I too am excited for GameID optimized runs, but I also wonder how strange this territory is for speed gaming. I think we could set almost any rules we want, but because Stardew Valley is new and the community is still small, it's hard to know what choices would attract or discourage more runners going forward.

I'm not familiar with Banjo-Tooie or N64, but I looked into the Jinjo manipulation a little bit since TrenteR mentioned it in another thread. It does seem to be a similar situation. There are 2 differences that I see:

The first difference is technical. I couldn't find a video anywhere, but as I understand it, copying the game file is a native capability of Banjo-Tooie on N64. With Stardew, we need to close the program and manually edit a game file (folder). It wouldn't surprise me if this is the first PC game speedrun that would allow this. But at least Tooie also starts runs from loading an existing file. And in our case it appears the segmentation doesn't even affect the timing.

The second difference is substantive - the effect of GameID based RNG on many Stardew categories is massive even with just the Cart considerations. In Banjo-Tooie, the Jinjo pattern matters enough that any world record run is going to use the optimal pattern, be it manipulated or lucky (~1/30), but I think the run is basically the same.

At this point, my opinion is in favor of GameID manipulated submissions. Starting the run from game load instead of new game feels wrong - but from a practical standpoint comparisons from random/unknown GameIDs wont be meaningful. I'm more interested in doing ID-blind runs/races myself, but manipulated runs make more sense for the leaderboard. Even submissions for 'blind' runs should probably include GameID for reference, but with another option or seperate category to indicate they were executed without knowledge of the ID.

The race yesterday was brutal. I would definitely say that any great time is going to need around 12 specific items from the Cart at times you can afford them. 10 hours without using the Cart is probably not too hard to replicate, but a great run might be in the 5 to 6 hour range.

I think I need to do another race/run before I can really try making a specific approach to the route. I agree with most of Asiago's observations from a few posts above. A few notes:

Asiago suggested blueberries before melons in summer - I would do 1 melon (and pepper and tomato) early for the bundle, maybe 6 corn, but the rest blueberries. A couple blueberry harvests will help increase farming skill to get more gold star melons at the end.

Amount of money needed is about 200k if you dont get saved by an early Truffle, Rabbit's Foot, or Egg + Maki Roll from the cart. Those are some of the big potential saves.

In regards to the Maki Roll recipe, it can be learned from the TV show on Summer 21 in the 1st year (but not 2nd year). So no need to buy it or befriend Linus.

I started off my run with about an hour of fishing. I'm not sure if I would do that again, but I need to level up fishing skill at some point to catch the Catfish, Tiger Trout and Sturgeon (and Pufferfish in the rare case I dont see it in a cart). The advantage to doing it early is for a chance at finding a Fire Quartz and getting to skip some of the Mines. And that possibility is very appealing to me.

I got 1 watering can upgrade, and 2 axe upgrades. I would also consider doing 1 pick upgrade in the future.

The prospect of using known seeds (edit: Game IDs) presents a lot of possibilities and some concerns.

Beyond some regular random elements such as weather, daily luck, and cart stocks - imagine the more fringe circumstances - finding a seed where you can dig up a Strange Doll on day 1 (sells for 1000g), and buy 50 extra Parsnips to kickstart the economy. Or knowing exactly which ripened crop locations on various days will yield a gold-star item to complete the quality bundle.

Exploring the game code and solving the puzzle could be a fun community endeavor. Runs might come down to who can plant and pick the quickest, possibly in combination with not fully manipulable random elements (perhaps hooking the right fish on the first cast). It's hard to guess how many people will be interested in competing each time a new golden seed is identified.

I have some familiarity with retro speed-gaming, but I'm not sure how game updates are handled in speedrunning modern games. Will using seed-specific runs increase the potential for making records unbreakable as updates come through?