I mean that might help solve the problem, but also introduce new ones.
- The inverse is now true; it's better to practice on CPs than on starters. If you win on starters but lose on CPs, you lose sets 1-2. If the reverse, you win sets 2-1.
- Starters not being available on game 1 also limits the stage list a lot in subsequent games.
I don't think it's quite that simple because while you
can win only on counterpicks, to do so means you must win on an opponents counterpick.
This means that, for CP-focused strategies to work you must be better enough than your opponent to overcome the stage disadvantage. If the opponent is of similar skill and has put time into their counterpick choice(which is very likely as the list of stages you want to CP to is going to be reasonably small, you know what stages your character is good on, and it's very important to win on them as you'll be a game down and fighting to avoid a loss) getting such a skill difference through stage practicealone, when split between all of the CPs, seems unlikely.
If you can't win on the opponents counterpick you have to win the starter. In addition to being neutral instead of disadvantageous you are able to control the starter to a degree with your strikes, and you'll never see the stage on the CP, so you can neglect your bad starters almost completely. If the list of CPs is reasonably large compared to starters you should end up needing to focus on fewer stages with a starter focused strategy than a CP focused on.
Let's say you have 5 starters and 9 counterpicks with 1 CP-ban. A CP focused strategy would have to learn at least 8 stages (if you strike the same stage every time), but the nature of matchups means you probably have to learn all 9, while a more balanced strategy could learn 3 CPs they plan to take opponents to and 4 starters (ignoring your worst, as you'd always strike it with 1 of your 2 strikes). the balanced strategy learns 7 compared to the CP strategy's 9. I'm no pro by any stretch so my numbers may be wrong, but even if I'm off by 2 putting effort into starters is at least equally viable before you account for the need to overcome disadvantage, which tilts things further towards starters.
As for the second problem, I think I have an idea that both mitigates that and allows mixed hazards to exist in a way that reduces the amount of logistical errors: Split hazards on and hazards off between starter and counterpick while disallowing CPing to starters.
This would cause there to always be 2 settings changes: 1 before the set begins and 1 after game 1, making the settings change part of the routine. By making something routine it becomes easier to remember, the players no longer have to consider their hazards setting based on stages played, they just know to do it before the first CP (possibly while the loser is thinking about where they want to go). This also limits where issues can occur, it will either happen in game 1 where everything is still neutral (or even during a button check, where it affects nothing), or on game 2, which is to say it won't happen on a hype game 5 of grand finals or some other worst case scenario. Both of these things lessen the risks of mixed hazards.
In return, mixed hazards gives more unique stages, which allows for more stages left over to fill the counterpicks. FoD/either yoshis/SV/T&C is already 4 unique stages for the starter (assuming the issues on the former 2 are patched), and I haven't even used FD yet, leaving every stage allowed by hazards off to fill the CP selection. As a bonus, having both FD and battlefield on the CP list (very possible with a few more adventurous starter choices like prism tower) would let both battlefield forms and omegas to be counterpicked to, making them easily to make legal.
There are some compromises to this (you lose certain otherwise very neutral stages as a starter pick like hazardless PS, can't be too picky about stages in general for the starter list, requires accepting the odd mishap caused by mixed hazards even if it is reduced in frequency by routine) but the way it builds seems like a strong middleground between many of the current rulesets disputes.
