Apparently impossible must get to the maximum, unless some unorthodox methods, such as to destroy all of 9 layer to see the diamond, and then take the last layer.

Logically, you can find a larger,

For example, the size of the diamond on each layer is recorded,

Then calculate the average value of the size of the known diamond is considered as the overall average,

Calculate the average of adjacent size diamond the size of the gap, and then by the number of data and the overall number of calculated the difference between the adjacent to the overall size of diamond, each floor are recorded and calculated so these prediction value will become more and more close to the actual value)

It can be inferred that the maximum size of the diamond (average +4.5* average difference). In order to reduce the risk, to see the diamond > = predictive value -1* average difference when it thinks it is the biggest.

Here is that the diamond the size of the arithmetic progression distribution, diamond size under natural conditions should is similar to the normal distribution, but I think since this is the artificial test should choose some of the gap between the more obvious distribution more uniform diamond. I think this more ground in arithmetic progression of distribution, of course, can also be from the front to the data to analyze closer to what the distribution model, but this amount of data is relatively small, this is not quite accurate.