Pretender Cost Calculator

[Psientist]

---------------------------------------
Current posted version: v1.03
---------------------------------------

After continued discussions in my Mytheology mod thread, I decided to get serious about balancing my pretenders for multi-player. EndoPerez provided some *great* general guidelines for costing modded pretenders so they are appropriately balanced. However, I decided something even more consistent was called for. So, I ran a multiple regression on the attributes of about 85 of the default pretenders, and created a spreadsheet that provides a recommended cost for a pretender based upon the attributes the modder types in.

Advantages: It's guaranteed to be pretty close; the formula from the regression is significant to .000. That's pretty accurate, and means that your new pretenders will be balanced in comparison to as many of the default pretenders as possible.

Potential flaws: The spreadsheet is based upon a *linear* regression, so it probably will under-cost pretenders with high levels in individual schools of magic (which isn't linear when you empower your pretender/commander). Attributes like awe, which are rare, may be undercosted due to the statistics. Also, the costs I have picked for special attributes and abilities are somewhat arbitrary, although collectively they work out pretty well. I only typed in about 85 default pretenders, as always, more data points means more accuracy.

Special thanks to Endo and all the others who've given feedback on my Mytheology thread. I'll be applying the calculator to all my custom pretenders.

[paradoxharbinger]

it would be neat to see if we could find formulas that would generalize to units in general, not just pretenders.

[quantum_mechani]

I played around with creating a pretender cost formula during the Dom3 bets test, but finally decided it was just far to rigid- you needed more exceptions than rules. Part of the problem is so many abilities/stats/magic paths synergize with each other in odd ways. For instance, the protection value on a cyclops is many times more valuable than on an immobile, fear is better on pretenders that can easily get awe, trample is mostly a downside for a pretender with hand slots but not bad otherwise, etc.

But the hardest part to account for is magic paths. Some paths/combos are simply much more valuable to start with than others. Death and blood are very weak paths to start with due to lack of good early buffs spells and blesses. Starting with fire, astral, or earth and decent dominion instantly makes a great bless chassis, making it more valuable. The value of pathcosts also vary a lot depending on whether the path makes it a bless pretender or not.

Good luck with the project, but I hope you aren't hoping for a breakthrough in balancing.

[Psientist]

that's a good point! There's probably no way to mechanically account for synergies, certainly not statistically! One may always have to apply a degree fo subjective adjustment to any guidelines.

[quantum_mechani]

I don't know if you are interested, but as far as statistics go there is another project I think a lot of people would get use out of. If someone came up with hard statistics on the ratios of what path's sites occur on each terrain type, that could revolutionize site searching. In theory, Edi's database should have all the information necessary, the only catch I can see is maybe knowing exactly how much more common 'common' sites are vs 'rare'.

[Psientist]

I just uploaded an updated version (v1.03) of the Pretender calculator. I decided there were some elements of the linear regression method which were just too quirky, so I used some different statistics-based tricks to come up with a weighting method that incorporates some of the comments posted here and my other thread.

Notably, it now:
- still does a measurably good job of predicting the scores for the default pretenders (whose average cost is about 78 plus or minus 44, by the way)
- provides a way for the modder to score "synergy" ... how well pretender attributes work together or against each other.
- compensates for non-linear scoring of magic levels
- tries to "build in" control over why a pretender's cost should be more or less than the predicted average

[Psientist]

Regarding cost calculators for units and magic sites and the like... if anybody wants me to run regressions or other statistics on any data, just (a) ask, and (b) give me the data typed up in spreadsheet format, with all categories converted to numbers (e.g., "absent"=0, "rare"=1, "common"=2, "very common"=3, etc.)

I also thought it would be interesting to do a listing of pretenders, spells or magic items by the combinations of magic schools, to see if there are any gaps ripe for modding (e.g., "There doesn't seem to be many astral+earth spells?" etc.).