why cdu has to be a lt investment, page-53

Mums and dads don't understand resource estimation
neither does the state, for that matter.
There is way too much room for obfuscation within the jorc guidelines.
How do two supposedly reputable companies give such wildly different estimates from more or less the same samples?
A clue is given by the disclosed inputs (other than samples which would be more or less the same less any subsequent drilling in the first one)
The first real jorc (the second resource statement) uses a search area which intuitively rings true to the companies own 'geological' description of the orebody, of rapid changes over short distances. The search area is 25-50m or something. The method is also tailor made to deal with unusually high, sporadic values and also offers probabilistic support to what is likely to be mined between samples.
the SP tanks because the result isn't up to expectations.
So what do they do?
They select an averaging method which uses a search area of up to 300m in one direction.
They also decide to contain the estimate within zones they have constructed. These are subsequently shown to be at least partially wrong, by them digging up mineralization where they didn't think there was any. They admit (either knowingly or unknowingly) that this method is inappropriate because of their claims that local nuggets are their big issue. Underestimate or not, the other method would deal with this problem better.
This algorithm they are now using is a linear weighted averaging one, and coupled with this vastly larger search area, this method will use samples from up to 300m away to influence local estimates. They already should know not to do this because they ’know’ the character of the orebody enough, from all the drilling they have done and the prior estimate. What this latest method does is smooth higher grade from afar into local block estimates, to give a flawed, but 'better' sounding result. The result overall result is biased to the high side and any half arsed statistician could prove that to be the case. Furthermore, an averaging algorithm, squeezed into narrow constructed zones , is no longer an average at the scale of the resultant block model, so the algorithm is being misused… It doesn't take a rocket scientist to understand that.
The co. have found an old school tie from down the gold coast way who has signed off on such a method, despite the result of the prior one and despite the companies own claims of ‘understanding’ their orebody. They have admitted as much in their supporting documentation (probably without realizing it) by disclosing their parameters and methods as required and then making completely discordant speculative claims about underestimation, whilst using a method which amplifies the stated issue.