Tweet
I imported the Dec '24 leaderboard to a spreadsheet before it got archived.
Here are the raw stats and some relevant calculations as of just over 3 hours before the end of the competition:
s6ZpH0F.png External link: https://i.imgur.com/s6ZpH0F.png
Columns A through H are populated with the leaderboard data via the formula in cell A1, visible at the top of the image - you can use the same formula in your own spreadsheet if you're so inclined. The full table updates automatically every so often, and any calculations you make in other columns using that data will also update automatically.
Column I is calculated by multiplying the number of puzzles completed by the average seconds per puzzle, then dividing by 3600 (the number of seconds in an hour).
Column J is estimated by multiplying the number of puzzles attempted by the average seconds per puzzle, then dividing by 3600.
The estimate in column J presumes that the average time it took you to decide to abandon the puzzle is the same as the average time it took you to complete a completed puzzle. Most players who give up on a puzzle probably spend more time than average on it, so column J is typically an underestimate (except when a player has no imcomplete puzzles, of course). The more puzzles solved, the less of an underestimate it likely is. The data in these columns can be thrown off by leaving a puzzle for a long time but then coming back to complete it later. Those with many solves are likely impacted much less.
Column K is calculated by dividing the average points per completed puzzle by the average seconds per completed puzzle.
Column L is estimated by dividing the total points by the estimated average seconds per attempted puzzle.
These columns are where you see who is playing best. Most players wouldn't bother optimizing for this stat. For example, if you go AFK during a puzzle, you'd probably finish it when you come back, just to get points. Some who are invested in the competition might consider abandoning certain attempts, especially if the record for a given puzzle is fast and when their time is more limited - better to get a fresh chance at an easy max score than stay stuck on a low-scoring puzzle. These columns are also impacted by being AFK and similar circumstances, so it's typically an underestimate, just like column J.
During this competition, I abandoned hundreds of puzzles. I did this to optimize for points per second on completed puzzles, leaving tens of thousands of points on the table, and to chase individual puzzle records. My original goal wasn't to get a million points - I didn't know how much effort I'd need to put in to win the trophy until about a week in, and I only broke the previous top score of 568618 about a week ago. At that point, what motivated me to continue was improving my estimated points per second on all puzzles.
I mostly played 12x12 as that's certainly the best format for optimizing points per second: small mistakes don't lower points per second by too much, and the ratio of easy puzzles to hard ones is sufficiently high. Lowering the size ruins the former attribute, while raising it ruins the latter. Playing other sizes is bad if your aim is to win the competition.
I probably won't play this competition again, as there's definitely nothing left to prove. Good luck to all future competitors.
Here are the raw stats and some relevant calculations as of just over 3 hours before the end of the competition:
s6ZpH0F.png External link: https://i.imgur.com/s6ZpH0F.png
Columns A through H are populated with the leaderboard data via the formula in cell A1, visible at the top of the image - you can use the same formula in your own spreadsheet if you're so inclined. The full table updates automatically every so often, and any calculations you make in other columns using that data will also update automatically.
Column I is calculated by multiplying the number of puzzles completed by the average seconds per puzzle, then dividing by 3600 (the number of seconds in an hour).
Column J is estimated by multiplying the number of puzzles attempted by the average seconds per puzzle, then dividing by 3600.
The estimate in column J presumes that the average time it took you to decide to abandon the puzzle is the same as the average time it took you to complete a completed puzzle. Most players who give up on a puzzle probably spend more time than average on it, so column J is typically an underestimate (except when a player has no imcomplete puzzles, of course). The more puzzles solved, the less of an underestimate it likely is. The data in these columns can be thrown off by leaving a puzzle for a long time but then coming back to complete it later. Those with many solves are likely impacted much less.
Column K is calculated by dividing the average points per completed puzzle by the average seconds per completed puzzle.
Column L is estimated by dividing the total points by the estimated average seconds per attempted puzzle.
These columns are where you see who is playing best. Most players wouldn't bother optimizing for this stat. For example, if you go AFK during a puzzle, you'd probably finish it when you come back, just to get points. Some who are invested in the competition might consider abandoning certain attempts, especially if the record for a given puzzle is fast and when their time is more limited - better to get a fresh chance at an easy max score than stay stuck on a low-scoring puzzle. These columns are also impacted by being AFK and similar circumstances, so it's typically an underestimate, just like column J.
During this competition, I abandoned hundreds of puzzles. I did this to optimize for points per second on completed puzzles, leaving tens of thousands of points on the table, and to chase individual puzzle records. My original goal wasn't to get a million points - I didn't know how much effort I'd need to put in to win the trophy until about a week in, and I only broke the previous top score of 568618 about a week ago. At that point, what motivated me to continue was improving my estimated points per second on all puzzles.
I mostly played 12x12 as that's certainly the best format for optimizing points per second: small mistakes don't lower points per second by too much, and the ratio of easy puzzles to hard ones is sufficiently high. Lowering the size ruins the former attribute, while raising it ruins the latter. Playing other sizes is bad if your aim is to win the competition.
I probably won't play this competition again, as there's definitely nothing left to prove. Good luck to all future competitors.