All points about scripts seems to be about making it easier to spend less time or that it needs to be possible by a human. What I like to do is to balance this up with a third weight that is the looks.
For M249 boosting, you can’t manually look back perfectly each shot like in my segmented speedrun. The look is quite slick thoug. But you could just as well do full maps looking backwards just as fast. But it would be harder and take more time. It would also be entertaining, but maybe in another sense.
With a gauss script you get a clean, straight forward and easy to follow speedrun.
However a one frame turn can easily be made with a one frame segment which leads me to my initial question. If someone badly wants the best segmented time, what is stopping him from doing frame long segments manually. Frame long segments that with repetition can be equal or extremely close to a TAS, without the need of any script. I’m thinking about something like this but a step further.
So in order to get segmented runs competable for real I think you’d need rules for the segmenting itself. Segment penalty was a solution for that. But nowdays not even SDA applies it? Most games seems to get shorter and shorter segments. Because it’s faster. And easier. I know the scenario above is extreme but that’s the direction it’s heading. And it’s very hard to rule it.
In my eyes SS is for competition. And segmented runs set their own rules and are “not really ever intended to compete and compare against others but compete agains the game while still showing skillful entertainment exhibits as much as possible.” - SW
One of the reasons I lost some interest was precisely because of this, there didn’t seem to be a good solution either; because of the ability to create frame long segments, like you say, one could essentially (with time on their side) make a TAS like run without any scripts which would nullify the script ruling originally meant to prevent TAS runs overlapping the segmented ones :d
May have to tackle this first. One of the reasons may be how easy it is to create small segments using scripts/bxt/bmod pro etc
I firmly believe segmenting can in fact be much harder (or easier), players that perform really well in an SS on a top-tier level may or may not be able to get fine tuned “perfect” strafes in segments etc, unfortunately there’s no good way of telling how much practice one might need for a specific segment. Those last milliseconds can really mean a lot, but may not be distinguishable from those with a few frames slower in the final video.
It seems natural that the community begins to reconsider segmented runs now that TASing is becoming more common.
Before this, it probably made the most sense to fuse the segmented & TAS categories together to create an entertainment focused speedrun without general limitations, much like how host_framerate runs have been made as “virtual TASes” for looks. But I’d like to think any type of speedrun done by a decent player/s that is well optimized would generally be considered “entertaining”.
People are obviously free to make whatever kind of speedrun they like. I think it’s fine if people want to make “experimental” speedruns (such as sv_cheats 1, fusion TAS, modified etc), but for submissions to the YouTube channel I think it makes more sense to have a logical, clean standard so that there is room for clear improvement.
Like quad said, you could technically beat older runs with tactics like extra scripting and segmenting where a run could otherwise be practically non-improvable. I don’t think anyone would actually do this, but even in a realistic scenario, it still makes it unclear to see how much time you *actually *saved over the old run if you saved *extra *time by using extra assistance; it’s basically a false timesave. Even for a TAS it’s clear to see exactly how much time has been saved over an older run because they have their own limitation which is the game itself.
For a comparison, a Quake segmented speedrun is competed for with the segment limitation of 1 per map and (generally) no scripts allowed. Is this applicable to a GoldSrc speedrun? Segment-per-map is not really a possibility as you would have to arbitrarily allow saveglitch segments and segmenting right before changelevels for NPC abuse sections. Scriptless is not viable because of the way jumping/ducking etc works differently so the autojump/duck etc is the solution to that, but is there a need for anything else? I’m not seeing scripted rocketjumps in Quake (in-case there’s some exception, I’m aware there is/was a minor gray area in their rules).
The only proper solution that’s been thought of as mentioned is the segmenting penalty. It makes players plan and regulate their segmenting wisely and it’s also not too radical for people to adapt to, but personally I’m not fond of the idea of a “fake” time. Maybe some sort of point-based system could be used instead?
About the segmenting policy: I’m not sure if this will work well in all cases because, for instance, Portal Done Inbounds had to use a huge load of tiny segments in some parts to get the desired movement (for instance the camera boosting parts).
Generally, competitive segmented speedruns don’t seem to me like a very good idea because so far new (goldsrc/source) segmented speedruns were made if some considerable amount of time was to be cut off due to a newly discovered glitch / etc. But then, this is probably caused by the current segmenting policy.
Definitely a downside to take into consideration since you could basically lose time doing something that’s supposed to save time, but I suppose if the improvement over the old run comes down to something as minor as that then the run might not be worth doing anyway.
Also, I don’t think that the “project” system (only working on a run when there is a considerable amount of time to be saved) would be any different by making segmented runs more competitive since it would be pretty pointless to put that much effort into something that is barely any different, just like it would be under the current system. It’s pretty much the same way in Quake also. I think it’s more about making all runs play under the same circumstances so that there aren’t “false” timesaves (using extra assistance to cut more time).
Segmenting rules have been talked about in past threads, I believe some sort of time adjustment per segment is the best solution that’s ever come about. It doesn’t necessarily need to be .5s but that ends up being a pretty good figure. I think the ideal figure might be slightly larger, but still less than 1.0s. Run times should have the information of the actual time, number of segments, and then the adjustment quotient for competitive and comparative purposes.
Also I don’t think one segment per level is all that different for HL as it is for Quake and other games. You still start the next segment or level with the health, armor, ammo, guns, ect left over from the end of the previous segment, only difference is the the position and velocity are also carried over for all but a few xen maps.
Another point to consider is entertainment; the little time save and many segments strat might be way cooler (and just simply new). This also applies to cutscenes (you might want to do some precise / random stuff to make the cutscenes less boring, that’d require a segment to not waste a huge amount of time on). This might seem like a minor issue, but one of the main purposes of segmented speedruns is entertainment.
I still feel like entertainment is a bit of a moot point to consider.
I don’t mean that speedruns should not focus on entertainment, of course. What I mean is that entertainment is entirely subjective and can be applied to all established categories. Again, with Quake segmented runs they are done 1 segment per-map and are still very popular for being highly entertaining.
In GoldSrc’s case, with the penalty system seemingly being the best option, it would most likely not be that different to what we see with segmented runs today. Especially with there probably being no point in doing another run of the original Half-Life + expansions (other than possibly Opposing Force), it means that in most cases we’re going to see runs of mods that haven’t been done before and therefore would allow runners more flexibility anyway.
All we’d really be doing by implementing this policy is making sure that in a worst case scenario, nobody can come along and take advantage of the unlimited segmenting, as well as persuading players to only use as many segments as they actually need. Otherwise as pointed out previously, you’re essentially leaving one of two “TAS-precision” doors open.
No restrictions = virtual TAS | Restrict scripts = more segments = virtual TAS | Restrict segments = more scripts = virtual TAS | Restrict both = a human run that’s distinct from a TAS
I think you might be missing one of the points that Yalter’s bringing up on entertainment in regard to cut-scenes/dead time. This I think is an easily amended small problem. In most cases it should be very easy to arbitrate when there is dead time that can’t be spent to gain any advantage towards the run. In such sections segment penalties could be ignored to allow people to add extra segments in to do some better and more messing around entertaining tricks and stuff.
This was something I thought about when making half-hour half-life in 2005. I actually needlessly added 3 extra segments for the test chamber just to do those tricks better with much less real time invested, figuring most people, particularly radix, would understand and not penalize me for it, but also wiling to take their 1.5s competition penalty as a worth while trade off. Ultimately I really didn’t care much in the end how it was seen as there was no competition. This was motivating from much of the comments on my OP4 run done prior being entertainment based and hyper critical and aware of what happened during dead time.
The other point is dicey and came up a lot when deciding what do during HL21 planning and execution. There’s quite a few times where different, extremely technical strats were possible to be executed by a human but only with adding extra segment(s) that ultimately would never come near to making up a .5s penalty. Quad and I agreed to roughly follow a .1s penalty guideline for these cases but still subjective to each case. For competition though, it’s hard to think of anything that could work, this might just be one area between human and TAS that gets left out when trying to standardize for human vs human competition.