### Table 4. Henry(TI =

"... In PAGE 7: ...Since all the results are inferior to 5 points, Table4 shows that Henry(TI = 2; 3; Pi = 0:4) and Henry(TI = 2; 3; Pi = 0:1) have approximately the same strength. We think that, provided Pi is low, its value has no great importance on move quality.... ..."

### Table 3. Henry(TI =

"... In PAGE 6: ...The results given in Table3 are slightly positive but, again, no certain sta- tistical conclusion can be drawn for most cells. However, by improving the refer- ence program by 6 points with a 10% time overhead, Henry(TI = 3; Pi = 0:4), can o er a good compromise between move quality and time.... ..."

### Table 1: Henry Classifying Modal

### Table 2. Table 2. Henry(TI =

"... In PAGE 6: ...Since most of the results are inferior to 5 points, no conclusion can be drawn from Table2 with su cient statistical con dence. The results of Henry(TI; Pi = 0:4) vs Henry(TI = 1; Pi = 0:4) are given in Table 3.... ..."

### Table 8. Results for Li and Henry data set neural network performance

"... In PAGE 15: ....2.2. Li and Henry data set The best network found for this data set used five hidden nodes, and was trained for 260 epochs (in steps of 10 epochs) ( Table8 ). There is some evidence that the network found was... ..."

### Table 6: HENRY(base=2, 2 lt;=R=11) vs HENRY(base=1, R=19).

"... In PAGE 4: ... Then, in a second stage, given that base=2, we aimed at determining the best value for R, performing 400 games per confrontation again. Table6 yields the results. ... In PAGE 4: ...Table 6: HENRY(base=2, 2 lt;=R=11) vs HENRY(base=1, R=19). Due to a different number of games per confrontation, the results given in Table6 are more accurate than the results given in Table 5. The result obtained by HENRY(base=2, R=4) is amazing because it shows a discontinuity that we are not able to explain.... ..."

### Table 4. Performance of regression models on testing data for Li and Henry

"... In PAGE 12: ... Whilst the MMRE indicates some modeling problems over the data set as a whole, the Pred(l) and AR indicators suggest that performance for specific observations is reasonably sound. The gains from robust regression are fairly small, but nonetheless robust regression could be seen as slightly better performing ( Table4... ..."

### Table 6. Henry(B = 2;2 lt;= R lt;= 9) vs Henry(B = 1;R = 19).

"... In PAGE 7: ... 5 7 9 11 13 2 +9 +8 +7 +4 0 5 0 +4 -1 -3 +3 10 -6 -5 -4 -10 -8 Then, in a second stage with B = 2, we aimed at determining more precisely the best value of R, performing 400 games per confrontation again. Table6 yields the results. Due to a di erent number of games per confrontation, the results given in Table 6 are more accurate than the results given in Table 5.... In PAGE 7: ... Table 6 yields the results. Due to a di erent number of games per confrontation, the results given in Table6 are more accurate than the results given in Table 5. The (B = 2; R = 4) result is amazing.... ..."

### Table 2: HENRY(TI=2,3,4, Pi=0.1) vs HENRY(TI=1, Pi=0.1).

"... In PAGE 3: ...ests. The results of HENRY(TI=n, Pi=0.1) vs HENRY(TI=1, Pi=0.1) are shown in Table2 . The results of HENRY(TI=n, Pi=0.... In PAGE 3: ...1). Since most of the results are inferior to 5 points, no conclusion can be drawn from Table2 with sufficient statistical confidence. 2 3 4 9x9 +3 +2.... ..."

### Table 3: HENRY(TI=2,3,4, Pi=0.4) vs HENRY(TI=1, Pi=0.4).

"... In PAGE 3: ...4) vs HENRY(TI=1, Pi=0.4) are given in Table3 . The results of HENRY(TI=n, Pi=0.... In PAGE 3: ...4). The results given in Table3 are slightly positive but no certain statistical conclusion can be drawn for most cells. However, by improving the reference program by 6 points with a 10% time overhead, HENRY(TI=3, Pi=0.... ..."