maybe I'm a stat junkie, but...

[sillyface]

I'm a newb here, and it's taking a while for me to grasp how much each factor (board size and difficulty, time) affects the scoring. I'm not sure either how difficult each board is, as I may have used a poorer approach on some than on others, how to judge?

I'm thinking that others may have a similar issue, so I've started recording my results as comments, X s (seconds) X t (turns) X p (points). eg: 492s 383t 69p

Hope others find it helpful. If I repeat a board several times it will become more useful, as we'll be able to compare the results and see, for example, that if a similar number of turns are used, and it's 100s quicker, it will make a difference of, oh, THAT many points. On that particular board.

[sillyface]

Just played a board where someone else has put up this same information in the comments.

Player x - 249s 192t 56p
Sillyface- 172s 110t 52p

I'd have to guess that the score given is dependent upon previous player's performance, it certainly doesn't seem to be intrinsic to the board layout.

[sillyface]

I'm starting to post five stats, the first three as before, followed by the fastest time under the bell curve, then the slowest time under the bell curve.

[FranklyMyDear]

I wonder if the scoring system has changed. When I first started playing, setting a record on a 5x5 board would garner 4 points. Now when I set a record on the same size board, I get 8 points.

[aquagirl555]

I'm starting to post five stats, the first three as before, followed by the fastest time under the bell curve, then the slowest time under the bell curve.

I like that idea, how about we do something like this:
268s 312t 132p
208h 1860L

I chose Upper case 'h' and 'l' because lower case 'l' looks weird with the numbers.

[sillyface]

I wonder about the scoring system too.

I've been using f before the fastest time, s before the slowest.

EDITED TO ADD e.g. 137s 163t 107p f105 s1580

[sillyface]

Trying to figure out the scoring system seems futile. From the same puzzle:
May 21st - 256s 213t 93p f133 s2230
May 25th- 149s 175t 113p f133 s2170

Obviously the data is somehow corrupted. The slowest ever time for a particular puzzle cannot get faster over time. I've previously noticed that the 'slowest times' under the bell curve are rounded off to five second increments; I now wonder if they're even generated from player's times. Higher 'slow times' do seem to affect scores, so yeah, futile.

Maybe the data isn't corrupted, maybe I've falsely assumed that data is used where it is not. Many other people seem to consider that I have a rather narrow idea of what is logical.

[FranklyMyDear]

maybe I'm a stat junkie, but...

Ain't nothin' wrong with being a stat junkie. I'm kinda one myself....and a popcorn junkie, too!

[loverlylady]

"Points are awarded according to each puzzle's difficulty, and the speed with which it is solved." I think difficulty simply refers to the size of the puzzle. Number of turns does not seem to be a factor at all.

[misseratic]

Points depend on how fast you complete a board compared to the average of times to date. You should always score maximum points for a record time, but may later score fewer points with a quicker time !

eg For 11 x 19 boards the maximum score is 199 points.
- player A takes 400 seconds = 199 points for posting the record time for that board.
- player B then takes 200 seconds = also scores 199 points, new record.
- player A tries again, takes 300 seconds but will score fewer than 199 points despite being quicker than their earlier attempt, because they are no longer the fastest.

Just a theory but 'slowest times' may appear to get faster if another player starts a puzzle but doesn't complete it (Save Progress button). Their board is saved for 24 hours, with the clock ticking and sooner or later they will hold the slowest time. Then if they fail to complete the board within 24 hours their time is dropped and no longer counts as the slowest time.

[uigrad]

The slowest ever time for a particular puzzle cannot get faster over time.

I just came across one of your puzzle comments. You recorded the slowest time as 1315 seconds, but now I see it is 1195.

My guess is that the fastest time you see is actually the fastest, but the slowest is at the 90 percentile mark, or something like this.

If it really showed the slowest, then you would see times above 10,000 seconds (3 hours) all the time, since you can come back to a puzzle up to 24 hours later.

I've played logic-puzzles.org for a while, and theorized that the slowest times don't show up there, using the same reasoning. I'm glad you actually are recording this data, as it validates my theory!

[uigrad]

When I do these puzzles now, I just record my time, the fastest time, and the median time. So, I write something like this:

180, f=80, median=360

Once you have this, then you know what all 5 tick marks are on the chart. The first one is the fastest time (80s). The second one is half way between the fastest and the median (220s). The middle one is the median (360s). The 5th spot is either the median multiplied by 5 or your current time, whichever is larger (360 * 5 = 1800s). The 4th spot is half way between the 3rd spot and the 5th spot (1080s).

If you play a lot of puzzles, you'll see values that are not whole numbers some times. These can only appear in the 2nd or 4th spot of the graph. The number in the second spot of the graph will be xxxx.5 if fastest time + the median time is an odd number. The 4th spot of the graph is almost always an integer, but can be xxxx.5 if the median + your time is an odd number and your time is larger than the median time multiplied by 5.

Every size of puzzle has a maximum and minimum amount of points that you can score. I haven't experimented much to figure out the minimums, but the maximums are:

8 points for a 5x5
26 points for a 7x7
56 points for a 9x9
116 points for an 11x11
161 points for an 11x15
199 points for an 11x19

To figure out your score, your percentile must be calculated. This isn't a real percentile, because it's just based upon values: your time, the best time, and the median time. So, it's kind of a faked percentile, but I'll still call it a percentile because it's easiest to think of this way. Here's how it is calculated:

Your percentile = 1.00 if your time is better than the fastest time.
Your percentile = 0.50 if your time is the same as the median.
Your percentile = 0.00 if your time is worse than the median * 5

If your time (t) is between the fastest time (f) and the median time (m), then your percentile is:

Your percentile = 1 - 1/2 * (t - f) / (m - f)

If your time (t) is between the median time (m) and point 5 on the graph (m*5), then your percentile is approximately:

Your percentile = 1/2 - (((t - m) / m) - 1) / 8

I've noticed my calculations are a little off for the scenarios where I'm between (m) and (m*5), so the formula I listed above for your percentile in those cases may be off.

Now that you know your percentile, you can multiply it times the max points available for the puzzle. As long as this value is larger than the minimum number of points for the puzzle, this is what you will get. If it is lower than the minimum value for the puzzle, then you get the minimum value.

So, back to our example, where I recorded the following for an 11x11:

180, f=80, median=360

The median will be 1 - 1/2 * (180 - 80) / (360-80)

That comes out to be 1 - 0.5 * (100 / 480) = 0.8214

Now you multiply that by the maximum points for a puzzle of that size (116):

116 * 0.8214 = 95.29

Now you round that to the nearest integer (95), and that is the number of points you get!

[uigrad]

Trying to figure out the scoring system seems futile. From the same puzzle:
May 21st - 256s 213t 93p f133 s2230
May 25th- 149s 175t 113p f133 s2170

Using the scoring information from the previous post, I can show you that these work out.

I took the slowest times you listed (2230 and 2170), and divided them by 5 to get the medians: 446 and 434.

Then I used your times (256 and 149) along with the fastest times (133 and 133) and the medians (446 and 434) to get your percentiles: (.8035 and .9734)

From there, I use the max value for an 11x11 (116) and get your scores (93.21 and 112.92).

You recorded 93 and 113, which is exactly what I get after rounding.

Obviously the data is somehow corrupted. The slowest ever time for a particular puzzle cannot get faster over time. I've previously noticed that the 'slowest times' under the bell curve are rounded off to five second increments; I now wonder if they're even generated from player's times.

It turns out you were correct. Points 2, 4, and 5 on the graph are completely fabricated based on your score, the fastest score, and the median score.

Higher 'slow times' do seem to affect scores, so yeah, futile.

Higher 'slow times' just indicate higher medians, and yes they do affect score. But your original post (several years ago now) prompted me to figure this out myself, so it was certainly not futile, haha!

[PIXEL_sm]

the minimum for a 11×19 is 20