Congratulations! You win the award for being the 10 million'th person to WANT a better fit than they got. Sadly, just wanting something is not sufficient. If it was, then the Cubs would have won a World Series long ago. If it was, then every person who ever bought a lottery ticket would have won the first prize. If it was, then we would have achieved peace in the mid-east long ago.
There is no magic to be found. Anyway, you can think of the regression estimate as the result of an optimization ALREADY. A further optimization won't help, because what you got is already the best you can get.
To get better results, you must get better data, with lower variability. You choose a better, more intelligent model, that better represents the data you do have.