Will report back in a few days.

Well, almost.

Tic Tac Toe appears to be a pretty good tupper diagnosis tool
We've played about 60 games of connect-four on 8x8 from May to July of which the tupper won more than 2/3.
I tried again today and I easily won 4 out of 4. Extremely weird. But it shows there's definately somethin wrong. Alice is much, much weaker than in May, or even January.

I think Fede's explanation bears some truth, I probably did want her to win in the beginning and put more focus into her rounds than mine. But even later when we played for me doing unpleasant chores she managed to win significantly more than 50%.

Concluson:
need to force moar