The need to wait out the demo after each reset comes up repeatedly in discussions, so I thought I'd chime in with a possible solution.

Would this community be open to adding a provision that allows using an agreed-upon romhack which locks the RNG from the very start, without watching the demo?

I've already made such a romhack, the technical part is trivial. But I understand that there may be reservations about using such a romhack.

What I can see as arguments against:

1. Unfair to the previous runners of the game, who had to wait out the demo.
2. Unfair to those runners who run the game using the original cartridge.

My personal view on these (albeit I'm not a CnD runner): no. 2 is untenable in the long run, as there are no new CnD cartridges getting produced, and the old stock is bound to dwindle over time. No. 1 is worth considering, but this kind of disruption would be true for any other new development (e.g. a new skip/strat), and these only receive pushback (in the form of relegating them to a separate category or banning them) if they're viewed as somehow making the speedrun markedly 'worse'.

What I can see as arguments for:

1. Stops you from wasting 16 seconds each time you reset (obvious), making pole climbs much less aggravating to master.
2. Makes the game more appealing to new and returning runners.
3. Can still be used on original hardware using flashcarts.
4. The hack can be easily distributed as an IPS patch or even a link to an online patcher with the prompt to provide the original ROM, which should make it easy to popularise.

So... if the esteemed moderators of this game are open to this idea, I would propose holding a community poll to see if it's tenable.

I'm sure this one isn't new, but I haven't seen any runners employing this one, so I thought I'd post it here.

You can skip the small hand near the end of the Genie level with a dash jump + sword slash. This may save up to 2.5 seconds depending on the small hand's cycle.

The best thing is, if you see the small hand turn big before you jump, you can go for the regular route on reaction.

Please vote whether you'd be okay with making Normal Any% the default category over Practice. As for the cut-off point for this poll - I'm suggesting implementing the change (or just ending the poll) when there's an overwhelming consensus (say, 90% of votes) with 10+ votes in total.

My arguments in favour:

1. Potentially the quickest category now thanks to the Cave boss skip.
2. Somewhat forgiving. Almost as easy to get into as Practice, and not as brutal as Difficult.
3. Seems just as popular as Practice, judging by the number of runners.

My arguments against:

1. The climb after the Cave boss skip introduces an extra element of randomness into the run.
2. There's a slight prevalence of recent runs on Practice over Normal.

Notes: Post a message with your vote. I don't think counting Likes as votes 'For' would be fair. Also, I do encourage non-runners to vote and give their reasoning too, but I won't be counting their votes just to be safe.

Current tally. For: 9 votes (Svipur, Demian_Neville, MrMonsh, veschii_nevstrui, GreyGurd, tepleyev, takojaki, PaulSaltine, ZeeGee_) + 4 votes from non-runners (PakLomak, EmilSinkler, JackPro, OopsKapootz). Against: 0 votes.

Vote passed.

So I've been trying out the SNES version, getting used to the controls, and I've found a neat little trick that allows you to glitch through uneven ledges.

One use for it that I found is skipping the Level 1 boss entirely:

The idea is to stand in a specific spot on the edge of a platform (here, the window is somewhat generous, about 3-4 pixels), facing away from the ledge, crouch, and go prone.

Remember to buffer another jump when you clip through the branch to jump over the Jag (to avoid triggering the fight). My visual cue for this one is that Harry's left toes should be a bit past the bigger gap in the cobweb when you're facing rightwards.

Another use for this technique that I found is doing a deathless Tikal Temple clip, saving about 3 seconds over the current method thanks to not needing to respawn (it's also a lot more consistent and easier to execute - well, for me, at least):

Hope people can find more uses for this one.

I've been routing out the Genesis version for a few days with the intention of running it, and here's a list of differences I noted from the SNES version.

1. Harry Jr. runs significantly faster in the SNES version. Shaves off several seconds every time you just run in a straight line.

2. Obvious one: the SNES version skips two boss fights. Couldn't find a way to skip those in the Genesis version.

3. The HP of the remaining bosses is much higher in the Genesis version, putting even more emphasis on ammo preservation.

4. An extension of #1. Some of the tricks become nigh pixel-perfect, compared to the SNES version. E.g. the jump to the Hourglass platform in Level 2.

5. You have to push the minecart in Level 3 to grab the rope in the Genesis version.

6. The autoscroller minecart level is about 40 seconds longer in the Genesis version.

7. There are additional human-possible skips and clips in the Genesis version saving about 20 seconds over the SNES version (these are missing in the current SNES TAS, haven't played the SNES version myself to verify it). E.g. the wall clip in Level 5 and the climb in Level 6.

8. Dying resets your slingshot ammo to 20 in the Genesis version - which, combined with #2 and #3, puts more emphasis on health preservation as well.

The idea is this.

Calm doesn't show off the later levels, and is pretty limited in terms of optimisation.

Total Demolition might be a tad too grindy to get into, and not too different from Crazed (the big ones being the Train level and the Library).

So Crazed is probably the optimal category for speedruns, with enough depth in terms of optimisation of movement, boss kills, etc.

... or at least add those in?

I've been doing runs of this one on the Genesis Plus GX core in RetroArch, so I thought I'd share my rationale for this to pre-empt possible questions.

It all stems from these testing efforts by Sonic 3 runners into whether runs on emulators offer any significant advantage over console runs: https://www.speedrun.com/s3k/thread/w89y7/2#vcgia

The gist of it is this: Fusion is usually quite a bit faster than a real console. Blastem is usually on par with a real console. Genesis Plus GX core (in RetroArch and Bizhawk) is just marginally faster than a console. Most other emus fall between GenPlus GX and Fusion.

Here's a quick comparison vid: