Friday, May 11, 2012

Are there multiple levels or tiers of Conserve MP?

I was recently looking into what exactly the proc rate is for the "Augments 'Conserve MP'" property of the full Goetia attire +2 set (5/5), which was estimated indirectly last year (a thorough overview of the set bonus, including the relationship between damage multiplier and the fraction of MP conserved, is presented at The Black Mage's Apprentice in two posts). This was done by estimating the effective rate of observed procs, or the product of the actual augment proc rate (which itself cannot be observed directly when the Conserve MP rate is not 100%) and the proc rate of BLM's Conserve MP trait, which is supposed to be 25%. If you believe the set proc rate is a multiple of 5, then 15% and 20% are plausible estimates.

While some of the reasonable assumptions underlying the analysis caught my interest, such as whether the occurrences of Conserve MP, resists, magic critical hits and the Goetia set bonus are all actually mutually independent (kind of similar to whether the Double Attack trait procs on two hits of a multi-hit WS; no one actively looks for these rare events so you have to use more clever methods of providing evidence of such), of greater interest was the suggestion that there are multiple "tier"s (or "ranks" or "levels") of Conserve MP.

I just checked wiki.ffo.jp for the Conserve MP article and found a claim of at least one more Conserve MP "rank" at level 76 for BLM and level 96 for SCH. The claimed Conserve MP rate was 33%, but I couldn't find any evidence supporting this. (But there must have been some reason for this to be posted?)

I then tried to look for evidence of another tier  on the BG forums. I totally forgot I already looked at the evidence for this last year (posts #154 and #155...), and test 1 showed that the estimated Conserve MP rate for BLM was well below the claimed 33% rate. (Due to the use of a non-MVUE for the conserve MP rate, which is a binomial proportion, for the sake of convenience and sanity, the nominal 95% normal approximation confidence interval based on the binomial proportion but plugging in the non-MVUE actually has a lower coverage probability, or 91% as I claimed previously...).   This was before the September 2011 version update raised the maximum job level to 95, so this doesn't rule out any new levels being introduced since then.

So as long ago as June 2011 there did not appear to be evidence of multiple tiers of Conserve MP. If this is to be checked again in the future, the data collection procedure as described on the BG forums (see the previous link) would be one way to amass evidence for or against multiple tiers.