I've compiled a list of all the games on the site by their activity (as of 2016-04-21 07:00 site time): https://docs.google.com/spreadsheets/d/10wG7y0e4jFtC-Tt9UKvjbnTlXfvzw-4vz0fKnNHFGQ0/edit?usp=sharing

The larger the number, the more active a game is. You can sort the columns by hovering above the headers and clicking the arrow. Here's a quick slice of it:

1. • Diddy Kong Racing -- 363.97
2. • Freedom Planet -- 187.62
3. • SUPERHOT -- 180.04
4. • Undertale -- 167.47
5. • The Legend of Zelda: A Link to the Past -- 145.59
6. • Super Mario Sunshine -- 131.69
7. • Super Mario 64 -- 119.03 8) - The Legend of Zelda: Ocarina of Time -- 115.37
8. • Grand Theft Auto V -- 111.15
9. • Portal -- 100.84
10. • Super Mario World -- 97.11
11. • Kalimba -- 84.56
12. • Super Mario Bros. -- 83.37
13. • The Legend of Zelda: Majora's Mask -- 74.43
14. • FAST Racing NEO -- 72.64

Level runs impact the overall activity score more than it should; I haven't came up with a way to resolve it yet since some games use only level runs (like PAYDAY 2). I included activity for Full Game runs separately as a temporary fix so the racing games don't hog all the glory.

I calculated it by using a weighted count formula, meaning the older the run, the less it's 'counted'. Runs from today are counted as a whole 1, a month old is about 0.5, two months is 0.25, three is 0.125, etc. I was only able to use runs that had dates their date recorded in someway (hence the 'AllRuns' column). If you want to know the exact formula: Int([CurrentDate]-Min([VerifyDate],[SubmissionDate],[RunDate]))^(2^(-1/30.4375)) -- It uses the oldest date of those three since run dates aren't always correct; it had to of been done before all of those dates. If one of the dates is missing, it's treated as 99999, and if they're all missing the run is effectively ignored.

Interesting work. But I'd suggest if the peak IL figure is 2-3x the peak full run figure, the scaling's a bit off.

Added a new column, "Freshness". It shows how 'fresh' a game is.

Example: Slither.io is the freshest popular game. All of its runs have been in the past week. Meanwhile Super Meat Boy, while it's still fairly popular, has lost a lot of its activity overtime. Most of the time, fresh games are recently released ones.

For the more active games, the freshest games are:

1. Slither.io
2. Dark Souls III
3. Hyper Light Drifter
4. Star Fox
5. Golf With Friends
6. Angry Video Game Nerd II: ASSimilation
7. Defunct 8) The Division
8. The Simpsons Game
9. Stardew Valley

And the least fresh games are: 10) Wave Race 64 9) Super Mario Sunshine 8) The Legend of Zelda: Ocarina of Time 7) Metroid: Zero Mission 6) Super Smash Bros. Melee 5) Grand Theft Auto: Vice City 4) Splatoon 3) Sonic Adventure (DX) 2) The Legend of Zelda: Twilight Princess

1. Super Meat Boy

Deadest games (under 0.10 activity): SSX Tricky - Most runs (230) and players (85) Combat Arms - Most full-game runs (59)

I've been playing around with this a bit, although I've gotten the value to be a bit of a better fit, my main problem with it is that the value itself is kinda meaningless. It works fine for comparing/sorting games, but looking at a single value, there's no way to conceptualize if say 84.76 is a high or low activity.

Normalizing to a percentage with the highest value being 100% is the one idea I've had so far to deal with that, but I don't really like that either for various reasons.

For reference, here's the adapted formula:

By considering the length of the runs, full-game and level runs can be combined without issue. The scaling parameters ( /2 and +1 ) are included to strike a balance between considering length and number of runs (ten 1 hour runs represent more activity than a single 10 hour run). Time dependence is increased to a month old run counting for 0.25 instead of 0.5 and so on.

Here's the list of resulting highest activity games (the values will be a bit off as the snapshot of the database used is a week old):

Nice job adding duration to the formula. It was something I considered; I didn't want records that were long enough to be let's plays affect the outcome too much. If you think it doesn't make much of an impact, then it's probably fine.

While making the 'freshness' column (which is ln([Activity]/[RunCount])/ln(2^(-1/30.4375))) I found that the column was roughly 'weighted average age of runs' per game. When the run count is 1, the freshness is the age of the run. It applied similarly to the Activity column, which could best be summarized as 'number of recent runs', albeit not exactly. If you multiply it by (1+h/2) though, it could be stretched as 'recent hours worked' (since you can argue people work at least an hour worth before getting that run) or something else.

I will mention that if the half-life is at 15.21875, a run a day, everyday, would eventually converge to 22.46 (when not multiplied by time). You could divide the activity value by it, add an extra decimal, and call it 'average new runs per day'.

Edit: It'd be a good idea to put a hard limit on how large (1+h/2) can get; otherwise it can be manipulated. There are a couple of Super Mario Bros. 'runs' that are 99:00:00 that screw the activity up. I'd clamp it to 5 (which would be 8 hours).

Added three new sheets, which can be accessed on the bottom of the screen: Top 500 Games Categories Top 500 Categories

The Top 500 sheets were added mostly so sorting isn't such a pain when you want to check out the most 'popular' games/categories.

The categories sheets use Lighnat0r's modified formula, so they're calculated differently from the games sheets. They also don't have the names of the categories because of API limitations. Use the example link to go to (presumably) the world record run and see the category. From what I understand I would have to crawl every single game's API for category information, which would be strenuous for both the server and me. Game names are still there.

I love how at the moment of this post Gears of War 2 is 469th place trumping all other Gears games. Honestly I am surprised by this because Gears of War is the fastest of them all.

I've included the Activity value as I described previously to the game statistics page as an experimental thing.

Not sure I like the word 'worked' to describe the value, it seems even more vague and inaccurate than activity. Regarding the 99 hour 'runs', in the calculations I simply ignore any runs in that range (99 <= hours <= 100) since all current runs that hits are fakes.

It even has the number of followers, that's awesome.

Nevermind, I forgot about the time adjustment. -- I was curious about the formula so I checked it manually. It rounds down the number of days since run/submission/approval for each run, and the final activity value is rounded to the nearest hundredth. It also seems to update a game when there's a change in runs, or at least a new approved run.

Had some fun with numbers. I only crawled the games I follow as examples. I may crawl more if it doesn't torque the server. Most anticipated games • Blinx 2: Masters of Time and Space (7 followers and no runs) • Mario & Luigi: Partners in Time (13 followers and no recent activity) • Yoshi's Story (22 followers and very low activity [under 0.10]) Most watched runs -- Mario & Luigi: Bowser's Inside Story (2 runs with 1274 average views) Least watched runs -- Command & Conquer: Red Alert - Counterstrike (3 runs with 7 views, mostly because they're ILs)

Edit: I've noticed something for the more active games. This formula: ln((x-1)/x)¤30.4375/2/ln(2), x being the activity value, results in how long it takes for it go down by 1. This could somehow be used to think of a name.

While entering IL times in for one of the games I moderate, I noticed the activity (and by extension, run count) can be manipulated to inflate a game's activity more than it should, only limited to how active the moderators are. Games that only keep track of full-game runs would be stomped by similar games that also keep track of best segs/ILs. If I added all of 8-Bit Armies's ILs same day as the first run, its activity would've been at about 27.74, and would probably put it in the Top 100. It comes down to how 'detailed' a leaderboard is. I mean really, it just shows more determination in keeping track of records, which isn't a bad thing. So in a sense it does show activity.

Ah four months ago...Dark Souls, Hyper Light, Stardew...people get bored quick, I guess.

Superhot too D: