File "Wijk2JBm", summary file from run by Jason Buczyna with Junior 10 in
full detail, and other chess engines less formally.  Since Junior 10 skips
evals and PVs for some depths that are not multiples of 3 when there is no
change, "o" and "i" mean non-atches "0" and matches "1" that were inferred.

TOPALOV 11ii1ii1 FULL MATCH (19. h5), Junior 10 delta = 0.15
TOPALOV 0o0oo1i1      MATCH (20. g6), Junior 10 delta = 0.14
TOPALOV 00oo1i10 PART MATCH (21. gxf7+ possible match), Junior 10 delta2 = -0.06
TOPALOV 11ii1ii0 PART MATCH (22. Be6+ possible match), Junior 10 delta2 = -0.04
TOPALOV 00oo0oo0 NON-MATCH  (23. b3), Junior 10 delta2 <= -1.20
TOPALOV 00oo0oo0 NON-MATCH  (24. Rhg1), Junior 10 delta2 = -0.58
TOPALOV 00oo0o0o NON-MATCH  (25. Bf5), Junior 10 delta2 <= -0.46
TOPALOV 00oo1i1i MATCH      (26. Bxc5), Junior 10 delta = 0.62
TOPALOV 11ii1ii1 FULL MATCH (27. Qh6), Junior 10 delta = 2.34
TOPALOV 11ii1ii1 FULL MATCH (28. Qxh7), Junior 10 delta = 2.40
TOPALOV 11ii1ii1 +0 FULL MATCH (29. Nd3), Junior 10 delta2 = -0.05 at depth 20
TOPALOV 11ii1111 FULL MATCH (30. cxb3), Junior 10 delta = 5.74.
TOPALOV 00oo0oo0 NON-MATCH  (31. Rxd3), Junior 10 delta2 = -0.09
TOPALOV 10oo0o0o PART MATCH (32. Be6), Junior 10 delta2 = -0.30
TOPALOV 11ii1i1i FULL MATCH (33. Bxf7), Junior 10 delta = 2.01
TOPALOV 11ii1ii1 FULL MATCH (34. Qg8+), Junior 10 delta = 5.81
TOPALOV 11ii1ii1 FULL MATCH (35. Kxa2), Junior 10 delta = 7.80

17 moves
 8 full matches (though 29.Nd3 was topped by 29.Rg4 at depth 19 or 20)
 5 part matches
 4 non-matches

Matches at various ply depths:
11: 10/17
12:  9/17
13:  9/17
14:  9/17
15: 11/17
16: 12/17
17: 12/17
18: 10/17

Moves ranked by delta/delta2:

TOPALOV 11ii1ii1 FULL MATCH (35. Kxa2), Junior 10 delta = 7.80
TOPALOV 11ii1ii1 FULL MATCH (34. Qg8+), Junior 10 delta = 5.81
TOPALOV 11ii1111 FULL MATCH (30. cxb3), Junior 10 delta = 5.74
TOPALOV 11ii1ii1 FULL MATCH (28. Qxh7), Junior 10 delta = 2.40
TOPALOV 11ii1ii1 FULL MATCH (27. Qh6), Junior 10 delta = 2.34
TOPALOV 11ii1i1i FULL MATCH (33. Bxf7), Junior 10 delta = 2.01

TOPALOV 00oo1i1i MATCH      (26. Bxc5), Junior 10 delta = 0.62 
TOPALOV 11ii1ii1 FULL MATCH (19. h5), Junior 10 delta = 0.15
TOPALOV 0o0oo1i1      MATCH (20. g6), Junior 10 delta = 0.14  

TOPALOV 11ii1ii1 +0 FULL MATCH (29. Nd3), Junior 10 delta2 = -0.05 at depth 20
TOPALOV 11ii1ii0 PART MATCH (22. Be6+ possible match), Junior 10 delta2 = -0.04

TOPALOV 00oo1i10 PART MATCH (21. gxf7+ possible match), Junior 10 delta2 = -0.06
TOPALOV 00oo0oo0 NON-MATCH  (31. Rxd3), Junior 10 delta2 = -0.09     
TOPALOV 10oo0o0o PART MATCH (32. Be6), Junior 10 delta2 = -0.30
TOPALOV 00oo0o0o NON-MATCH  (25. Bf5), Junior 10 delta2 <= -0.46 
TOPALOV 00oo0oo0 NON-MATCH  (24. Rhg1), Junior 10 delta2 = -0.58 
TOPALOV 00oo0oo0 NON-MATCH  (23. b3), Junior 10 delta2 <= -1.20

Only 3 matches in the 0--2.00 range.  This looks pretty random.
Of course, we need much more data to establish "random" empirically...

And in terms of an "information gained" metric, this run on Junior might
even give a negative number!  That is, it would conclude that Deep Junior's
programmers had more to learn from Topalov's play than vice-versa.
A negative number wouldn't prove absence of cheating---and would say little
or nothing about "key move" cheating---but it would argue that the 
information gained from the matches was worth little.

The two matches 19.h5 and 20.g6 are significant (but note that they can
be argued as dependent, "part of a strategy"), and 20.g6 would be 
identified as a "swing move" (in a "complex position" in Guid-Bratko's
terms), as (surprisingly to me!) would 26.Bxc5!  But that's it---the
other matches have no "swing" and are pretty obvious, I can say as an IM.
Thus even this /single run/ CAN say that the only meaningful "key move"
cheating would have been at moves 19--20, which were not mentioned
in the article!

The match rates to Fritz 10 (see other results file) and my own Fritz 9
were more significant.