Today's update on GitHub is about the fine tuning of the four pre-selected subsets of impossible patterns.
The selections are now based on the first 50,000 T&E(3) puzzles in mith's list of 158,276 min-expands (instead of 20,000 before).
Two new patterns have been added to the lower part of the list.
Globally, the four subsets are quite stable when puzzles are added to the analysis. Here are the numbers of puzzles (among the 50,000) where each pattern is "useful" (i.e. allows at least one elimination - when all chains lengths are restricted to 8). The partition into 4 sets appears quite naturally.
- Code: Select all
Trid = 49089 (it's not 50000 because even if the pattern is present, it many not be useable by ORk-chains).
- In Imp630-Select1:
EL13c290 = 10863
EL14c30 = 6746
EL14c159 = 4241
EL14c13 = 3243
EL14c1 = 2737
- In Imp630-Select2:
EL13c30 = 1409
EL10c28 = 1375
EL13c179 = 1295
EL13c176 = 1159
EL13c234 = 925
EL13c171 = 742
EL10c6 = 738
- In Imp630-Select3:
EL13c259 = 577
EL10c8 = 509
EL13c172 = 413
EL13c187 = 378
EL10c4 = 378
EL13c175 = 371
EL14c19 = 360
- In Imp630-Select4:
EL14c93 = 273
EL15c97 = 234
EL14c154 = 160
EL10c10 = 157
EL13c170 = 155
EL13c168 = 153
EL13c19 = 150
This seems to support my idea that only a few of the 630 impossible patterns in two bands (or two stacks) are really useful (those in Select1 or Select2) - although quantifying this claim requires more investigations.
Note: a friend and user of SudoRules (and expert in finding whips) remarked that the recent updates are only for the few people interested in very rare puzzles. This is absolutely true (except for the generic ORk-chain and ORk-ultra-persistency rules). But notice that finding Trid-ORk-whips is not more difficult than finding ordinary whips; which should allow this expert friend to solve most of the puzzles in mith's list.
Notice also that these recent updates also illustrate the original nature of CSP-Rules as a research tool.
.