[Feature Request] Weighted / Point-Based Training Decision

[Nirioppai]

Add an optional weighted scoring system for training decisions, where each “training element” (e.g., Friendship Training, Supporter presence, condition checks, etc.) contributes points. The automation then totals points per available action (train types / race / recreation) and picks the highest-scoring option based on user-defined weights.

Benefits
Custom training behavior → users can tune priorities per strategy (speed-focused, balanced, etc.)
More flexible logic → multiple signals can contribute to a decision instead of hard overrides
Easier experimentation → tweak weights instead of rewriting rules

Sample:

[steve1316]

This seems interesting to have! And thanks for the sample, it helps in putting it into perspective.

For the Kitasan Black Entity, I assume you mean to mark specific support cards with higher/lower weights in this. Then I think Supporter means just any support card being under the correct training like a Speed supporter in Speed training as opposed to a Speed supporter in a Guts training.

I don't think we can do specific support card Entities like El Condor Power SSR, Kitasan Black Speed SSR, Fine Motion Wit SSR, etc. but we can do it generically by looking at that tiny stat square at the top left of their portraits.

I will see if I can work this in with our current systems.

[Nirioppai]

Thank you! Yes, I’m actively using the app, and I think this could be a really solid improvement.

Just to clarify: the weights / point-based training metric isn’t originally my idea—other players already use a similar way of thinking. My problem is I’m pretty bad at calculating the trade-offs in my head mid-run, so if the app can compute it consistently (based on configurable weights), it would be way better than doing it manually.

If the team can get this working well, I think it can improve run consistency a lot, for example:

• More reliably entering Classic year with supporters ready for rainbows without sacrificing overall stats, since you can pick the best-value training more objectively.
• More reliably reaching 100k fans before Classic year December to prep for Senior year events—because if the app shows all available trainings are under a certain score/weight threshold, it’s easier to justify “just race instead” rather than forcing a bad training.

That said, my main concern is scope: this might be a pretty big feature and could conflict with (or override) the current training sequence / decision flow you already have. If you think that’s a risk too, maybe it’s better introduced as an optional layer (e.g., a toggle, or a “weight suggestion” displayed alongside the existing recommendations) rather than replacing the current logic outright.

[butteredping]

just to chime in, out of the various auto projects, umaplay has probably the best weight system in regards to making smart decisions, using YOLO to weight stats/events/supports and adjust based on current stats and energy to avoid overcap. not the same but might have bits you don't have to invent from scratch.

https://github.com/Magody/Umaplay/blob/b438842b9903260175b1acedc6589fcdbd81fc17/core/settings.py#L95

[steve1316]

just to chime in, out of the various auto projects, umaplay has probably the best weight system in regards to making smart decisions, using YOLO to weight stats/events/supports and adjust based on current stats and energy to avoid overcap. not the same but might have bits you don't have to invent from scratch.

Magody/Umaplay@b438842/core/settings.py#L95

YOLO cannot make weighted training decisions by itself. It can only be used to see what is on the screen depending on what dataset it is trained on. From what I see, their training decisions is based off of their own weighted system via scoring what is available on screen. Which is what we are doing.