Statistics Levels statistics - Research for monthly reports

Gepeto · Oct 6, 2016

EDIT @ 10/18/2016 - I have changed the name of this topic as it was first intended to be a simple report of stats for september 2016 from the website itself. Eventually it turns into a real statistic methodology dicussion because of the competences of members arround. That's great as it is now possible to rank levels according to several criteria importances and not only each of them separatly. I will publish some results/threads for the last couple of monthes (for history) before regular and monthly releases.
__________

At the end of August @vince started a thread about sorting levels on a specific date range. I really loved that! (Thanks Vince!). More than one month after I think it could be great to do that each month. Originally I have posted that on the previous thread but it is maybe more relevant and organised in a special thread.

As one post cannot have more than 10000 characters I am going to do that in several parts.

So Here are the stats for september (well... much more now), from the 09/01/2016 to the 10/06/2016 included (MM/DD/YYYY). You can click on the links of the titles for a direct access

____________________
20 most viewed

N° Title Author Views Rate Likes Comments Date
1 Praying For Rain nGord 1K 3 1 58 10/02/2016
2 Deserted Curse meko 650 3 2 23 09/20/2016
3 Opening Doors meko 604 7 4 22 09/12/2016
4 The Mountain meko 590 3 4 21 09/09/2016
5 Limbo richardfu_ 587 10 4 14 09/02/2016
6 Begin Again meko 564 3 2 20 09/05/2016
7 Headroom ZachMan 562 0 1 0 09/08/2016
8 Transformers (re) Gepeto 500 3 1 21 09/07/2016
9 Glyph Walk richardfu_ 496 11 3 19 09/16/2016
10 No way to Win TR O 494 5 3 29 09/12/2016
11 Special Forces Vol I S. Nagy 468 4 1 19 09/04/2016
12 Padlock Game v2 TR O 467 3 0 28 09/18/2016
13 OOOOOOO SCAAARYYYYY!!!! Scare Crow 419 0 0 24 09/21/2016
14 Expedition IV richardfu_ 406 10 1 16 09/12/2016
15 STEPPING UP Mike_CN 387 0 0 24 09/24/2016
16 Castle KPACABA 371 0 1 11 09/11/2016
17 Choice meko 357 7 2 19 10/03/2016
18 Mountain Depth richardfu_ 353 7 5 10 09/10/2016
19 Expedition III richardfu_ 353 9 3 9 09/06/2016
20 Cursed Tower richardfu_ 343 2 0 11 09/06/2016

Gepeto · Oct 6, 2016

20 most commented

N° Title Author Views Rate Likes Comments Date
1 Praying For Rain nGord 1K 3 1 58 10/02/2016
2 No way to Win TR O 494 5 3 29 09/12/2016
3 Padlock Game v2 TR O 467 3 0 28 09/18/2016
4 STEPPING UP Mike_CN 387 0 0 24 09/24/2016
5 OOOOOOO SCAAARYYYYY!!!! Scare Crow 419 0 0 24 09/21/2016
6 Deserted Curse meko 650 3 2 23 09/20/2016
7 Opening Doors meko 604 7 4 22 09/12/2016
8 Swamp Khudrat 223 5 0 21 10/05/2016
9 The Mountain meko 590 3 4 21 09/09/2016
10 Transformers (re) Gepeto 500 3 1 21 09/07/2016
11 Begin Again meko 564 3 2 20 09/05/2016
12 Choice meko 357 7 2 19 10/03/2016
13 Glyph Walk richardfu_ 496 11 3 19 09/16/2016
14 Special Forces Vol I S. Nagy 468 4 1 19 09/04/2016
15 Super Mario Bros V3 sawdust 316 5 2 18 09/30/2016
16 Crouching Tiger TR O 316 8 4 17 09/27/2016
17 Plan B LooKing? 220 4 0 16 09/23/2016
18 Expedition IV richardfu_ 406 10 1 16 09/12/2016
19 Mini Outpost richardfu_ 337 13 1 15 10/03/2016
20 Flipped sawdust 247 2 1 15 09/23/2016

Gepeto · Oct 6, 2016

20 most rated

N° Title Author Views Rate Likes Comments Date
1 Mini Outpost richardfu_ 337 13 1 15 10/03/2016
2 Glyph Walk richardfu_ 496 11 3 19 09/16/2016
3 Path of Courage richardfu_ 328 10 2 13 09/30/2016
4 Expedition IV richardfu_ 406 10 1 16 09/12/2016
5 Limbo richardfu_ 587 10 4 14 09/02/2016
6 Dark Waters richardfu_ 238 9 3 10 10/02/2016
7 Expedition III richardfu_ 353 9 3 9 09/06/2016
8 Crouching Tiger TR O 316 8 4 17 09/27/2016
9 Choice meko 357 7 2 19 10/03/2016
10 Claw Machine Box TR O 274 7 1 12 10/01/2016
11 Chateau Sawdust sawdust 220 7 3 8 09/23/2016
12 Opening Doors meko 604 7 4 22 09/12/2016
13 Mountain Depth richardfu_ 353 7 5 10 09/10/2016
14 Ship Picking richardfu_ 172 6 2 5 09/27/2016
15 Expedition V richardfu_ 273 6 2 7 09/22/2016
16 Fishing richardfu_ 254 6 3 7 09/18/2016
17 Swamp Khudrat 223 5 0 21 10/05/2016
18 Jump pipe Muslim Arizzy 154 5 1 11 10/02/2016
19 Super Mario Bros V3 sawdust 316 5 2 18 09/30/2016
20 Sudoku MScript 161 5 2 10 09/28/2016

Gepeto · Oct 6, 2016

20 most liked

N° Title Author Views Rate Likes Comments Date
1 Mountain Depth richardfu_ 353 7 5 10 09/10/2016
2 Ancient Temple Lazarus 146 1 5 1 09/04/2016
3 Safebreaker2 cheinz 74 4 4 4 10/02/2016
4 Rails MScript 116 3 4 5 09/29/2016
5 6 Piece Burr sawdust 142 4 4 4 09/28/2016
6 Crouching Tiger TR O 316 8 4 17 09/27/2016
7 Opening Doors meko 604 7 4 22 09/12/2016
8 The Mountain meko 590 3 4 21 09/09/2016
9 Limbo richardfu_ 587 10 4 14 09/02/2016
10 Dark Waters richardfu_ 238 9 3 10 10/02/2016
11 Electric Cute LooKing? 137 3 3 6 09/30/2016
12 A lofty goal Jose Luis Galiano 137 2 3 6 09/30/2016
13 Chateau Sawdust sawdust 220 7 3 8 09/23/2016
14 Cae11 Chuckthulhu 192 5 3 9 09/19/2016
15 Fishing richardfu_ 254 6 3 7 09/18/2016
16 Glyph Walk richardfu_ 496 11 3 19 09/16/2016
17 Swivel Muslim Arizzy 193 3 3 13 09/13/2016
18 Manage Your Time Muslim Arizzy 79 2 3 1 09/12/2016
19 No way to Win TR O 494 5 3 29 09/12/2016
20 Screenshot_2016-09-08-23-01-41-47 Muslim Arizzy 172 5 3 4 09/08/2016

cpw · Oct 7, 2016

Nice statistics. @richardfu_ basically owned half of the Most Rated chart

I always find it interesting that the most viewed/commented levels are usually either about astonishing bugs (the wow factor), or have uncertainties (i.e. too hard, or solvable in many ways) that open the door to a lot of discussions (but not necessarily ratings). The most rated ones however usually have a clear direction, are solvable by most people, and often action-packed (because @richardfu_ ).

vince · Oct 7, 2016

cpw said: ↑

Nice statistics. @richardfu_ basically owned half of the Most Rated chart

I always find it interesting that the most viewed/commented levels are usually either about astonishing bugs (the wow factor), or have uncertainties (i.e. too hard, or solvable in many ways) that open the door to a lot of discussions (but not necessarily ratings). The most rated ones however usually have a clear direction, are solvable by most people, and often action-packed (because @richardfu_ ).
Click to expand...

I agree, many levels get a lot of views for bad reasons.

Gepeto · Oct 7, 2016

That's interesting. Here is how I see things:

the most commented = there's something to speak about

the most viewed = there's something interesting to see

the most rated = there's high quality content

the most liked = there's something... to like

To me, "the most viewed" means something relevant where "the most commented" could be tendencious. But I agree that "the most rated levels" is more relevant than anything. On the other hand I've seen a default in the rated method: it doesn't take care of the ratings. So you could have 20 rates for a level with a rating of 3 Fortunately , for real, levels are rated when people prefer to add a positive rating. But the prefered way to consider ratings should be an equation with rates and ratings (if you've got an idea on that tell me )

Everything is based on the orders proposed by the website for each category. It could be better with an access to raw data (base) but it is not possible. Maybe I'll try to figure out the rating more than just the number of rates but that's less obvious to code as it needs to reorder everything with this new result not just from what the website tell us on that.

Otherwise, maybe someone could find an algorythm that I can use to melt all the data and build a chart of the 20 best levels of the month?

vince · Oct 7, 2016

Gepeto said: ↑

That's interesting. Here is how I see things:

the most commented = there's something to speak about

the most viewed = there's something interesting to see

the most rated = there's high quality content

the most liked = there's something... to like

To me, "the most viewed" means something relevant where "the most commented" could be tendencious. But I agree that "the most rated levels" is more relevant than anything. On the other hand I've seen a default in the rated method: it doesn't take care of the ratings. So you could have 20 rates for a level with a rating of 3 Fortunately , for real, levels are rated when people prefer to add a positive rating. But the prefered way to consider ratings should be an equation with rates and ratings (if you've got an idea on that tell me )

Everything is based on the orders proposed by the website for each category. It could be better with an access to raw data (base) but it is not possible. Maybe I'll try to figure out the rating more than just the number of rates but that's less obvious to code as it needs to reorder everything with this new result not just from what the website tell us on that.

Otherwise, maybe someone could find an algorythm that I can use to melt all the data and build a chart of the 20 best levels of the month?
Click to expand...

What I have done is counting the total number of stars, that is to say number of ratings multiplied by rate. It gives a better idea than the average rating. The number of views is not very relevant. By the way, I always wondered how hundreds of people can view a level without rating or writing a small comment. I think at most 10 guys give rates to levels in the forum...

cpw · Oct 7, 2016

vince said: ↑

What I have done is counting the total number of stars, that is to say number of ratings multiplied by rate. It gives a better idea than the average rating. The number of views is not very relevant. By the way, I always wondered how hundreds of people can view a level without rating or writing a small comment. I think at most 10 guys give rates to levels in this forum...
Click to expand...

Many are probably just visitors and don't even bother to register. They are just looking for more content and have no interest in discussing or studying the physics of Mekorama

Gepeto · Oct 7, 2016

@vince From what I have understood you was parsing also the number of rates, not the rating number, isn't it?

There's a difference between them because the "ratings stars" (red frame on the picture above) is not the same than the rating notation (shown by the arrow) . Maybe (rates x ratings) = valuable notation? But then we could have (3 rates x 5 ratings = 15) < (10 rates x 2 ratings = 20) witch is bad...

I don't know how the views are counted as many time I came to a level without downloading and nothing changes in the views count (at least on the moment). The same when I download it. Maybe there's some delay.

Many people just come and download I think. They are not obliged to rate so, that's part of the deal Maybe some don't even speak english. My kid used to came to the website on my back but he doesn't read english (french is coming soon ), so no need for that. What we don't know is if a downloader needs to be registered to count as a view. Otherwise, to rate people needs to be registered. Maybe that's the point

sawdust · Oct 7, 2016

vince said: ↑

What I have done is counting the total number of stars, that is to say number of ratings multiplied by rate. It gives a better idea than the average rating. The number of views is not very relevant. By the way, I always wondered how hundreds of people can view a level without rating or writing a small comment. I think at most 10 guys give rates to levels in this forum...
Click to expand...

Well, I like math so I applied some math to the view to rating ratio. If you look at the "Members Online Now" section of the homepage, you will notice that most of the time there are about 3x as many guests as members at any given time so we can reasonably estimate that about 3/4 of views are from guests. Those are views that CANNOT leave ratings unless they sign up. Also, not all views necessarily mean downloads, not all downloads mean wins, and not all wins mean ratings. We have about 4 filters in between views and ratings that can be summed up with the following formula.

(Ratio of member views to guest views) x (ratio of downloads to views) x (ratio of wins to downloads) x (ratio of ratings to wins) = percent of views that return a rating.

Let’s assume 25% of views are members, 50% of views are downloads, 75% of downloads are wins, and 25% of wins are ratings.

.25 * .5 * .75 * .25 = 2.3% of views can reasonably be expected to ratings.

Note that downloads, wins, and ratings are complete guesses based on members only, NOT guests since they were counted out in the first ratio. I also believe the estimated ratios I put are pretty optimistic. If someone else knows how to get these actual stats that would be awesome, but I’m not a web guru like @vince and @Gepeto ;-)

vince · Oct 7, 2016

@Gepeto

Maybe (rates x ratings) = valuable notation? But then we could have (3 rates x 5 ratings = 15) < (10 rates x 2 ratings = 20) witch is bad...
Click to expand...

That is exactly what I have done. It is not so bad because as you said, 2 stars ratings are extremly rare.

cpw · Oct 7, 2016

vince said: ↑

@Gepeto That is exactly what I have done. It is not so bad because as you said, 2 stars ratings are extremly rare.
Click to expand...

Because you know......People would rather give 1-star ratings right away if they dislike a level so much

B Hill · Oct 7, 2016

Gepeto said: ↑

Everything is based on the orders proposed by the website for each category. It could be better with an access to raw data (base) but it is not possible. Maybe I'll try to figure out the rating more than just the number of rates but that's less obvious to code as it needs to reorder everything with this new result not just from what the website tell us on that.

Otherwise, maybe someone could find an algorythm that I can use to melt all the data and build a chart of the 20 best levels of the month?
Click to expand...

Oooh, now we are getting into territory I am familiar with. I have used statistical quartiles to rank applications based on a combination of metrics- total number of launches, total application running time, date of the very first launch, date of the most recent launch, time elapsed between launches, etc. The key to this analysis is normalization- and guess what? The website already provides you the normalized data! Woohoo!
Let me explain- the website already gives you a ranked list of levels based on different criteria- rating, number of views, number of comments, etc. For each criteria, you can take each list and break it down into quartiles. For example let us say there were 100 levels uploaded in the last month (to make math easy). To split these into 4 quartiles, quartile 1 (Q1) is levels 1 to 25, Q2 is levels 26 to 50, Q3 is levels 51 to 75 and Q4 is levels 76 to 100. Assign each quartile a score- Q1 would be 4 points, Q2 3 points, Q3 2 points and Q4 no points. Do this for each criteria and add up the points for a total score. Example:

My Level1- lots of views and ratings, OK average rating, not many comments
Code:
Criteria Type:  Views    Rating      #Ratings     #Comments
Quartile:         Q1        Q2         Q1          Q4
Scores:            4         3          4          0
Total score=11
My Level2- low number of views, rating and comments, but high average rating:
Code:
Criteria Type:  Views    Rating      #Ratings     #Comments
Quartile:      Q3          Q1          Q3           Q3
Scores:        2           4           2           2
Total score=10
So in this example, My Level2 has a very high average rating, but not much activity (views, comments, number of ratings) on the level. It has an overall score just below my other Level1 that had a slightly lower average rating but much more activity.

I could probably whip up a spreadsheet showing an actual example if you have the complete ranked lists of levels for the last month.

Gepeto · Oct 7, 2016

Wow... I do love this conversation For now I am back to coding to get the whole data over a range of dates, but of course, everything is from "now" . So I do not take just the 20 of each categories given from the website, but I am grabbing all data in a range in order to do whatever I want with that (I'll check if I come to the same result than the website ). So @sawdust and @B Hill your advices are really precious. @B Hill I would not be surprised that you work in a Big Data company

I'll dig on math and stats and let you know. For now I try to end up something good with the data

B Hill · Oct 7, 2016

OK here is a spreadsheet I whipped up. I used the data above with the 20 top Levels in each of 4 categories. What this really represents is activity on the levels, since the actual star rating for each level is not available. But this gives a very good picture of the most active levels above. The first table is the raw quartile scores, scroll down in the page to see the final ranked scores.
Mekorama Levels Scored by Quartiles

OpenOffice version

B Hill · Oct 7, 2016

If you can't view the file above here's the final result:

Final ranking: Title Author Total score

Opening Doors meko 12

Limbo richardfu_ 11

The Mountain meko 10

Glyph Walk richardfu_ 10

Praying For Rain nGord 8

No way to Win TR O 8

Deserted Curse meko 7

Expedition IV richardfu_ 7

Mountain Depth richardfu_ 7

Crouching Tiger TR O 7

Transformers (re) Gepeto 6

Padlock Game v2 TR O 6

OOOOOOO SCAAARYYYYY!!!! Scare Crow 6

STEPPING UP Mike_CN 6

Choice meko 6

Dark Waters richardfu_ 6

Begin Again meko 5

Mini Outpost richardfu_ 5

Special Forces Vol I S. Nagy 4

Expedition III richardfu_ 4

Swamp Khudrat 4

Path of Courage richardfu_ 4

Chateau Sawdust
4

Ancient Temple Lazarus 4

Safebreaker2 cheinz 4

Rails Mscript 4

6 Piece Burr sawdust 4

Headroom ZachMan 3

Super Mario Bros V3 sawdust 3

Claw Machine Box TR O 3

Fishing richardfu_ 3

Ship Picking richardfu_ 2

Expedition V richardfu_ 2

Electric Cute LooKing? 2

A lofty goal Jose Luis Galiano 2

Cae11 Chuckthulhu 2

Castle KPACABA 1

Cursed Tower richardfu_ 1

Plan B LooKing? 1

Flipped sawdust 1

Jump Pipe Muslim Arizzy 1

Sudoku Mscript 1

Swivel Muslim Arizzy 1

Manage Your Time Muslim Arizzy 1

Screenshot_2016-09-08-23-01-41-47 Muslim Arizzy 1

B Hill · Oct 7, 2016

Gepeto said: ↑

. @B Hill I would not be surprised that you work in a Big Data company
Click to expand...

I do not, I work in healthcare, on the IT side. But I program for Haiku OS in my spare time.

Gepeto · Oct 7, 2016

I was reminded of Quartiles on Wikipedia @B Hill according to your sheets this method seems relevant. So I am digging on the subject to try to implement something in my code. I didn't get yet what decides the score you assigned to each quartile.

Astral_Mage · Oct 8, 2016

Reputation plays an important role in the statistics as it seems, but it makes sense though :3

Log in or Sign up

Forums

Statistics Levels statistics - Research for monthly reports

Gepeto MekoStudio Architect Staff Member

Gepeto MekoStudio Architect Staff Member

Gepeto MekoStudio Architect Staff Member

Gepeto MekoStudio Architect Staff Member

cpw Retired Moderator

vince Famous Member

Gepeto MekoStudio Architect Staff Member

vince Famous Member

cpw Retired Moderator

Gepeto MekoStudio Architect Staff Member

sawdust Retired Moderator

vince Famous Member

cpw Retired Moderator

B Hill Active Member

Gepeto MekoStudio Architect Staff Member

B Hill Active Member

B Hill Active Member

B Hill Active Member

Gepeto MekoStudio Architect Staff Member

Astral_Mage Well-Known Member

Share This Page

N°	Title	Author	Views	Rate	Likes	Comments	Date
1	Praying For Rain	nGord	1K	3	1	58	10/02/2016
2	Deserted Curse	meko	650	3	2	23	09/20/2016
3	Opening Doors	meko	604	7	4	22	09/12/2016
4	The Mountain	meko	590	3	4	21	09/09/2016
5	Limbo	richardfu_	587	10	4	14	09/02/2016
6	Begin Again	meko	564	3	2	20	09/05/2016
7	Headroom	ZachMan	562	0	1	0	09/08/2016
8	Transformers (re)	Gepeto	500	3	1	21	09/07/2016
9	Glyph Walk	richardfu_	496	11	3	19	09/16/2016
10	No way to Win	TR O	494	5	3	29	09/12/2016
11	Special Forces Vol I	S. Nagy	468	4	1	19	09/04/2016
12	Padlock Game v2	TR O	467	3	0	28	09/18/2016
13	OOOOOOO SCAAARYYYYY!!!!	Scare Crow	419	0	0	24	09/21/2016
14	Expedition IV	richardfu_	406	10	1	16	09/12/2016
15	STEPPING UP	Mike_CN	387	0	0	24	09/24/2016
16	Castle	KPACABA	371	0	1	11	09/11/2016
17	Choice	meko	357	7	2	19	10/03/2016
18	Mountain Depth	richardfu_	353	7	5	10	09/10/2016
19	Expedition III	richardfu_	353	9	3	9	09/06/2016
20	Cursed Tower	richardfu_	343	2	0	11	09/06/2016

Log in or Sign up

Useful Searches

Statistics Levels statistics - Research for monthly reports

Gepeto MekoStudio Architect Staff Member

Gepeto MekoStudio Architect Staff Member

Gepeto MekoStudio Architect Staff Member

Gepeto MekoStudio Architect Staff Member

cpw Retired Moderator

vince Famous Member

Gepeto MekoStudio Architect Staff Member

vince Famous Member

cpw Retired Moderator

Gepeto MekoStudio Architect Staff Member

sawdust Retired Moderator

vince Famous Member

cpw Retired Moderator

B Hill Active Member

Gepeto MekoStudio Architect Staff Member

B Hill Active Member

B Hill Active Member

B Hill Active Member

Gepeto MekoStudio Architect Staff Member

Astral_Mage Well-Known Member

Share This Page