Parsers don't kill experimenting, they facilitate it and improve it. They make testing and experimenting objective and functional instead of biased and subjective. In fact, a parser would help you experiment with stats and certain builds to legitimize them. The alternative is that the player who balks at "cookie cutting" is the sole determinant of how viable their "unique" combination is, which is inherently biased and other players will pick up on and discriminate against that bias to the point where the setup gets MORE flak than it actually deserves just to prove a point.

It's easy to discriminate against setups that look gimp, but harder to discriminate against setups that hold their own on the parser even though they look ridiculous. No one makes fun of setups that perform nominally lower than the established setups as long as there is novel benefit to the combination. The parser doesn't tell us much about your setup that we didn't already know, really. It just validates it objectively.

You may pitch to us a red mage who nukes like a mage and melees like a DD. You may say you do a lot of damage after "experimenting with it." We see your nukes hitting for 85 and your melee hits wailing at a whopping average of 9 per hit, and while the 3 dds are each doing 28% of the damage, you're doing 10%, which is 4% more than the whm casting banish intermittently when they're bored. The parser tells us what we already know: that a RDM/WAR is not even half a meleer and not even half a mage, so we're better off getting a full one of either.

If your "experiments" validate your odd combination without a parser, but if a parser would be the end of your odd combination, then your experiments are invalid. The parser should back up your experiments, not crush them.