Answering part of my own question: attacking bonus are affected by terrain halving etc...
There is one river+forteress hex east: bonus from one winterized + moutain div divided by 4.
There is a river on the other units: bonus from HQ attacking city and 3 winterized units halved.
My new count:
* base odds: +4.05 (precise fractional odd)
* city: -1
* factory: -1
* HQ attacking city: +1/2
* Blizzard: -6
* HQ support: +2
* Winterized units (3): +3/2
* Winterized units (1): +1/4
* Moutain div: +1/8 (?)
* Fractional odds: +1
Anyway, the final result is closed to what the panel shows... but not exactly. I wonder how the rounding is done for partial odds modifications (I'd say the fractional odds should apply, but who knows)
It's hard to see the rivers that are being attacked across. I make it 22.25 to 11 which is a DRM of 4.045. (You never round, just truncate everything after the third decimal place in MWiF.) I agree on all the boni, so it looks like a net DRM of zero with a fractional chance of .420 for a DRM of +1.
All 3 hexsides have rivers. The eastward facing hexside has a fort (new construction). But I do not know which units are attacking across that fort hexside.
Calculating this stuff is really difficult. At this point I trust the program (having delved into the details of many reported bugs on calculating land combat odds).
---
Not being asked about using winterized units might still be an issue.