#5 - Testing the attributes

Since my last update, i have implemented most of the attributes in Cycling - The board game.

Quick recap: the game is about getting to the finish line first. Each turn, each player rolls a dice (D6) and moves the amount of squares. The odds of the dice are rigged based on the player's strength (higher strength = better odds). Next to the dice roll the player can gain bonuses to move additional squares. These bonuses are based on their attributes (which can range from 1 - 20).

Now that i have implemented them, i have to check if they add fun to the game. Fun is of course hard to measure, but what i am aiming for mostly is balance between the different elements. For example, if one attribute is much more important than others, the others are irrelevant.

To test them, i left my computer running for a night and simulate over 600 different races with different AI players (with every time different attributes and strength). I now had a lot of data points to analyse with.

I decided to use a multiple regression analysis, where i analysed the impact of the strength and all the different attributes on the outcome of the race. This led to the following results:

R-squared: 
0.43

Coefficient of different attributes

Strength	-4,59
Climbing	-3,60
SoloRiding	-3,06
GroupTechnique	-3,04
Flair	-1,57
Sprinting	-0,93
Restoration	-0,84
Starting Position	-0,73
Determination	-0,51
Composure	-0,36
Luck	0,12

Conclusions

• This stuff really makes me feel like a nerd, which is nice ;-)
• The R squared means that more than half of the results are not explained by this model
• The fact that all attributes are negative is correct (the higher the attribute, the lower your total final time)
• The attribute luck has a different sign, which basically means that a lower attribute is better (although the coefficient is very small, so technically it means that it is not relevant)
• Strength is more important than all the different attributes
• Climbing, solo riding and group technique are most important
• Composure and determination hardly matter at all

I am somewhat happy with these outcomes, but find it difficult to determine my final goal here. I think it is not a good idea to have all attributes at (roughly) the same value, since then it won't matter which one you improve at all. 

On the other hand, an extremely large difference between attributes, means that your strategy should be to simply focus on the best attributes, which is also technically not really fun.

For now, my next move will be to make the lowest three attributes somewhat more influential so that they matter more. When that is done, i am just going to continue playtesting manually and get a feeling for how the game is running. At that point i will also upload a new version, so i can (hopefully) get some feedback on the game itself. 

In the meantime, please feel free to try out the game yourself (link) or even better, let me know what your ideas would be to further balance the game.

Thank you once again for reading!