I would suggest that the order which one uses techniques to advance a puzzle depends on how likely the technique is to be effective. As a first cut, I've started to program up many of the techniques discussed on the site, including:

Naked Values (1-4 cells/candidates)

Locked Candidates (Block/Line, Block/Block)

Hidden Values (1-4 cells/candidates)

NxN Fish (X-wing, Swordfish, Jellyfish, Squirmbag, Fillet-o-Fish)

XY-wing, XY-chain, XY-ring, XYZ-wing, WXYZ-wing, and VWXYZ-wing

SueDeCoq

Almost Locked Sets (xz with the 1st set consisting of 1-4 cells)

X-cycles with 4-7 nodes, (Turbot Fish=5nodes, Grouped and/or Fillet-o-Fish)

Broken Wings

Uniqueness Tests (1, 2/2B, 3/3B, 4/4B) - when did type 5 appear?

Simple Colouring (1, 2, 3, partial Grouped)

I still hope to include Unique loops, BUG/BUG-lite, AUR and AUL. I may also include multi-colouring, but I would expect X-cycles to cover both simple and multi-colouring. With my solver if I run simple colouring after X-cycles I get no further reductions for Top1465 which suggests this may be the case. The idea is that I don't necessarily care to solve the puzzle (if that's the issue I can just run suexk and get the answer in milliseconds), but to know if I've done all that this human can do.

Back to the question of hierarchy. If I run my solver and order the results based on the number of puzzles which a technique advances (maybe next time I record the actual number of times a technique is used), I get the following order for Top1465 where the number of puzzles is shown in parenthesis. The numbers in braces are the order for 1000 randomly generated puzzles using suexg.

1) [3] Hidden single (1420)

2) [4] Locked line/block (1396)

3) [1] Naked single (1060)

4) [2] Naked pair (933)

5) [5] Locked block/block (671)

6) [7] Naked triple (599)

7) [13] Hidden Pair (548)

8) [18] Grouped type 1 colouring (306)

9) [6] Generalized WXYZ-wing (304)

9) [8] ALS-xz rule with at least 2 cells per ALS (304)

11) [12] ALS-xz rule with at least 3 cells per ALS (283)

12) [10] X-wing Fillet-o-fish (277)

13) [17] ALS-xz rule with at least 4 cells per ALS (257)

14) [11] Generalized VWXYZ-wing (237)

15) [15] Swordfish Fillet-o-fish (179)

16) [20] Grouped Turbot Fish (126)

17) [22] Hidden Triple (122)

18) [9] XY-wing (114)

19) [19] Turbot Fish (113)

20) [23] Grouped type 3 colouring (105)

21) [14] Generalized XYZ-wing (97)

22) [16] X-wing (78)

23) Type 1 Unique Rectangles (49)

24) Type 4/4B Unique Rectangles (46)

25) Swordfish (33)

26) [21] 4-node XY-chain (22)

27) SueDeCoq (19)

28) Type 2/2B Unique Rectangles (19)

etc.

This looks like an advertisement for ALS, grouping and fillets. Probably the most significant differences between the Top1465 and random draws are X, XY, and XYZ-wings being more prevalent and coloring less so. Two notable underachievers: no ungrouped Type 3 colouring advancements were found and only 4 XY-ring advancements. I probably still have bugs in my code and I only actually solve 622 of the 1465 puzzles. Has anyone seen similar results?

It would be nice to know which techniques actually crack the puzzle (ie only beginning techniques are required following use of the technique(s)) - something for the future.