Model card: explain how Minimum and Recommended specs feed the model and which ones the app passes

README.md

@@ -122,7 +122,9 @@ All four tests can be rerun with `evaluate_demo_end_to_end.py`, `evaluate_extern
 3. **The model for everything else.** Games without a curve use this model: 18.5% on 132 end-to-end points.
 4. **Matching the right game.** Whole-word title matching cut the end-to-end error from 35.1% to 15.2% with no model change. The old matcher sent "Assassin's Creed Mirage" to "RAGE".
 
-The app can also be worse than the model. Its store-spec path passes fewer inputs (no VRAM, release year or genre), so on F1 24 it is 54% off while the model is 29% off.
+The app can also be worse than the model. Its store-spec path passes only the recommended and minimum GPU scores. The recommended CPU, VRAM and RAM fall back to defaults (13,000 PassMark, 6 GB, 16 GB), and release year and genre are left out, so on F1 24 it is 54% off while the model with all inputs is 29% off. Passing the full parsed specs is the next fix to test.
+
+**How the store specs are used.** 8 of the 21 features come from the Minimum and Recommended specs: both GPU scores and the user's headroom over each, VRAM headroom and a VRAM cliff penalty, a demand index from the recommended GPU, CPU and RAM, and CPU headroom over the recommended CPU. A page with only one GPU gets the other at a 1.65× ratio. The same specs also generated most training targets (see Training data), so the model mostly learned how far a PC sits above the recommended GPU.
 
 ### Compared with other FPS estimators