The parameterisation of the metatstic cancer network given in Fig 2.7 (and repeated in Fig 3.7) is incorrect. Using Pearl's 1998 text (p 196-197) as the source, the CPTs for nodes S and B have been reversed (no doubt due to the node renaming from A,B,C,D,E to M,S,B,C,H!). Also, in our text we have P(M)=0.9, whereas in Pearl's version, the prior for metastatic cancer was 0.2.

The parameterisation of this clustered BN in Figure 3.8 is incorrect.

• P(Z=FT|M=T)=0.04 (not 0.16 as printed).
• P(C=T|Z=TF)=0.8 (not 0.6 as printed. The 0.6 comes from the Neapolitan, not Pearl version).

A number of entries in Table 4.2 are incorrect.

• When Therm = F , Bel(FeverLater=T) = 0.018 (not 0.525 as printed),
  EU(TA=yes)=45.40 (not 45.41 as printed).
• When Therm = T, EU(TA=yes)=44.12 (not 44.1 as printed).
• When Therm = T and Reaction = T, Bel(FeverLater=T) = 0.033 (not 0.273 as printed).
  The reason behind this change in belief is not that having a reaction reduces fever, but rather that the observation of a false reaction implies that aspirin has been taken, hence the chance of having a fever later is much reduced.

Table 9.4 entries for the poor qualitative outcomes should be -20, not -5 as printed. This corrected utitity function was used to compute the expected utilities in Table 9.5.

Figure 5.5 has the arcs from BPP_Win to OPP_Final and BPP_Final. The direction of both arcs are incorrect -- BPP_Win is a deterministic function of the final hands of both players.