New Doom 2 Speedrun Record Breaks The 20 Minute Barrier

The run isn't perfect...

Booo. Just kidding, I will never reach his Doom skills in a lifetime or two. I can barely finish Doom 2 in one sitting.

This run looks so effortless - I know it wasn't, but god damn half of it seems like a casual stroll! Amazing stuff!

When that final counter stopped at 19:59 he must have gone apeshit!

Impressive. Most impressive.

Thats awesome, I didn't think it was possible to complete Doom 2 under 20 minutes.

Wow. Freakin' awesome!! Doom2 Done Quick in Real Life! Come to mention it, anyone recall offhand the total time on D2DQ? does this beat it?

The blooper reel is great too. Good editing job there!

That is fantastic! I can't believe how many tricks and aggressive strategies have been incorporated into the route.

Damn, that's pretty impressive. Even more so is that the record for a 20+ year old game has been improved by a whopping 13% in just 10 months.

And yeah, not perfect... he had to do a jump twice at the end of MAP06, for example. But some of those rocket jumps are pretty damn impressive (MAP14, MAP29).

Rez said:

Wow. Freakin' awesome!! Doom2 Done Quick in Real Life! Come to mention it, anyone recall offhand the total time on D2DQ? does this beat it?

The blooper reel is great too. Good editing job there!

Yep, D2DQ was 21:16. So actually even ZM's previous record beat it.

Pretty awesome that the first D2 TAS (even it was a crude form of TAS) has been beaten.

Next goal: beat 16:17 :)

19:59 will be broken in a few years. Now that we're so used to complex and lca, these "normal" monsters are a joke.

Magnusblitz said:

Damn, that's pretty impressive. Even more so is that the record for a 20+ year old game has been improved by a whopping 13% in just 10 months.

Dropping from 23:25 to 19:59, I calculated 14.66 % difference. :)

Congrats, Zero-Master!

Fredrik said:

Pretty awesome that the first D2 TAS (even it was a crude form of TAS) has been beaten.

You're assuming this one wasn't?

Grazza said:

You're assuming this one wasn't?

Ha ha, well aware of it, but not sure if it should count... :)

Why, was that demo TASed?

Girlich and Winterfeldt were proven to use tools in some of their runs, so without being able to prove which ones were legit, all of their demos were removed from CN tables.

What tools were even available in 1995? Did they use some gross hacking to slow down their entire system or something?

princetontiger said:

19:59 will be broken in a few years. Now that we're so used to complex and lca, these "normal" monsters are a joke.

Considering he was on track for a 19:52 or 19:53 on one of the failed runs in his blooper reel, it could be faster than a "few years." Depends on how things play out from here :P

princetontiger said:

19:59 will be broken in a few years. Now that we're so used to complex and lca, these "normal" monsters are a joke.

show me

Linguica said:

What tools were even available in 1995? Did they use some gross hacking to slow down their entire system or something?

As I understand it, the consensus is that Girlich and Winterfedlt were able to record in slow-motion and to connect together separately recorded demos. It's all a bit murky because obviously they didn't advertise what they were doing or how, and even after they were caught the details were deliberately kept between a few Compet-n insiders (of which I was not one, though I picked up a few things), as it was felt that making them public would only encourage other potential cheaters and help them to avoid detection. Such security through obscurity doesn't seem warranted anymore.

Slow-motion was achieved, I believe, by reversing engineering the .exe and modifying the timer so the game ran at a reduced but stable framerate. Some of Girlich's demos show quite blatant signs of being recorded in slow-motion.

Piecing together demos to create long runs in vanilla Doom relies on save abuse. I've not tried it, but I believe the technique goes something like this:

1. Start recording and play until you fuck up or die.
   
2. Decompile the resulting demo using LMPC
   
3. Find a tic in the demo just prior to your error, and insert a SG (save game) instruction.
   
4. Recompile the demo and play it through to create the save.
   
5. Record a new demo, using the -loadgame parameter to start recording from the save point.
   
6. Play until you fuck up again
   
7. Decompile the second demo in LMPC and append the contents to the demo from step 2.
   
8. Repeat this process until you have completed the episode/game/whatever.
   

Was this possible in 1995? Certainly. Girlich released the first version of LMPC at the end of August 1994, so he already had a good enough understanding of the demo format by then.

But wouldn't that savegame thing immediately desync? Given that active monsters all reset to inactive when you load a savegame, etc.

Well, saving at map start would be enough cheating for an episodic/full run.

Hmm yeah. The technique as I described it won't work, as the RNG isn't actually stored in the saves. I'm certain there was some connection technique that used savegames though. Possibly it also involved hacking the .exe to change the initial index of the RNG.

Really? People have enough confidence now to consistently pull off those moves at the end of Map24 and Map27? Call me just a teeeensy bit skeptical.

AndrewB said:

Really? People have enough confidence now to consistently pull off those moves at the end of Map24 and Map27?

Why yes they do.

AndrewB said:

Really? People have enough confidence now to consistently pull off those moves at the end of Map24 and Map27? Call me just a teeeensy bit skeptical.

Go watch some speedrunning live streams. You'll be amazed.

While Zero Master hasn't been streaming his attempts, watching Looper in action proves that it's possible. You can tell just by looking at the demos that ZM has spent a huge amount of time working out how to execute those tricks consistently; he's not just winging them.

It seems that speedrunners across many games are becoming much better. Not just because the list of known tricks for each game always expands, but because runners have learned how to practice more efficiently. That means spending a lot of time working out consistent setups for tricks and of course being willing to grind huge amounts of runs in a focused way. There's also the mental hurdle of believing that something is too hard until you actually see it being done, and ZM deserves a lot of credit for blazing the trail. As amazing as this result is right now, I wouldn't be surprised if several people have pulled off sub-20 runs of Doom 2 five years from now.

A nice example outside of Doom is the Phantoon first any% route in Super Metroid, which requires hitting a frame-perfect wall jump (alternatively, a very difficult horizontal bomb jump) 10 minutes into the game. This was long considered TAS only, yet a few months ago one person managed to use it in a run, and after that, other runners simply had to force themselves to learn it. Super Metroid has some other tricks that once were considered TAS only, like the baby metroid skip. There are countless other examples from other games.

Ok. I accept that.

AndrewB said:

Really? People have enough confidence now to consistently pull off those moves at the end of Map24 and Map27? Call me just a teeeensy bit skeptical.

Imagine the hundreds of failed runs before that ...

wow, that's crazy. speedruns improved so much lately, with this first sub-20 min (19:59, lol!), plutonia done, scythe2 done...