Mousehunt Analytics

The Art of Hunting

The art of hunting is keep learning and improving your strategy. First, consider what your goal is. Gold? Point? Catch a specific mouse? Collect certain loots? Or combination of some of the aforementioned goals?

Second, realize that there are 5 parameters in the game: location, cheese, base, weapon, time. That's all. You need to know where to hunt,  how to attract and how to catch. These are three main questions applicable to all goals. If you want to catch a specific mouse, you might need to know when to hunt. If you want to collect certain loots (to access new areas or to craft something), you might need to know which mouse drops the loots.  The five parameters are your decision.

Lastly, you need to optimize. Collect stats and analyze them purposefully so that you take appropriate actions. This lesson, Mousehunt Analytics 101, will address key points for you to start thinking scientifically and strategically about mousehunting.

Mousehunt Strategy Process

  1. Set your goal
  2. Test your parameters
  3. Analyze the results
  4. Take action

Is statistics safe to use?

You can use statistics because the dynamics behind Mousehunt is probability. One developer mentioned this in Mousehunt Staging Forum:

The sequence of events involved in going on a hunt and determining the results is now complete in its first form and ready for more thorough testing.

For clarity here is the general break down of dice rolls that determine your success in a hunt. Obviously there are a few more variables in the equation such as your hunter's title effecting your luck against the tougher mice, etc.

Sequence:
  • Based on your location you encounter a mouse.
  • Based on the cheese you are using and your trap's accuracy you attract or fail to attract the mouse you encountered.
  • If the mouse was attracted an equation takes into account a variety of factors such as the power of your trap to determine the probability if your trap being successful. A roll is then performed based on that probability. This equation makes no mouse impossible to catch.
  • Based on where you were hunting and the type of mouse another roll is preformed to determine if you receive loot and to what quantity.

However, you should be aware of statistical errors as you are dealing with many "dice roll" random variables. It is possible that you toss ten coins and all turn out  to be head (though very rare). Moreover, you are trying to use statistics from your sample data to make inferences about an aspect of the game. For instance, if your sample data says that gold per hunt is 550 whereas, in the long run, according to the developer's dice and probability theory, you should earn in average 520 gold per hunt, then 30 gold per hunt is a margin of error (Wikipedia)

It is important to note that you will never know the 520 figure, but calculating margin of error is possible and introduced in the Mousehunt Log Summarizer in a format of "confidence intervals".

About Confidence Intervals

The Mousehunt Log Summarizer calculates 95% confidence intervals. This means that there is a 95% chance that the (unknown) parameter (520 figure) falls in the confidence interval.

As you gather more hunting data, your experimental value should move closer to the true value of 520, and the confidence interval should become smaller, reflecting a greater chance of getting a more accurate estimate in a long run.  In a short run, a confidence interval could be larger as you add more data that enlarge the standard deviation of sample data.

Let's assume that you catch 95 mice out of 100 attractions. You can be quite confident about your 95% catch rate. Unfortunately, shortly after that you miss 10 mice from 10 consecutive attractions. Now you are less sure about your catch rate. C.I. can reflect that by showing a larger confidence interval.

Standard deviation of sample data (the extent to which sample data is deviated from the sample estimate), in theory, should be constant in a long run. If your setup in a certain location gives a 70% attraction rate over 5,000 hunts, it should also give a 70% attraction rate over the next 5,000 hunts.

More Readings


If you are mathematically inclined and interested in technical details about statistical  inference, please read the following wikipedia articles:
Statistical Inference - is the concept of estimating a parameter (520 figure). I use interval estimation.
Central Limit Theorem - says that sample data is normally distributed. Thus, we have to have some, but not total, confidence in estimation. A 100% confidence interval is infinite (there is no chance that the parameter falls outside the confidence interval) and therefore useless for estimation.
Confidence Intervals -
Confidence Intervals for Proportions - show the formula used to calculate C.I. for catch rates and attraction rates. Those rates are hunt/hunt proportions. (see glossory).

The Mousehunt Log Summarizer by Pooflinger

This tool will do the math part for you. Your part is translating the stats into action.

Here is the link to the program > Mousehunt Log Summarizer

Save your hunter journals in local text files

This is an interim solution. Save them in Notepad (please don't wrap text) in separate files for separate trap/cheese/location setups. Be careful to exclude hunts from events such as giveaways or halloween because you may be facing a different mix of mice and this does not answer your long-term objectives.

Before Using the Metrics

How to use of Metrics

You will NOT try to maximize any single metric blindly.  You will need to balance some or all these metrics to achieve your goal(s). You can benchmark your metrics with fellow mousehunters with similar goal(s).

A metric should be able to tell you to "do something". Generally speaking, we are trying to optimize something we can control (directly or indirectly). I personally believe that the King's rewards is independent and purely random; this is to say that the King's rewards cannot be influenced by the five parameters; King's rewards are therefore irrational to be included in a statistical workout, even in per cheese figures calculated by the log summarizer.

Loot, on the other hand, is something we can control indirectly. For example, you can change location (one of the five parameters) in order to collect certain loot (such as map pieces or potions).

Per Hunt VS Per Cheese

Per Hunt = per event that you are trying to attract a mouse either by the horn being sounded or hourly trap checks. There are three possible outcomes: (1) failed to attract a mouse, (2) did attract a mouse but failed to catch it, and (3) did catch a mouse.
Per Cheese = per piece of cheese used as a result of hunting. Not a result of cheese trades or King's rewards.
Per Catch = per hunt that results in a successful catch.

In economic sense, per hunt figures are much more meaningful than per cheese ones. This is because the number of hunts is more time-based than chance-based and offers greater comparability. Why?

Suppose that you have two choices:

  1. Cheese effect: extremely fresh; attraction rate: extremely low; catch rate: medium
  2. Cheese effect: extremely stale; attraction rate: extremely high; catch rate: medium

We can expect that the two choices result in similar profit per cheese and point per cheese (because, in both choices, cheese very rarely goes stale). Choice (1) is wasting time, but does not waste cheese. Remember that time is an opportunity cost. Taking in account "lost opportunity" (though you don't lose any cheese) is perfectly rational. 

In summary, choice (2) is obviously a better choice of strategy, which can be suggested by per hunt figures, not per cheese ones! 

Some Metrics

Attraction rate

Attraction rate = number of attractions / number of hunts

Why does it matter?

  • Opportunity cost. If you cannot attract anything, you will not catch anything. You are losing time.
  • Cheese cost. If you cannot attract, cheese might go stale. 

How to improve?

  • Use cheese that is generally more attractive (usually more expensive). For example, upgrade from cheddar to marble.
  • Use cheese that is more attractive to mice in the location you are hunting. For example, radioactive blue cheese in the Mousoleum.
  • Use cheese that is generally more attractive to the mice you wish to catch. For example, zombies are said to be more attracted to swiss cheese than to brie cheese. Lycans are more attracted to moon cheese than to radioactive blue (rumors say that lycans are more seen in nighttime, so someone expects that using moon cheese in night time would maximize your chance to attract Lycans).
  • Try a trap that has more accuracy bonus. For example, change from explosive base to target base.

Catch rate

Catch rate = number of catches / number of attractions

Why does it matter?

  • Profit. If you can attract but cannot catch, you will lose your bait cheese for sure. And in certain locations, you run a risk that the mouse might steal extra pieces of cheese, gold or even points from you.
  • Opportunity cost. You might miss a mouse or loot you are looking for (usually rare ones, e.g. potions). If this is the case, catch rate is very important: a low catch rate will make the chance to reach your goal even slimmer.

How to improve?

  • Use a trap that is more powerful. This is a general rule for common mice. A Trebuchet is better than a Sticky Glue trap in the town of Gnawnia, for example.
  • Use a trap that has a better luck factor. This is particularly useful against overpowering mice such as zombies. Swiss Army trap is said to be more successful in catching zombies in the lab than regular Deathbot.
  • Use a trap type that is more effective in a particular location (especially the Mousoleum). For example, shadow type traps are more effective than physical type ones in catching ghostly mice.

Cheese per hunt

Cheese cost per hunt = total cost of cheese (baited or stolen) / number of hunts
Cheese used  per hunt = total cheese used (baited or stolen) / number of hunts

Why does it matter? 

  • If your cheese and trap's accuracy bonus combined is not 100% attraction rate, then your cheese might go stale.
  • In certain locations, mice might steal your cheese.

How is that useful?

  • Based on amount of gold you have at the time of buying cheese, cheese cost per hunt can help you estimate the number of days you stay in a location that does not sell cheese without traveling back to buy cheese again.
    • For example, you have 30900 gold; traveling cost = 900 gold;  cheese cost per hunt = 300 gold; number of hunts per day = 50. Then you can expect to stay there for (30900-900)/(300*50) = 2 days.
  • Cheese used per hunt can strengthen your understanding of Mousehunt Analytics.
    • Low cheese used per hunt means that your attraction rate is low but cheese rarely goes stale.
    • High cheese used per hunt means that your catch rate is low and many mice steal multiple pieces of cheese. In locations where mice do not steal, cheese used per hunt is never greater than 1.

FAQS

How is total power of the trap calculated?

Please refer to Mousehunt Glossary.
There are power and power bonus factors from both the base and the weapon.

How does catch rate work? I have a dilemma between Luck and Total power of the trap.

Once the mouse you encounter is determined, given there are 3 behind-the-scene steps to determine the catch rate:
A) Trap type (physical/shadow/tactical) vs Mouse type.
B) Total power of the trap is compared with the mouse's strength to determine the catch rate.
C) Luck determines the catch rate by its own regardless of the mouse's strength.

Because steps B and C are statistically independent (wiki), it does not matter which step comes first and which step comes after. Step A remains a mystery about its sequence and whether or not it is independent.

Catch rate chart
 Trap \ Mouse  Weak
 Strong  Overpowering
 Low power, low luck  decent (pp, l)
 dreadful (p, ll)
 dreadful (p, ll)
 Low power, high luck  good (pp, ll)
 poor (p, lll)
 poor (p, lll)
 High power, low luck  very good (ppp, l)
 decent (pp, l)
 dreadful (p, ll)
 High power, high luck  very good (ppp, ll)
 good (pp, ll)
 poor (p, lll)
p = power has a little role in determining catch rate
pp = power has a considerable role in determining catch rate
ppp = power has a significant role in determining catch rate
l = luck has a little role in determining catch rate
ll = luck has a considerable role in determining catch rate
lll = luck has a significant role in determining catch rate

What does the chart say?
  • High power can comfort you in catching mice that are in par with, or weaker than, the trap. But it helps very, very little in catching mice that are far more powerful than your trap.
  • High luck is necessary to catch mice that are far more powerful than your trap.
  • Depending on the location and your objective, you need to decide carefully when choosing between alternative traps.
    • Your objective: to catch a zombie (overpowering) in the lab to get the Mousoleum cloth. You need luck.
    • Your objective: to maximize gold in the meadow (mostly, weak mice). You need power.
You don't need to believe this, as it is from the "input" point of view. Use different traps, see your hunter's journal ("output" of the five parameters) and compare stats to confirm your understanding. Mousehunt Log Summarizer is my recommended tool for you.

Is gold per hunt always lower than gold per cheese? Point per hunt always lower than point per cheese?

This is a common misperception. You should read the cheese per hunt metric again if you are still in doubt. If that does not help, then let me explain a bit further. It is possible that cheese used per hunt is higher than 1 in these scenarios:

  • you are hunting in locations where mice steal extra cheese in addition to bait cheese
  • attraction rate is very low and/or the cheese effect is very negative.
That's why sometimes gold per hunt can be higher than gold per cheese and point per hunt can be higher than point per cheese.

Is gold per hunt always lower than gold per catch? Point per hunt always lower than point per catch?

Almost "Yes" to both questions. All catches are qualified as hunts. If you catch a mouse in every hunt, then per-catch, per-cheese and per-hunt figures will be the same. Otherwise, per-hunt is lower than per-catch figures.

Can't find what you want?

This article will be updated from time to time.
Feel free to Visit this thread to post questions and discuss with fellow hunters. If your comment is published to this article, you will win some Super Brie from me as a thank you.
Comments