by andthen » Sun Jan 26, 2014 1:30 am
Hi Carbob and Moti
Back testing with estimates/predictions is not really possible, as they are dynamic and change from draw to to draw, using the previous draws to set their levels. So you can't run a set of estimates for this week's draw and expect it have to any meaning/relevance in the past.
History differences are determined each week for each level by the previous draw which pushes numbers through the levels. Each number has a difference, based on it's draw history. Each level is a draw. Therefore every new draw each level will change.
Whether a history level difference is negative or positive and the size of it (eg low -1 or high - over -30) is determined by the numbers drawn in L0. There are many tools you can use to estimate likely differences in each levels, one is the sum of differences of the segments for a level (for eg a level with total sum of differences of over 300 in my lottery is generally a 60% chance of being a negative). You can also look at the numbers per segment...if a level has mainly numbers in segment 1, and none in Segment 3 it is likely to be a positive level.
The only way to test a theory using what if simulation or any other method of predicting history differences is to hide a draw, enter your predictions and then unhide the draw and see how close you came.
The other factor that will have a bearing on your simulation results is the settings you have in your occurrence threshold, found in tools. i think the default setting is 7.5%-7.5% which i have found is too high for my lottery. the higher the OC, the narrower the simulation results will be.
A