coloin wrote:A new member [FLOCKMAN] has expressed an interest [via pm] in making sudokus with high SE rating - and he has made a few new ones ......
Which has rekindled publicizing some work i did.
It has been said that the area with 22,23 clues are well searched and there are few new puzzles to be found with high ratings.....
However by changing a few clues and testing - it seems that this is not the case !!!!
Here are a few new ones - found by repeated vicinity search +1 or +2 on known puzzles .....some time ago.... but it took a determined search
as you can see a few SE 11.5 ... all well and good ....
however a search around Champagne Dry - brought a few new puzzles ...43 with skfr > 10.6 , 30 already in file01
except 13 were new .... including a 11.2 ....
which made me wonder..... how these puzzles were new ....
untill these facts explained it
{-1+1} of champagne dry - 68 puzzles
{-2+2} of champagne dry - 13,470 puzzles
{-3+3} of champagne dry - ? at least 15,000,000 puzzles - and i only generated / rated the first million
so it seems that there may well be a few more out there ....
and the 11.2 may well appear in a well known g.r.emlin game
Hi coloin,
Now that I am back, I intend, after cleaning of several things, including a 3 months search of 17 clues puzzles, to process the new findings of FLOKMAN, but some comments on your post
It has been said that the area with 22,23 clues are well searched I wrote this and I would do the same to-day.
a) The yield of a new attack in this area is very low compared to what can be achieved with more clues
b) and the chances to get high ratings is still lower
but in any area, new puzzles can be found meeting the requirements to enter the data base of potential hardest, especially in the bottom part of the file where several puzzles created in the past have no been included.
A new comer has more chance to find something in widely explored area. In my search, all puzzles with skfr ER<10 are lost as seed for the next cycles and a seed used will never come again.
Any error in data handling means that the corresponding seed is definitely lost.
Many of the recent games have used patterns existing in the data base of potential hardest and the results generally go in the same direction:
a) most of the high ratings are already in the data base,
b) a few number of new >=11.0 have been produced
c) but the majority of the new ratings are limit down for the data base.
The current pattern proposed by patrice has more clues, an area not yet deeply searched, It will be interesting to see how many new puzzles (within the data base criterai) are produced
PS: the chance to find a new 17 are extremely low, what does not prevent me to check whether this is true in a very long process