Probably because it's a multi-person game?
Most of the axioms of mathematical economics don't hold in a 3-person trading world. Once the model has more than 2 traders the math gets really nasty. Perhaps the same kind of problem is present in progaming a "Bridge Engine?"
I conjecture a dedicated "Bridge Programmer" would know the quick and dirty answer to your question.
The best chess engines play equal or better than the best human.
The duplicate bridge playing computers play at a relatively very low level. Maybe class C in chess terms.
There are many millions of bridge players on the internet.
You would think someone, somewhere could program a much better bridge playing computer. [they are called Gib at the site I play]
Of course, they know all cards played but when defending against a contract [especially defending against 3 No Trump] they are gadawful
Duplicate Bridge is very complicated but chess is more complicated than Duplicate Bridge.
Why the disparity?