Marble Madness Record History:

July 31 2005, 3:13, Elliott Feiertag
October 27 2008, 2:54.2, Andrew Gardikis
March 18 2014, 2:52.1, Steve Barrios
March 21 2014, 2:50.9, Steve Barrios
May 22 2014, 2:48.84, Big Walsh
July 24 2014, 2:48.68, Steve Barrios
October 9 2014, 2:47.51, Steve Barrios
July 24 2015, 2:46.930, Toad22484
July 25 2015, 2:45.530, Toad22484
December 5 2015, 2:45.465, Steve Barrios
October 21 2016, 2:44.998, Steve Barrios
April 24 2017, 2:44.271, Steve Barrios
July 19 2017, 2:44.198, Steve Barrios
July 20 2017, 2:43.664, Steve Barrios
October 25 2017, 2:42.407, Spef
November 1 2017, 2:42.203, Spef
November 15 2017, 2:41.127, Spef
August 6 2019, 2:40.883, yelsraek


This speedrun of Marble Madness times out to 2:40.883 in Yua, and has 0 deaths.  It was done live on http://www.twitch.tv/yelsraek and retimed afterwards.

A *LOT* has changed in the two years since Elipsis registered the first ever 2:43.  At that time, he offered up that a 2:42 would be possible with multiple golds in a run.  Then, in the latter half of 2017, an influx of high-level runners hit the scene, including Kinnijup, MikeIsMyIke, and Spef.  Kinni and Mike both posted 2:44s in a couple of months, while Spef pushed the game even further, becoming the first runner to break not only the 2:42 barrier, but the 2:41 barrier as well!

As for me, Elipsis taught me how to play Marble Madness at AGDQ 2017.  A couple of months later, I got a 2:45 and put it down for a long while.  When I picked it back up in October 2018, I wanted a 2:43 of my own!  By the time I got there, I was starting to see some of the frame rules Spef was hitting, so I kept grinding and grinding and grinding and here we are!  The 2:40 barrier has been breached!

Frame Rule explanation:  In Marble Madness, the game will only advance from one race to the next every 16 frames.  At 60 frames per second, this equates to 0.266 seconds.  For the first five races of the level commentary, time saved or lost will only be in these intervals.  For the sixth and final race, the platform cycles rotate every two frame rules, which is every 32 frames or 0.533 seconds.  Timing ends on an exact frame rather than a frame rule, because loss of control occurs on the frame that the 6000 appears over the marble, which is the accepted timing method.

Level Commentary:

Practice Race:

Elipsis previously described three different possible "55" frame rules that you can land on, "slow", "fast", and "TAS".  I'm on the "fast" frame rule here, about 6 frames shy of the TAS frame rule.

Beginner Race:

The beginner race is a race of three parts:  the draw bridge, the bumpy road, and the ice.  I hit all three of them pretty smoothly and got within one frame rule of the TAS, matching my gold for this race.

Intermediate Race:

Intermediate is the only race where I feel like I left a frame rule on the table.  It's nearly imperceptible, but some of the cornering could be a bit tighter.  A segment at the bottom is usually done blind from the pipe to the carpet, but on this attempt I could see the top of the marble the whole time, so I could tell I was a little slow overall.  As a result, a frame rule that I am capable of hitting somewhat frequently was missed.  At this point in the run, I'm now one frame rule behind Spef's PB.

Aerial Race:

The first bar to clear for the Aerial race is getting to the catapult before any pistons spawn.  The second bar to clear is getting the early hammers pattern, where you cleanly sneak between hammer #4 and hammer #3 (if you number them from front-to-back).  This involves some of the most aggressive cornering you see in the run.  Spinning on tight corners can also help, because the screen stops scrolling for the 2 or 3 frames that the marble is falling.  This can delay the hammer spawns and give you a chance to make the cycle with a few frames to spare.  This Aerial race matches my gold.  Despite this, I'm still one frame rule behind Spef's PB.

Silly Race:

I'm going to paraphrase Elipsis's description of Silly here:  This one is the run killer, getting a top time in Marble Madness is tremendously difficult primarily due to this race.  The Silly Skip shortcut is, optimally, a pixel-perfect endeavor... and the variability of the preceding birds mean that there is no consistent setup or approach to nailing the trick, which has now become mandatory.  So not only do you have to go for it to have a shot at improving the world record, but the required inputs to execute it end up being slightly different every...single... time.

My movement in this race is so aggressive that I occasionally go off the top of the screen coming out of the first pipe.  This ends up costing me time in the long run, because my lack of sight leads to the one wide turn before the ramp up to the birds.  That wide turn cost me one frame rule that I have gotten a couple of times before, but I refuse to beat myself up about it because so many attempts die to the skip anyway.  Having one slow turn in a run with a clean silly skip is A-OK in my book.  Overall, this was two frame rules faster than Spef's PB, so I'm now ahead by one frame rule and into Ultimate on 2:40 pace.

Quick aside:  I wear a Heart Rate monitor while I stream, and the Silly Race always gives away my nervousness.  I started the Silly Race at 75 bpm, and by the time I get control at the start of the Ultimate Race, I'm up to 128 bpm!

Ultimate Race:

Ultimate may not be the hardest race in the game, but it certainly has the hardest turn in the game.  The turn in the bumpy section has a very tight range between missing the ramp completely on the left and bouncing off the screen edge right (as Elipsis did in his run).  Beyond that one turn, Ultimate is all about keeping the screen moving.  At the bottom of the race is the floating platform section, which has two cycles: fast and slow.  You have 18.88 seconds (1131 frames) from the fade to black at the end of the Silly Race to get to the bottom of the screen scroll and catch the fast cycle (the same cycle as the TAS).  One frame late, and you lose two frame rules waiting for the adjacent platform to appear in front of you.  Elipsis lost these frame rules because of the screen bounce, so in effect that one bounce cost him one full second.

My run arrives on frame 1131, the very last frame on the fast cycle, and the rest is history.  An aggressive finish that is one frame slower than the TAS brings the run to a finish.

Final Time: 2:40.883

Many thanks are in order.  First, to Elipsis, for the constant support and for being an amazing co-op partner.  To Spef, MikeIsMyIke, and Kinnijup for always being willing to explain their knowledge of the game in such rich detail.  To the ever-expanding Marble community which, as of August 2019, has over 80 runners on the leaderboard.  Last but not least to Games Done Quick, who not only let Elipsis and me perform a Co-Op run at SGDQ 2019, but also let me showcase a 1-player 2-controller run that better than I could ever have dreamed.

Where Do We Go From Here?

As speedrunners, we like arbitrary goals.  So, what would it take to get 2:39?  My sum of best is 2:39.8.  Since, that is less than one frame rule away from 2:40, that means I need every frame rule I have ever gotten, all in one run.  I've gotten my fastest Intermediate frame rule once EVER, and my fastest Silly frame rule twice EVER.  The Marble Madness TAS is a 2:36.266, but almost all of that time is from pushing off-screen and doing blind movement in Intermediate, Aerial, and Silly that I cannot imagine replicating.  But we shall keep digging as always!  My Beginner was 15 frames off the TAS, and none of that movement is blind.  The fact that three whole seconds can be saved with LOTS of offscreen movement means that we humans could probably scrape an extra frame rule with a LITTLE offscreen movement.  I think we will see 2:39 someday.  Today, we got 2:40, and I couldn't be happier to show it to the world.

Marble Madness Co-op Record History:

January 11 2017, 3:17.298  - Toad22484 and Steve Barrios
September 4 2017, 3:17.097 - chessjerk and yelsraek
January 8 2019, 3:14.349 - yelsraek and Steve Barrios
June 28 2019, 3:07.704 - yelsraek and Steve Barrios

(Written commentary by Elipsis - please see audio track 2 for joint audio commentary from both runners)

Strong co-op runs of Marble Madness are always elusive due to the difficulty of gathering regionally disparate players.  So far, 3 out of the 4 records have taken place at GDQs.  This latest entry took place at SGDQ 2019 in the private practice room.  While the time here is 3:07, it is worth nothing that real-time 12 seconds in any two player run will always be lost rewarding the winner of the previous race with an additional five in-game seconds (2.4 seconds per race times 5).  Yelsraek and I knew that we could improve our previous 3:14 run this time around, as we were both in top form from our earlier practice for our GDQ run.  Many single player strategies here go by the wayside in favor of consistent, predictable, collision-free pathing.  I am proud to say that we have finally produced a run that is clean and unambiguously worthy of SDA publication.

Level Commentary:

Practice Race:

A small time loss here of about 1.5 seconds and the only death for red (played by Elipsis) in the entire run.  It's the usual offscreen death that top runners are all too familiar with.  We elected to keep the run going in these cases because of the discovery of a relatively fast and completely reliable backup strat that also shows off the alternate path.

Beginner Race:

It's easy to slap the balls together here if you aren't careful, and this run showcases very tight movement as well as some of the careful planning that yelsraek and I did on the backend.  Of particular interest is the almost telepathic link between the two of us at the bottom of the final section, where you see red make an adjustment with nearly the same inputs to give blue a run at the goal without a collision.

Intermediate Race:

This one is super fun because yelsraek and I are actually racing.  Specifically, whoever goes down the pipe _second_ has to perform the difficult "hot drop" maneuver - new technology that is only viable in the two player run of Marble Madness that allows the second marble onto the carpet to make up some time by suiciding into the goal.  It saves  time, looks cool as hell, is hard to do, and still technically doesn't count as a death.  I'm really happy that we got to show it off in this run.

Aerial Race:

Blue intentionally hesitates here in order to desynch the marbles and eliminate the chance of a game freeze.  This delay doesn't even cause a time loss, as the catapault has a set cooldown time and will always seperate the marbles by a second or so.  Blue now has the difficult job of going through the pegs on an unusual cycle, since red activates the pegs first.  The hammers aren't as predictable in co-op (because we aren't frame-rule level consistent, just yet!) so there is just a tiny hestitation to make sure that both of us get to the goal safely.

Silly Race:

With perfect play, red has the slightly longer track when the two diverge - and a slightly more optimized movement allowed yelsraek to get a bit ahead of me by the time we were clear of the bird section.  But BEHOLD the power of the pixel-perfect silly skip!  I was able to guide the red marble through a beautiful one-bounce silly skip here and make up all of that time and more.  This was an important time save because it came from the trailing marble, and allowed us to proceed on just as soon as the blue marble crossed the goal line.

Ultimate Race:

This one is always frightening.  The red track is hot garbage and always threatens a time loss.  yelsraek hangs back on the blue marble at the start just a tiny bit as planned in order to avoid a ramp collision.  yelsraek gets down to Steely (the black marble) first and begins the age old ritual of deadly combat.  This may seem unneccessary, but it turns out that the game WILL freeze if both players enter the final platforming section while the black marble is still on screen.  (Don't believe me?  Watch CGDQ!)  yelsraek trades his life (the only blue death) for a Steely kill, but believe it or not this isn't even a time loss.  My job as red at this point becomes to immediately get the platform cycles started by scrolling the screen, and as long as yelsraek can make the first cycle - it doesn't cost us time.

On the other hand, having both marbles share a single cycle ultimate race finish is _very_ nerve racking.  We lost a lot of runs here because there just isn't much room to make the platforms and not collide with each other.  Fortunately, tons of practice paid off, and we were able to complete the double one-cycle without a yelschoke.

Final Time: 3:07.704

I am extremely happy with this run.  We hit all the high notes (hot drop intermediate, silly skip, double one-cycle ultimate) and avoided significant mistakes.  This blows away the number two and three times (which also involve yelsraek and I with other partners) and creates a monsterous (in marble terms) 10 second gap in the leaderboard.  That figure alone should tell you something about how much faster this run is than anything that came before it.  Without the more minor errors in Practice and Intermediate, I think we'd be looking at about a 3:04 as the best reasonably possible time - but it's hard to say decisively, as the two player space has not been nearly as thoughly explored.  Still, I love this run and hope you all enjoy watching it as much as we did making it.

Much thanks and love to my Marble Brother yelsraek for pushing this game to the next level and collaborating with me to make this incredible time a reality.