Statistical analysis in MouseHunt

From MHWiki
Revision as of 17:21, 7 September 2009 by Cbbrowne (talk | contribs) (Basic Theory)

Statistical analysis has become a pastime among many mathematically inclined regulars on the MouseHunt forums as a way of determining which trap/base/cheese combination works best in certain areas, determining how much success a hunter should expect to have when using a certain combination in a certain area, and determining a law of averages that sets proper expectations for hunting that smooths out the instances of "good luck" and "bad luck".

Basic Theory

The underlying assumption of statistical analysis in MouseHunt is that the game's behaviours are ultimately deterministic, that the game can be predicted based on statistical probabilities rather than the probabilities varying based on "luck," or "I just changed my trap," so it is therefore logical for the hunter to make decisions that maximize his odds of attaining goal (such as gaining gold/points, or trapping that rare mouse) as efficiently as possible.

Analysis allows the hunter to make informed statements about the effectiveness of his trap/base/cheese setup.

Rather than making an exaggerated and almost certainly false statement such as, "I'm having a really crappy day in the Mousoleum. All of my cheese is being eaten and I'm not catching anything", the analytical hunter can say, "I'm having a pretty cruddy day in the Mousoleum. I'm only averaging 543 gold per cheese and 712 points per cheese, today, which is less than we'd expect. However, I was racking up 1379 GPC and 1686 PPC earlier in the week, which was better than expected, so I suppose things are evening out now."

The analytical hunter aims to tabulate data over the long term rather than just a few hunts in order to make informed opinions on setups, areas, and strategy that are not biased by brief runs of apparent good or bad luck.

Popular Statistics and Terms

Along with the typical abbreviations used by most hunters (such as "DDB" for the Digby DrillBot), MH statistical analysis has its own vocabulary.

GPC, PPC - Gold per cheese and points per cheese.

  • Sally has 112,000 Gold when she enters the Mousoleum with 300 pieces of Radioactive Blue, and she has 471,000 gold when she runs out of cheese. Her GPC during her stay in the Mousoleum is 1196.66, i.e. (471000 - 112000) / 300 = 1196.66

GPH, PPH - Gold per hunt and points per hunt. Some hunters prefer to tabulate their statistics based on the number of hunts they've engaged in instead of the amount of cheese they've used.

  • Dwight has 890,000 points when he enters the Training Grounds. After 250 hunts, he has 1,100,000 points. His PPH for those hunts is 840, i.e. (1100000 - 890000) / 250 = 840

Run - the experiment or batch of hunts that the hunter is referring to, e.g. "I am doing a Lab DB/exp/Swiss(500) run right now" or "My latest run in the TG netted me 952 PPC".

Sample size - the amount of hunts or cheese used in order to determine the statistics that have been tabulated. Larger sample sizes, by default, result in more accurate statistics than smaller sample sizes.

Cost Analysis

Cost analysis is about as complicated as MH statistics can get, because it takes every single expense into account in order to determine whether or not a run (or series of runs) would be ultimately profitable.

A prime example of this is when a hunter determines whether or not to travel to a shop, buy a new trap, purchase a bunch of cheese, and then travel to a hunting location to resume hunting. Cost analysis aims to answer the following question: "Is spending my gold on travel cost A, new trap components B, and cheese C going to increase my profits over the long run, considering my plan to hunt in location D?" Needless to say, this type of analysis requires an extraordinary amount of data and a lot of math.