Great, I think we can close this one.

BTW, there was another error where the scripts expected initial h values 
to always be reported. This is also not the case and has been fixed.

Thanks for clarifying. The dead_ends parsing should be fixed now. Can 
you parse successfully now?

Evaluations and Evaluated States are not the same.
Evaluated States is the number of states that were evaluated.
Evaluations is the number of heuristic evaluations performed.
They differ if:
1. More than 1 heuristic is used, or
2. A non-standard search algorithm (like LM-A*) is used

So they should both be collected and treated differently.

Yep, it's rather new. Runs 81+ use a different revision than runs 1-80.

Also, why is there sometimes a second entry about "Evaluations:"?

Plan length: 16 step(s).
Plan cost: 280
Initial state h value: 179.
Expanded 555494 state(s).
Reopened 0 state(s).
Evaluated 4507985 state(s).
Evaluations: 5051190
Generated 12725632 state(s).
Dead ends: 1360280 state(s).
Expanded until last jump: 555491 state(s).
Reopened until last jump: 0 state(s).
Evaluated until last jump: 4507964 state(s).
Generated until last jump: 12725598 state(s).
search space hash size: 4507985
search space hash bucket count: 6291469
Search time: 153.57s
Total time: 153.63s
Peak memory: 740080 KB

Apparently the code to print the number of dead ends was added after that 
version.

That's strange. Somehow the dead ends value is not printed in run 00081:

Do you have special code if the plan cost is 42?

Solution found!
Actual search time: 0.4s [t=0.41s]
board p2 slow0-0 n2 n0 n1 (0)
move-down-slow slow0-0 n2 n1 (6)
leave p2 slow0-0 n1 n1 n0 (0)
move-up-slow slow0-0 n1 n3 (7)
board p1 slow0-0 n3 n0 n1 (0)
move-up-slow slow0-0 n3 n4 (6)
leave p1 slow0-0 n4 n1 n0 (0)
board p1 slow1-0 n4 n0 n1 (0)
move-up-slow slow1-0 n4 n6 (7)
leave p1 slow1-0 n6 n1 n0 (0)
move-up-slow slow1-0 n6 n8 (7)
board p0 slow1-0 n8 n0 n1 (0)
move-down-slow slow1-0 n8 n4 (9)
leave p0 slow1-0 n4 n1 n0 (0)
Plan length: 14 step(s).
Plan cost: 42
Initial state h value: 24.
Expanded 4763 state(s).
Reopened 0 state(s).
Evaluated 25566 state(s).
Evaluations: 29708
Generated 65743 state(s).
Expanded until last jump: 4649 state(s).
Reopened until last jump: 0 state(s).
Evaluated until last jump: 25157 state(s).
Generated until last jump: 64204 state(s).
search space hash size: 25566
search space hash bucket count: 49157
Search time: 0.4s
Total time: 0.41s
Peak memory: 8580 KB

Exactly. I guess the dead ends output was added later?

Looks like it expects to find but fails to find a "dead_ends" entry?

I got the following error when running   ./downward-resultfetcher.py lm-opt/
in    /home/downward/erez/hg.fast-downward.org/new-scripts  (on habakuk):

Traceback (most recent call last):
  File "./downward-resultfetcher.py", line 431, in <module>
    fetcher.fetch()
  File "/home/downward/erez/hg.fast-downward.org/new-scripts/resultfetcher.py", 
line 257, in fetch
    file_parser.parse(props)
  File "/home/downward/erez/hg.fast-downward.org/new-scripts/resultfetcher.py", 
line 147, in parse
    self._apply_functions(props)
  File "/home/downward/erez/hg.fast-downward.org/new-scripts/resultfetcher.py", 
line 157, in _apply_functions
    function(self.content, props)
  File "./downward-resultfetcher.py", line 206, in get_iterative_results
    assert same_length(values[x] for x in group2), values
AssertionError: defaultdict(<type 'list'>, {'cost': [42], 'dead_ends': [], 
'expansions': [4763], 'generated': [65743], 'plan_length': [14], 
'initial_h_value': [24], 'evaluations': [25566], 'search_time': 
[0.40000000000000002]})