Offer alternative algorithms for assigning platoons

[skiadas]

✔ Enhancements → Be specific. Suggest implementation when possible.

The current algorithm assigns a platoon to the person with the weakest unit for that platoon. This system can be unfair to those members with lowest gp for two reasons:

1. They will tend to be asked to fill in more platoons
2. They are more likely to need some of the units they are asked to platoon, necessitating the use of a separate exclusion spreadsheet.

I propose the following algorithm:

1. For each member, standardize the power values for their units (subtract the mean, divide by the standard deviation). Then very negative scores indicate units that are weak relative to the units on this member's roster (and hence less likely to be used by that member).
2. Assign platoons based on these standardized scores. This will tend to ask members to platoon units that are low within their rosters.

[PopGoesTheWza]

Interesting algorithm. I will study and simulate what is the impact. Note that there is also another enhancement about recommendation (#24 ) So I need to find how best address all need.

• best usage of missing units (at least 1 but less than total requested)
• priority on assigning rare units (no more than 3 more than requested)
• fair distribution of the rest of requested units according to each member roster.

[DarkPhoenixSniper]

@PopGoesTheWza so on the Breakdown tab, you could include the Light/Darkside character count, then also the Average there so can be referenced during the Platoon assignments instead of calculating it during the Assignment function.

[DarkPhoenixSniper]

@PopGoesTheWza maybe another hidden sheet for the Z scores.

I believe I have the arrayformula working, but need to store the values somewhere else. I'll keep you updated

[PopGoesTheWza]

@DarkPhoenixSniper i’m Working on a OO design to better address platoon recommendation.the current code is an terrifying function which loops over platoonsat least three times and is cluttered with specific tests an rules.

[DarkPhoenixSniper]

I need to investigate a little more... but the Algorithm assigned 3 of the bottom 4 of a character in a test. it chose number 5 instead of number 4.

But it might be because he filled 10 other spots where he had a weaker character....

[DarkPhoenixSniper]

@PopGoesTheWza informed me that the Z-score caluclation was not implemented in the latest build. He is still coding the solution.

[DarkPhoenixSniper]

Implemented a zScore option in the latest version of TB Sheet Version: 2.200409

[skiadas]

Oooh nice! I'll check it out.