How many matrixs did you recommend for task 2?

[ebeebs]

Originally Posted by IIRC

I don't like VaR because it does not include the tail risk...is that pretty much hitting the nail on the head? Seems simple, but I'm asking. Thanks.

Right, if your worst 5 scenarios are -100, -110000,-140000,-150000,-200000, and your VaR value is -100, it really misses everything below that and doesn't show how bad it can really be. Obviously this is an extreme example, but the point is the same.

[IIRC]

Originally Posted by ebeebs

Right, if your worst 5 scenarios are -100, -110000,-140000,-150000,-200000, and your VaR value is -100, it really misses everything below that and doesn't show how bad it can really be. Obviously this is an extreme example, but the point is the same.

Thanks ebeebs. You have some good posts.

I've noticed the variance for the extreme good outcomes, i.e. top 5% is more than that of CTE 95...is there a talking point here? I can't think of one...other than the reward for landing in the top 5% is disproportionately greater than the risk of landing in the bottom 5%. But this sounds more like a supportive arguement for VaR, which I'm avoiding.

[wangwwzj]

wow, the discussion is useful.

[MJWood407]

Now that I have my MMR, I can ring in...

Maybe the real test here in using/not using VaR is whether or not you as an actuary have the ability to find a better option than the one suggested by your client. YOU are the professional and are being paid for your expect advice, even when it may differ from the one that's paying your check.

I suggest that you look at what VaR reports when there are a high percentage of scenarios that result in the mine not reopening. Is that a good risk metric (Yes METRIC, not MATRIX) in that situation?
Good luck...

[gametime50]

This may be answered somewhere else, but I am getting mixed messages on whether we are supposed to include actual numbers for the risk limits (e.g. CTE(80) > -40, Mean > = 20) or just list that we would use CTE(80) and the mean as metrics. Then once we analyze the data more, we can come up with reasonable limits for this information. In the control cycle, we typically evaluate the risks and tail scenarios before setting the limits.

Any thoughts?

[ebeebs]

Originally Posted by gametime50

This may be answered somewhere else, but I am getting mixed messages on whether we are supposed to include actual numbers for the risk limits (e.g. CTE(80) > -40, Mean > = 20) or just list that we would use CTE(80) and the mean as metrics. Then once we analyze the data more, we can come up with reasonable limits for this information. In the control cycle, we typically evaluate the risks and tail scenarios before setting the limits.

Any thoughts?

Don't need to include numbers for Task 2. In fact I don't think you ever really need to include something like "only use if mean > x" even in the other tasks.

For 2 you pretty much list the metrics you want to use, give a little description of what they are in english, then say why you are going to use it.

[gametime50]

Originally Posted by ebeebs

Don't need to include numbers for Task 2. In fact I don't think you ever really need to include something like "only use if mean > x" even in the other tasks.

For 2 you pretty much list the metrics you want to use, give a little description of what they are in english, then say why you are going to use it.

Thanks for the advice...one question though.

How do the metrics impact any decision, without giving them values?

In other words, how do I choose between a set of 100 scenarios, if I don't have prescribed $ risk limits and $ expectations of profit, etc.

[komorgan]

Originally Posted by Actuarial007

Hi,

How many risk measurements would you recommend for task 2? I recommended expected ending cash, % ending <0, VaR95 and CTE95 on my first attempt, but failed. Is that too many?

Thanks for your help.

Did you talk about all the metrics that are available in the workbook or just the ones that you recommended?

[rolenkid2000]

Did anyone use mean and standard deviation to calculate coefficient of variation?? I then used this to initially rank my scenario results before looking at CTE(95). I didn't do this until my 4th attempt but I think it made selecting the ultimate combination of assays and minimum to mine much easier.

[ebeebs]

Originally Posted by gametime50

Thanks for the advice...one question though.

How do the metrics impact any decision, without giving them values?

In other words, how do I choose between a set of 100 scenarios, if I don't have prescribed $ risk limits and $ expectations of profit, etc.

It's all a balancing act, with explaining your picks. In task 2 you just say what you want to use and get a little feel for the tool. You don't really have the information to be making cutoffs, you just want to say what you're looking for (a high mean, for example).

When actually picking your MTM and assay number, you need to take in all your metrics into consideration. Sure you'll narrow them down, but I never said "I'm only accepting a mean above level X." You may pick a combination with a higher mean while taking a little worse CTE, or a high %end<0. Just explain how you balanced each metric to make them all in a reasonable range.

Graphs are great on Task 3 to illustrate how you pick your final MTM and assays. I recommend a line graph for each metric with the metric on the y axis, # assays on the x, and 4-6 lines of different colors for different MTM amounts.

Then you can reference your graphs and directly compare different combinations.