1 reply to this topic

[HRR Insanity]

Using your existing database of games (since the last patch), calculate an Elo score for all 'Mech variants.

For each game, assume each 'Mech variant is a 'player' and proceed through all of the games calculating Elo score for each variant. At the end of the database, you will have a relative 'effectiveness' score for each 'Mech variant.

You can then use this 'Mech Elo to help balance matchmaking.

This will minimize the need for a weight class based Elo which is potentially flawed. Example: Players who only play a RVN-3L and win a lot will be over-rated when they drive another less effective chassis/variant (like the RVN-4X or the SDR-5K).

PS: You could also extend this concept to weapons and equipment to estimate the relative benefits of having ECM, AMS, or any of the weapons on a 'Mech. Boated items (6xPPC) would proportionally be granted a 'win' multiplier because each weapon would be counted as individual even if a single 'Mech all had the weapons.

Edited by HRR Insanity, 21 February 2013 - 08:27 PM.

[HRR Insanity]

HRR Insanity, on 21 February 2013 - 08:03 PM, said:

Using your existing database of games (since the last patch), calculate an Elo score for all 'Mech variants.

For each game, assume each 'Mech variant is a 'player' and proceed through all of the games calculating Elo score for each variant. At the end of the database, you will have a relative 'effectiveness' score for each 'Mech variant.

You can then use this 'Mech Elo to help balance matchmaking.

This will minimize the need for a weight class based Elo which is potentially flawed. Example: Players who only play a RVN-3L and win a lot will be over-rated when they drive another less effective chassis/variant (like the RVN-4X or the SDR-5K).

PS: You could also extend this concept to weapons and equipment to estimate the relative benefits of having ECM, AMS, or any of the weapons on a 'Mech. Boated items (6xPPC) would proportionally be granted a 'win' multiplier because each weapon would be counted as individual even if a single 'Mech all had the weapons.

Hmph. I guess, when used together, this approach could be used dynamically to calculate a "real world" BV. Possibly use the 'Mech and weapons on it to calculate a scalar that could be applied to a given individual's overall Elo ranking.