire population anyway! So if that could be integrated into your data gathering process the database should be valid for derived statistics on All Models. That would be a huge advance.

However, quick summary of the structures - and open to any suggestions

Models table:
model_id (primary key), model_name, last_login, current_camscore, camscore_change (during an update it is current_camscore - new current_camscore value), last_update (last time the profile was visited - indicates renames that aren't found or models left), country_id, views (relates to how many times her profile has been viewed).

Camscore History table
Model_id, Camscore, Date, entry_id (primary key).

Countries table
country_id (pk), name, short name, flag, small_flag, continent_id

Continent table
continent_id (pk), name
Self explanatory - continent_id matches to the name of the continent. Didn't go with Middle_East or Central_America. They are regions, not continents. New country additions have the value "Unknown" for continent, as does the "Unknown" country.

Submissions Table
Url (pk)

Country history
country_id, avg_camscore, date, id (pk)

most_viewed (searched)
model_id (pk), count

Monthly - models camscore at the start, and end, of the month.
id (pk), model_id, start, end, date.
...

The Continent table is the one I might question, whether it might be better replaced with MFC region (or adding a Region table, if you or MyCamGirl can capture data from the online Model List) allowing a mapping to continents for NA, C/SA, Asia, Africa, and Australia, ... then break Europe/Russia into separate continents (it's by far the largest region):

Europe/Russia 56.3%
NorthAmerica 18.8%
Central/SouthAmerica 17.5%
Asia 4.8%
UnitedKingdom 2.0%
Africa 0.3%
Australia 0.3%

And the UK could even be introduced into the continent of Europe :shhh:

 

[MyCamgirl]

IE9 chokes on the attempting jquery function "show()"

Strange things can happen on IE. Try to debug using Firebug lite for IE. Usually it shows JS warnings that IE doesn't catch.

I think the AJAX requests, which is where you have the show(), are not traceable with the lite version. But if the show() fail, it will tell you why. You can also try to find what issues IE has with the jQuery version you use.

Traditionally could use a Bayesian method to calculate the average, but I couldn't be arsed as I don't know enough about it...

In case you wanna give a try...

( (avg_num_votes * avg_rating) + (this_num_votes * this_rating) ) / (avg_num_votes + this_num_votes)

A PHP script simulating votes is available here: http://codepad.org/gSdrUhZ2

I was comparing data and stumbled on a strange situation. There is a top model (I won't name here) who has her MFC profile restricted. She does not appear on my database and on yours the last update over 15 days ago.

If she appears on your DB, then your crawler was able to access her profile at some point. But for some reason it can't anymore. But my case is even stranger. If I don't have her on my DB, it means she never came online since I started collecting data (about 2 months ago), which is impossible because according to Google cache she was online recently and I've also seen her name listed on MFC top models in the past 2 weeks. I don't believe this is a case of inflated camscore.

As this girl seems to want to keep the low profile, despite being a top model, this bug is actually good in this case. But this means that if you use my data, you might miss some camscores. By the way, I don't intend to manually add girls. I will just let the script to automatically insert any girl who comes online. The model's data might be restricted in some countries/regions if the model ask.

 

[SweetMystery]

Maybe for a restricted profile, they've blocked access to crawlers/bots/caching/etc? It wouldn't make sense if you have a restricted profile which could still be access through caching... Maybe this was something done recently, too?

 

[Sevrin]

How will you handle the fibs in the "country" field?

Many models who appear when I select to view exclusively Europeans show up with countries such as the US or Antarctica in their profiles. There's also a fair bit of intraeuropean fibbing with models in southern Russia showing up in Poland (for reasons which escape me), for instance. I'm glad that MFC's regions work based on actual locations, because I don't really care where models are pretending to be located, but it does mess up country statistics on your site quite a bit.

 

[Zoomer]

How will you handle the fibs in the "country" field?

Many models who appear when I select to view exclusively Europeans show up with countries such as the US or Antarctica in their profiles. There's also a fair bit of intraeuropean fibbing with models in southern Russia showing up in Poland (for reasons which escape me), for instance. I'm glad that MFC's regions work based on actual locations, because I don't really care where models are pretending to be located, but it does mess up country statistics on your site quite a bit.

Hmm, you know I'd never even noticed the regions on MFC - been so long since I'd looked at the filter! :lol:

If models are displayed as from one region, whilst claiming to be another on their profile, then I don't think there is much I can do about that.
I assume the region by MFC filter is based on the model's IP - not what the model claims on their profile - and that is something it isn't possible (or certainly shouldn't be possible) to access.

@MyCamGirl - cheers, will look at the extension and hopefully find what's wrong
As for the model, I think she's been banned perhaps? No profile, no way of finding her - someone posted on this forum that she'd been renamed, but no sign of that profile either. If it's a country block/IP block then you see a green page afaik saying you can't view them, not the default MFC profile page for non existent profile names (or their 404 page as it is therefore!).

@Harry, always interested. Refactoring a few of the tables at present - things got untidy as things were added without a huge amount of planning. Most of the stuff I do is "spur of the moment" and not planned, so I don't always give it the best structure on the first iteration

 

[MyCamgirl]

@SweetMystery: Googlebot isn't forbidden from accessing her profile. But Google is a bot apart. I asked a friend in Germany to access her profile and it is restricted there too. Supposing Zoomer's crawler can't access it, then her profile is also restricted in UK. Plus, in Portugal. I don't see any reason to block countries like that. But it is up to her.

But if she blocked my country from accessing the profile and also her room, she will never appear as online models to me. This explains why she isn't on my DB.

@Sevrin: I don't know what Zoomer will do about that, but I'm not collecting this kind of info exactly because it is inaccurate. I still don't know if I will insert manually the correct info on the girls I know for sure where they are from. If they are listed as Congolese or whatever, it is because they don't want to tell where they are.

Regarding those Polish/Russian models... If they are the models I know, they have Poland on their profiles because they were born there, but are living in Russia. They are not really lying. If you ask them, they will tell you where they were born and where they are now. Some girls will even explain why the moved to Russia.

@Zoomer:No, she was not banned. But yes, seems she changes name frequently. You have on your DB the correct name though. I get the green page, so I'm trying to access the correct name and the account is active. Otherwise I would be redirected to main page.

 

[Tesla]

Thank you so much for making this, Zoomer! It's fascinating and really helps put in perspective how *every* model goes through peaks and valleys. I especially love to check it out when my score is going up, tee hee! :dance:

 

[Harvey Birdman]

just wanted to say this is pretty cool Zoomer. Sure brings to light the ups and downs thru the months of any camgirl.

 

[Zoomer]

Update again. Some of these are a few days old, the other is much different

New URL - Camscorestats.com
The existing url will continue to work for another day or two whilst this "news" propogates. Then I will implement a permanent url redirect - so anything going to the old url will be (silently and without you doing anything) redirect to the new one This should therefore allow Search bots to also get to the new url and update accordingly. Any issues, please say!

Other things:
Comments.
Report feature.


Comments - yes, finally got comfortable with using jQuery and Ajax, so there is now the ability to post comments on a models history page. Emotes are acceptable (you type the code, they'll appear when next viewed), and a rudimentary word filter (fuck, cunt, whore, bitch, shit replaced with f***, c**t, wh***, b*tch, s***) to prevent un-necessary vulgarity aimed at individuals.

Also implemented a "Report message" system as well, so if you see anything objectionable - then report it! Grey icon means un-reported (so click to report), red means reported (can't click it anymore) and green means it has been reported but found to be acceptable. Whilst I encourage open, honest and frank comments (either positive or negative), it doesn't mean people are free to be a c*nt or have arguments. It isn't a discussion forum for starters! Secondly, posting anything that isn't directly related to feedback isn't really useful and may be deleted. Definite NO's are posting personal information, attempting to post links to any material regarding a model etc. I encourage the use of the report feature where appropriate, but also urge consideration. I do say feedback, positive or negative, is welcome - so flagging up a "I tipped this girl x tokens and she didn't do what she promised" or "In her private she refused to do what was agreed" are very acceptable. Of course, "fucking bitch whore stole my fucking money" is not. I will try to be consistent, but for example, some removed posts so far have been:

1) The same guy, using the same IP but three different names posting 3 times in one models profile in 7 minutes. As I say in the "rules" (linked from the comment box) I don't like spam - and if you're changing your name to make a comment on a model you've already commented upon, then you're changing your name for a simple reason - because what you're likely doing is spamming!
2) Someone posted what may, or may not, have been the real name of a model along with their conspiracy theory about her being an MFC exec's wife.
3) Someone calling a model a "fucking bitch". If your comment is motivated out of intense dislike/hatred, then it's not welcome!

As usual, suggestions are welcome for other ideas/thoughts. With the new url I have a few other things to implement in the coming week, before finally looking at what country data is available so far. Thanks to MyCamGirl there is the potential to have a vastly increased quantity of model's camscores, but there is a downside - it is just the id and camscore - no other data, so still requires extensive webcrawling to extract the rest (although this could be less frequent!).

 

[AmberCutie]

Something I loved as a feature of our individual pages was seeing where I ranked in terms of active profiles. Today it shows I am rank 41 overall.


I assumed that meant where I would show up on the current Profiles Homepage since that list is the currently active models. But I show up as 19th on there.



What is the difference in how you're pulling the info vs. how MFC displays it on the Profiles Homepage? I used to check the Profiles Homepage often to see where I showed up, but started checking camscorestats instead. Now I realize it wasn't the same info I thought I was getting/looking for.

 

[MyCamgirl]

I assumed that meant where I would show up on the current Profiles Homepage since that list is the currently active models. But I show up as 19th on there.

Yes, it is because he lists all the models, whether they came online today or in the last summer. But it is also because of the models he can't get info anymore. There is at least one model who is ranked better than you on camscorestats.com, but right now you have a better camscore. As he can't access her profile to update the info, she will be listed with that camscore forever.

btw, sorry... I know the question was to Zoomer, but I thought I could answer.

@Zoomer: If you know what girl I'm talking about, you can now use that API to get her current camscore. She is not missing on my DB anymore

 

[Zoomer]

Someone pointed this out to me the other day as well They'd said a girl had ended up 48th on the profiles...she was about 150th for me

The main difference will be timing - the rank on the history page shows everyone who's been logged (just because they've not been online recently doesn't mean their camscore isn't there any more) whether that's last 3 days, 30 days or 300 days...MFC's page is definitely who is online in the last 3 days

Compare it to the highest camscore page with 3 day filter instead
The differences still observable are again due to timing, and also blocked models (Miumiu for example is blocked, whereas at the time of collection DakotaTaylor hadn't logged in for 5 days - but logged in just a short while ago on MFC). Devon will appear on mine, whereas not on MFC - the reason for that is more exact treatment of time. For me - 12.01am on the 2nd until 11.59pm on the 5th is just under 96 hours in reality but treated as 72. On MFC it obviously treats it as 95hrs 58 mins and discount her

I can alter the ranking page to factor in last login time to give a more accurate reflection of how you measure against current - but the "Highest Camscore" page does this. So what you see on the profile page is a "historically, compared to all models known, you are here... "

@MyCamGirl - yep, I've got to implement that yet as I'll still have to visit profiles for names and other info.
There are ways to minimise the issue - with infrequent visits for names, country etc - however, I've not re-written in order to do this yet - and I don't to just overload by banging them in without having the fuller data than camscore and id - so it'll be a few more days (basically I just have to sit down for an hour and go through it).

Another concern on top is my shared hosting... I've no idea at what point they may say "you are using too much resources" Dedicated hosting is too expensive when it comes to something made whilst messing about, but I'm hoping your API reduces the amount of load (due to fewer queries during an update) - but at the same time it may cause a far larger sql table. But I'm just sticking my fingers in my ears regarding that and hoping for the best

 

[AmberCutie]

Ok, cool. Thanks for the clarification!

 

[Zoomer]

Blocked a few IP's today. If you are blocked yet have no clue about what I am talking about next - then please message me. I'll see what I can do to re-allow you to view the site again.

However, if you are someone blocked and were running software directed at the site, then please message me and explain what you are attempting to achieve.

One Adelaide based IP address is visiting 1 page per second, nothing too bad.
One Trend Micro address was making 10+ page requests per second. At least I think it's Trend Micro - running software to detect vulnerabilities? Or just their IP range in a really bad DOS effort?

The cumulation of 10-15 page requests per second is too much to allow to continue. This isn't a professional site, nor does it have dedicated hosting. It is a "hobby" site hosting deal, and such volumes will result in it being pulled... hence why I am blocking IP's.


p.s. No member has to worry about "fast clicking", such as viewing a few profiles in succession. These addresses visited THOUSANDS of pages in minutes...
Whether someone is attempting to replicate the site by scraping the content... :whistle:

 

[inkydoo]

Well, being the dork that didn't notice this entire thread , I didn't realize its creator was on here (of course, I shouldn't be surprised I guess).

In any case, just wanted to say thanks, Zoomer. As the kind of person who's fascinated by these sorts of things (and have written a few programs to measure similar statistical stuff), I know and appreciate how much hard work went into it. Keep up the good work. :handgestures-salute:

 

[Sevrin]

Since there seems to be seasonal variation, I was wondering if it would be possible to generate an aggregate camscore for all models, so that it would be possible to see whether individual score variation is due to seasonal factors. Since camscores do not vary during days when models are not logged in (example), I don't know what would be the best way to handle that. Would it be possible to flag only those days when a model is logged on and ignore other days? I don't know if you keep track of every day logged in or only the latest day.

 

[Zoomer]

I don't quite get what you mean :?

Or at least if I do, I think it's very simple. The average (mean) camscore for any individual day, irrespective of the time of year etc, is 1000.

So it'd just be a line of y = 1000.

 

[Sevrin]

Not average, aggregate, i.e., the total of all camscores. But if the mean score is 1000, then that would not give us much information other than how many models logged in on a given day multiplied by 1000.

 

[Zoomer]

I feel the need to post something regarding comments

Comments... they are supposed to give members (models included) a voice where usually they have none. Comment a critique on a models profile, she'll delete it. This was therefore supposed to offer the opportunity to post your observations, good or bad. Supporter or... well, that's where it becomes tricky.

Whilst 90% of the comments are fine - awesome indeed - this is a gentle reminder to everyone (including a model who had a pop at another model!).

Comments aren't there to let you be a fucking cunt towards someone (I surprise myself with my eloquence!).

I do remove such garbage if I read it, or if people flag it up. It is entirely possible to leave feedback either good or bad without resorting to personal insults and being nasty. Do NOT get personal as you cannot be objective! If you cannot, then your comment will disappear if flagged or if I read it. It is not a vent outlet. It is not a location to have a free-for-all because you hate someone. If that's the motivation to post, then don't bother I do IP ban repeat offenders (although it is easy to get around it).

There is a provision to close comments on a specific model (just as there is a provision to close recommendations and view/search count), and it will only be used in the event there is repeated abuse of the comments section to bash someone. Be objective in your comments, not objectionable - leave personal attacks and insults out of it!

As always - Models are free to request to be removed from the site at any time and for any reason. Such requests will be honoured once you identify yourself for sure (either here through having a confirmed model account here and PM'ing me - or via MFC in a PM/MFC Mail to member account Zoomer123). I will ask why you wish to be removed, simply to keep track of what concerns there may be - however, the response does not affect being removed whatsoever.

I thank all genuine comment makers and those reporting nasty comments. If anyone disagrees with an "okayed" comment, please contact me about it via the form (state the models name the comment is on obviously!) and I'm willing to listen to concerns regarding the comment (also feel free to PM me here). Anyone who flags a comment they made - the comment will be deleted as I assume it means you wish it to be removed (IP's match on commenter/person flagging the comment).

p.s. Freedom of speech is something granted by Governments, it is irrelevant on the website. You are free to comment, but not "Free to be a douchebag". If I decide to remove a comment, tough. Perhaps examine your motivation and content of posting rather than trying to concentrate on fighting a case of "but I should be free to be a prick to whomever I like"... :roll:

 

[ComicOzzie]

In reading some of the comments on your site, it is clear that there are many men out there with a serious hatred of women in general.

 

[Zoomer]

There is. It is quite challenging to try and keep it okay without just deleting negative things and seemingly like a censor of anything that's not "WOOO HOOO MY FAVOURITE".

Worst is, obviously, that there's no way to validate what people are saying. I've let one or two comments slide due to having read certain things here (with proof) about what they've done and a feeling that - yes - the comments are therefore justified. Otherwise I try to respond to put comments into perspective.

Sometimes I just want to close all comments, but comments are a good way of making a model visible when she may otherwise be invisible. Comment on someone with a camscore of 200? She appears on that front page for a while...same for searches etc. Currently wondering about views/recommendations. It's very lopsided (same names for nearly 2 months!), so may just make it more randomised...it's tough.