Point score calculation change?

[ScalvinC]

Off and on, some of us compare points achieved for different levels of logic puzzles. This was going on for a few years, so we have a sense of what was a pretty good score, what was not. It's been about a year since I tuned in, but it seems like the point results are WAY higher for similar performance a year or so ago. Like the difference of 80-100 ish being pretty decent to several hundred points reflecting mediocre or even kind of poor performance. Did the algorithm for points change? Or did we just get 10x better for taking time away If it's changed, can you provide typical ranges of point achievement?

[JoshGrams]

And you're accounting for the new puzzles (mostly) not having as many high scores yet? Because if you meet or exceed the current record, you'll get the highest possible score regardless of whether the previous record was any good or not. If you get around the current average you'll get around half the highest score.

AFAICS it must fit a curve to the distribution of the existing times for that INDIVIDUAL puzzle, and calculates a high time cutoff based on the distribution? Then I suspect it just linearly interpolates your score based on where it fits between the current high time cutoff and the current record time, with using fixed low/high score values that are based on the size and difficulty.

So I'd think that if you have several years of data you'd be able to tell if the limits changed, right? Like the 3x4 Challenging go from 3 points to 80 points (80 if you match or exceed the current record), 4x4 Challenging goes from 12 to 290 points, and so on. Were you recording the then-current average & fastest times along with the scores? Because if you don't have those, you have no idea what the score for a particular performance should be, especially for a new puzzle. It'll vary depending what times everyone else has made on that puzzle.

[ScalvinC]

Thanks for pointing out that there is not a standard rubric for establishing the point values, but it's based on the performance of people who've tried the puzzle, if I understand your explanation. That's a little disappointing in that it's impossible to compare performance on one puzzle to another unless they both have had a large number of participants to somewhat even out the comparison base. Your idea of tracking the average and fastest times, and other metrics, will probably be the way to go for a better comparison across puzzles.

[JoshGrams]

Yeah, it's based on the performance of everyone who has completed the puzzle, for sure. Which is both disappointing (if you want to compare performance between different puzzles you have to compare times rather than scores) but also cool (at least I find it fun to be able to see the relative difficulty of individual puzzles based on their accumulated average times).

Sometimes you'll see a puzzle with comments going back to 2016 but its average time is 30-50 seconds faster than many of the other puzzles of that age... and you're like oh, I had a really good time (for me) on this puzzle but that's because it's easier than a lot of the Challenging ones. Or you'll do a puzzle with a worse average time but you did faster than your personal average time for that size/difficulty and then you can feel really proud of it.

Dunno. I usually just track times so I can tell if I'm getting better and don't worry too much about the scores, but on most sizes of puzzles I'm not anywhere near the top end, so tracking my personal times feels like the healthiest thing... especially since I usually get faster only after I slow down and take a while really analyzing what kinds of deductions I'm struggling to see, heh.