Table 1.

Averages of log scores (left) and averages of their ranks (middle) from San Juan predictive distributions (see Figure 8) over the forecasting weeks. These have been negated so that lower average scores (and ranks) are best. The right table shows mean absolute errors (MAE). The alphabetic labels A–P are the anonymized names from the CSV file provided by the CDC at the end of the contest period. Our hybrid GP/GLM was comparator “E” in that file. Superscripts denote the top three in each column

San Juan Method	Average Score			Average Ranked Score			MAE
	Peak	Week	Season	Peak	Week	Season	Peak	Week	Season
A	Inf	Inf	Inf	15.08	16.19	14.15	221.83	8.15	4965.6
B	1.54	2.77	1.67	9.92	6.46	35.62	50.65	7.81	1009.6
C	1.24	3.11	1.69	5.85	8.00	6.23	42.96	6.83	989.6
D	1.48	Inf	3.29	5.73	14.73	9.23	45.42	9.56	1306.2
F	Inf	Inf	Inf	14.88	12.38	15.96	59.51	35.71	982.0
G	1.08	3.21	21.51	6.92	9.00	25.00	326.61	25.01	723.5
H	1.55	3.55	2.30	8.38	8.62	8.92	49.76	5.81	1117.4
I	Inf	Inf	Inf	17.15	15.46	16.85	42.52	7.29	949.4
J	1.34	3.20	31.59	8.15	8.69	5.77	34.48	5.87	939.1
K	Inf	Inf	Inf	11.54	12.85	14.88	49.06	6.06	1151.7
L	Inf	Inf	Inf	14.46	16.19	16.85	54.12	15.69	1361.3
M	5.17	Inf	7.66	14.31	16.19	13.69	86.75	10.25	1915.0
N	Inf	4.15	1.93	15.08	8.92	7.00	273.21	8.60	1048.5
O	1.52	32.56	1.87	7.23	34.38	6.00	124.96	5.89	2712.1
P	Inf	Inf	Inf	12.23	14.62	15.77	68.00	8.98	1166.0
hybrid/E	10.91	22.55	1.96	35.38	5.08	6.85	32.37	5.88	826.8
hetGP	30.94	11.91	11.38	14.08	11.92	13.00	225.86	14.25	1667.2
GLM	20.92	2.75	2.14	25.15	6.15	6.69	28.64	6.21	3803.7
baseline	1.41	2.54	Inf	8.46	24.15	11.54	55.45	6.83	845.0