### Re: croco-puzzle

Posted:

**Wed 10 Dec, 2014 8:48 am**Does anyone have the rules of today's Aquarium puzzle on Crocopuzzle? I pressed "Initialise" assuming the rules would be presented but I can't see them.

Thank you.

Thank you.

Discussion forum for UK puzzlers. The UKPA is the UK representative of the World Puzzle Federation.

http://forum.ukpuzzles.org/

Page **28** of **28**

Posted: **Wed 10 Dec, 2014 8:48 am**

Does anyone have the rules of today's Aquarium puzzle on Crocopuzzle? I pressed "Initialise" assuming the rules would be presented but I can't see them.

Thank you.

Thank you.

Posted: **Wed 10 Dec, 2014 12:31 pm**

David,

Perhaps you've already worked this out for yourself?

We've seen this type before (as 'water tank' ?) in past WPCs.

"Fill cells with water so that the row/column clues are satisfied.

Within an outlined area, cells with water cannot have empty cells beside or beneath them"

Perhaps you've already worked this out for yourself?

We've seen this type before (as 'water tank' ?) in past WPCs.

"Fill cells with water so that the row/column clues are satisfied.

Within an outlined area, cells with water cannot have empty cells beside or beneath them"

Posted: **Wed 10 Dec, 2014 5:24 pm**

Oh dear. I forgot that initializing started the clock. It must be unprecedented for a U1 puzzle not to previously have appeared as a Preisratsel.

Posted: **Wed 10 Dec, 2014 6:52 pm**

david mcneill wrote:It must be unprecedented for a U1 puzzle not to previously have appeared as a Preisratsel.

Or a "Rätsel ohne Preis", as it is now called...

Posted: **Wed 10 Dec, 2014 7:23 pm**

david mcneill wrote:Oh dear. I forgot that initializing started the clock. It must be unprecedented for a U1 puzzle not to previously have appeared as a Preisratsel.

It's probably a little bit late, but you can practice the ten puzzles of the design contest...

Posted: **Sat 07 Feb, 2015 11:55 pm**

I'm wondering if today's daily puzzle has a broken key?

I can't find anything wrong with my solution, and nobody has officially solved it yet (only up for about an hour, but even for a fairly tricky puzzle I'd expect a handful by now).

I can't find anything wrong with my solution, and nobody has officially solved it yet (only up for about an hour, but even for a fairly tricky puzzle I'd expect a handful by now).

Posted: **Sun 08 Feb, 2015 8:25 am**

sknight wrote:I'm wondering if today's daily puzzle has a broken key?

I can't find anything wrong with my solution, and nobody has officially solved it yet (only up for about an hour, but even for a fairly tricky puzzle I'd expect a handful by now).

Key is fixed. Sorry.

Posted: **Sat 04 Apr, 2015 7:49 am**

I'd like to point you to my Abstract Adventure, where one of the goals of the design was to create an almost completely language independent adventure. Ideed there are only three places where you can find language: First, the title, which is the german version of "Abstract Adventure". Second, a note, that all data can be lost in rare circumstances, if you are not logged in. And last the congratulation message at the end, where I had no idea, how to express this in an abstract way.

I hope you'll enjoy it!

I hope you'll enjoy it!

Posted: **Wed 30 Dec, 2015 10:27 am**

Does anyone know if there are rules issued for the Chatroom puzzles recently featured on Crocopuzzle? I haven't tried them yet.

Posted: **Wed 30 Dec, 2015 4:13 pm**

Well they are there on the puzzle app itself. I went ahead and patiently typed them into google translate. To summarise:

All telephones are connected to a white circle and a black circle via horizontal/vertical paths connecting the centres of the cells. The numbered clues in the circles indicate how many turns the path makes before it reaches the telephone. All cells are used by some path.

All telephones are connected to a white circle and a black circle via horizontal/vertical paths connecting the centres of the cells. The numbered clues in the circles indicate how many turns the path makes before it reaches the telephone. All cells are used by some path.

Posted: **Wed 30 Dec, 2015 5:34 pm**

Thanks Tom. Can we assume that none of the paths can cross nor revisit the same cell more than once?

Posted: **Wed 30 Dec, 2015 5:52 pm**

Yes - the app won't let you do that

Using all the cells is important, as I found out with today's puzzle

Using all the cells is important, as I found out with today's puzzle

Posted: **Thu 31 Dec, 2015 2:18 pm**

detuned wrote:I went ahead and patiently typed them into google translate.

I hope you'll never need to do that again, because in a few days the Ü-Rätsel will switch from java applets zu javascript applets. The later ones do provide the interface and instructions in english (thanks to Lasse for translating). You just need to go to the "Benutzerkonto", look for "Sprachwunsch" and check "english" there. You can test it allready, at the puzzle of the week, selecting one of the puzzle types, that have been converted allready - the list of converted types is listed in the second paragraph on that site.

When introducing a new type, it might be, that the english translation is not available in time. But in that case you can download the german texts from GitHub and just copy & paste into google translate.

Posted: **Mon 18 Jan, 2016 11:28 pm**

In case anyone missed it (like me!) the croco apps are now back up and running. Looking good so far!

Posted: **Tue 26 Jan, 2016 8:23 pm**

I thought the Masyu app was a bit odd - kept putting crosses on the grid if the mouse twitched the wrong way. All the rest seem to be working fine, with some interesting puzzles too.

Posted: **Thu 17 Mar, 2016 4:10 am**

It looks like the croco-puzzle Marathon is running over the next fortnight.

Some time between 18th March and 4th April you can choose a 24-hour period, within which you solve as many of the 87 puzzles (3 each of 29 types) as possible. Ties are broken by total overall time.

Each puzzle has a 10 minute limit. Wrong answers are ignored. You can have breaks in between puzzles however you like. They expect fast solvers to finish everything in ~3 hours.

Some time between 18th March and 4th April you can choose a 24-hour period, within which you solve as many of the 87 puzzles (3 each of 29 types) as possible. Ties are broken by total overall time.

Each puzzle has a 10 minute limit. Wrong answers are ignored. You can have breaks in between puzzles however you like. They expect fast solvers to finish everything in ~3 hours.

Posted: **Tue 27 Jun, 2017 8:32 am**

I have just been sent an e-mail congratulating me on becoming a second Dan. Second Dan used to be for a long-term average of over 2200 and I'm nowhere near that. Have the rules changed?

Posted: **Tue 27 Jun, 2017 10:10 am**

david mcneill wrote:I have just been sent an e-mail congratulating me on becoming a second Dan. Second Dan used to be for a long-term average of over 2200 and I'm nowhere near that. Have the rules changed?

Well, the rules have changed - I removed some part of the formula which proved to press the rating down over time instead of stabilizing it, what it was ment to do. I changed this retrospectifly.

But concerning the email: This was a bug in the software. Meanwhile it's fixed and the grades have been reset to where they belong to.

Posted: **Tue 27 Jun, 2017 6:53 pm**

Well I seem to have done very well out of this - I've been pushed up by 100 points or so. With a little less carelessness I think 3rd dan could be in sight!

Is it possible for you to explain a little more about the formula Berni? This kind of thing does interest me a lot.

Is it possible for you to explain a little more about the formula Berni? This kind of thing does interest me a lot.

Posted: **Tue 27 Jun, 2017 8:29 pm**

detuned wrote:Is it possible for you to explain a little more about the formula Berni? This kind of thing does interest me a lot.

Well the formular with an explanation (in german though) can be found at Rating description. In the formula "dauer" is "duration", "min" is the fastest and "med" is the median of all solvers. With this you get your "daily rating". The rating is then adjusted towards that daily rating by a certain percentage, which depends on how many puzzles have been there in the last 30 days.

What I changed back is the 2 in the formula. In 2013 I thought it would be better to have a look at the strength of the contestants to decide on this value, but I underestimated the feedback effect. With some experimenting I found out, that the rating of the highest players starts to drop constantly (being at about 2300 in year 2040), while with a 2 it will stay at about 2700 (using the results of the last view years periodically for estimation).

I hope this brief explanation is understandable...

Posted: **Tue 27 Jun, 2017 9:17 pm**

Yes, I think I understand. Instead of fixed logistic curves defined by a fixed parameter of 2, this exponent was variable. Do you know to what extent this value varied away from 2 before your change?

I'm also curious about the updating factor f - is that also a fixed parameter?

As an aside, I have looked into this topic before. I experimented with the beginning of a system based on Bayesian probability whereby log(solving time) can be rescaled and fitted to a normal distribution, using the observed mean and standard deviation of the log(solving times) for a particular puzzle. The solver's rating is also fitted to a normal distribution and given by the cumulative distribution. In that way you can tailor the variance to individual solvers, and therefore update the ratings in a dynamic fashion. My goal was (and is!) to overcome the problems experienced by systems like Elo whereby you have exactly one of (a) converge on your rating very very slowly, or (b) run the risk of getting absolutely hammered for the occasional bad puzzle.

I had one updating mechanism based completely on the assumption that a player's true skill was ultimately fixed - which is not true in reality and ultimately leads to ratings being biased by performances from long ago which may no longer be relevant. This got people to the right level pretty quickly, but I don't think would work in the long run.

I also believed I could adapt this system to fairly normalise other competitions, for example the GP series. Of course you could do exactly the same with the croco formula and get better results than simply dividing through the best solver's score.

One of these days I'll get around to finishing the work!

I'm also curious about the updating factor f - is that also a fixed parameter?

As an aside, I have looked into this topic before. I experimented with the beginning of a system based on Bayesian probability whereby log(solving time) can be rescaled and fitted to a normal distribution, using the observed mean and standard deviation of the log(solving times) for a particular puzzle. The solver's rating is also fitted to a normal distribution and given by the cumulative distribution. In that way you can tailor the variance to individual solvers, and therefore update the ratings in a dynamic fashion. My goal was (and is!) to overcome the problems experienced by systems like Elo whereby you have exactly one of (a) converge on your rating very very slowly, or (b) run the risk of getting absolutely hammered for the occasional bad puzzle.

I had one updating mechanism based completely on the assumption that a player's true skill was ultimately fixed - which is not true in reality and ultimately leads to ratings being biased by performances from long ago which may no longer be relevant. This got people to the right level pretty quickly, but I don't think would work in the long run.

I also believed I could adapt this system to fairly normalise other competitions, for example the GP series. Of course you could do exactly the same with the croco formula and get better results than simply dividing through the best solver's score.

One of these days I'll get around to finishing the work!

Posted: **Wed 28 Jun, 2017 2:23 pm**

detuned wrote:Yes, I think I understand. Instead of fixed logistic curves defined by a fixed parameter of 2, this exponent was variable. Do you know to what extent this value varied away from 2 before your change?

It was in between 1.69 and 2.38. Distribution: 5x 1.7, 72x 1.8, 286x 1.9, 817x 2.0, 536x 2.1, 221x 2.2, 261x 2.3, 39x 2.4.

I'm also curious about the updating factor f - is that also a fixed parameter?

It's basically calculated to have puzzlers on level after one year, provided, the frequency of the puzzles is the same as in the last 30 days. If there is a puzzle every day, the maximum jump is 50 points (having rating 0, scoring first). In that case f = 1/60, because (1/60)*3000=50.