This is stats Sydney Metro <$10 circa 2004 which only confirms that barriers do not matter with quality horses. When you add all horses it halves the strike rate. The spike in the ROI on the outer B10+ can only be from odds.
Barrier Wins Starts StrikeRate ROI
1 230 1262 18.23% 88.75%
2 241 1291 18.67% 80.70%
3 256 1306 19.60% 87.43%
4 241 1301 18.52% 82.98%
5 237 1253 18.91% 85.17%
6 204 1189 17.16% 79.99%
7 194 1125 17.24% 82.14%
8 158 948 16.67% 81.81%
9 143 820 17.44% 89.07%
10 108 567 19.05% 101.99%
11 65 378 17.20% 92.99%
12 46 249 18.47% 110.29%
13 22 111 19.82% 121.74%
14 7 65 10.77% 68.46%
15 4 13 30.77% 209.62%
16 2 12 16.67% 112.50%
17 0 3 0.00% 0.00%
18 0 3 0.00% 0.00%

Having said this there is definate bias at tracks at specific distances and rail positions. In UK Racing I read somewhere that you can get a book with all the bias data. Some trainers scratch purely on drawing certain barriers. From memory some tracks were the outer, others the inner and some had the middle barriers. Beton