I am posting this to further illustrate findings after I ran the GAT engine with the results from the latest (at that time) Canadian 6/49 run. I used all standard options with 13 numbers. I am attaching the following screen shots. I would like to hear the opinions.
The first two screen shots represent overall balance comparison at the beginning of the run, with just slightly over 100000 GATs and over 10 millions. White fields represent GATs with 0 numbers correctly predicted and dark green 4 or more numbers correctly predicted. To my, and probably everyone's surprise the difference is slightly in favour of the beginning of the run (Beginning 7 whites and 9 dark; over 10 millions 9 whites and 8 dark). Note that the reds are originally light green, which means1, 2 or 3 hits. However, the difference is statistically insignificant and I would say that, based on this sample, that there does not seem to be either advantage, or disadvantage in running very high number of GATs. Here is the situation at the beginning:
http://prntscr.com/kz55ij and when the number of GATs reaches 10 million
http://prntscr.com/kz54b4.
Below are three examples of success; one is from the group 6, only 45th in the rank, but achieved six correct numbers; one from the group 5 with 4 correct numbers; and finally one from the group 3 with 4 correct numbers. None were in the top 5, with the one from the group 6 being only 45th in the rank. That ranking, of course changed after this draw, they all advanced upward. Here are the data. From the column 6 numbers and hits
http://prntscr.com/kz5k0y, the same with respect to the statistics and time delay for the same GAT
http://prntscr.com/kz5kn2. Then, the GAT from the column 5 with respect to the numbers and delays
http://prntscr.com/kz5pfv, the same GAT with respect to the statistics and hits
http://prntscr.com/kz5qot.
Below are two examples of failures. Both are from the column 3. Example 1 hits and statistics
http://prntscr.com/kz62cr. The same GAT with respect to numbers and delays is here
http://prntscr.com/kz637i. Here is another example of failure. Hits and statistics are here
http://prntscr.com/kz64du and the numbers and delays here
http://prntscr.com/kz654e.
I believe that we need to develop some hints for choosing a good GAT and avoiding a bad one. Ranking, time delay, blue curve on hints slope may mean something, but may not be the only hints.