Fun with the game TinyRails

I play TinyRails, and I thought it would be fun to write a program to help me figure out the optimal train configuration for passengers. This is my attempt.

1. LINQPad v7 - the free version should work fine
2. tiny_rais.csv - CSV file with a list of train cars.

tiny_rails.csv is the main input to the script. There is a sample in the repository. The file describe the train cars to be considered when picking the optimal set. The file has the following columns in order from left to right. This is tedious to create and maintain. The Tiny Rails wiki has a more complete list, but not in a format that this script can use.

NOTE: all values are integers. TinyRails reports floats and integers, but only integers are used here.

The Notes column is parsed by the script to adjust any bonuses, but the parser is dumb. It specifically looks for columns with values like +5 Entertainment. If the value is like +4 Food when equipped on Tuesday it is ignored.

Brute force.

The script tries every combination of cars listed in tiny_rails.csv and some user values The following controls are exposed, and need to be set.

1. PASSENGER_COUNT - set this to the number of passengers your train supports without any cars equipped.
2. CAR_SET_COUNT - set this to the number of train cars to consider.
3. COMBINATIONS_PER_SEC (optional) - estimate of combinations per second this computer can execute. Varies by machine. Execute the script a few times, and pick an average. This script will print out an expected end time.

This is a brute force algorithm. Factorials are used when considering how many different combinations that need to be considered.

$\left(\! \begin{array}{c} n \\ r \end{array} \!\right) = \frac{n!}{r!(n-r)!}$

The number of combinations to explore quickly becomes untenable. For example, on my machine I can evaluate 2,750,000 combinations per second (roughly). Lets plug-and-chug (ha!) using the above formula to figure out how long certain values for n and r would take.

These numbers were randomly picked from an Excel spreadsheet to illustrate the point that this is a brute force algorithm. The CAR_SET_COUNT has an outsized impact on the number combinations to evaluate. Pick reasonable values if you want answers in a reasonable amount of time.

CAR_COUNT is determined based on the CSV file.

1. Type must be Car.
2. Include must be 1.
3. CAR_COUNT plus the value of the Quantity column. If the quantity is 3, then CAR_COUNT is plus 3.

I recommend you only include high-value cars. Avoid cars like Cargo cars, low-level, or cars with mediocre stats, e.g. Old West Passenger Car.

The bonus percentage is based on the number of passengers and the four attributes: food, comfort, entertainment, and facilities. The goal is to maximize all four of these values. If the train can support 50 passengers, the train needs 50 food, 50, comfort, 50 entertainment, and 50 facilities to get the maximum bonus.

Given a list of cars, the algorithm emits every combination of cars. Each combination is scored to determine the bonus based on the passenger count and four attributes. The cars selected will most likely increase the PASSENGER_COUNT. The algorithm considers bonus percentage not the dollar amount (another TODO).. The combination is enqueued into a priority queue. Once all of combinations have been tested, the top 10 combinations are dumped in order from highest to lowest bonus.

The priority queue tracks the top 1,000 combinations. It evaluates these 1,000 combinations and add up all of the cars that appeared in these winning combinations. The list is dumped in order from most selected to least selected. The intent is to make it easier to determine what cars to include or exclude in future runs. This will improve the execution time of the algorithm.

1. Engines and cabooses are not considered, only cars. Some engines maximize an attribute.
2. The algorithm maximize the bonus percentage, but it should maximize the dollar amount per passenger.
3. Combination generation is single threaded, but modern machines have multiple cores. Find a way to divide and conquer across multiple threads. (Seems easy-ish, but probably hard [very hard].)
4. Algorithm is way to brute force, what can be done to reduce the search space sooner.
5. Import the Tiny Rails cars from the wiki to make updates a little less painful.

Yes, but it is all in the name of fun! 😆🚂