WEBVTT Kind: captions Language: en 00:00:06.930 --> 00:00:11.050 When something is titled “The World’s Hardest Game,” it certainly has a lot to 00:00:11.050 --> 00:00:14.290 live up to. 00:00:14.290 --> 00:00:20.539 On December 15th, 2007, a 16-year-old developer named Stevie Critoph, who went by the online 00:00:20.539 --> 00:00:25.500 alias Snubby, released a very special game to his website, snubbyland.com. 00:00:25.500 --> 00:00:30.460 At the age of 12, he started making short flash games and cartoons, uploading much of 00:00:30.460 --> 00:00:35.620 his work to both Newgrounds and Snubby Land to share his creative passion with the world. 00:00:35.620 --> 00:00:40.840 His games were simple and his animations were crude, but there was certainly something special 00:00:40.840 --> 00:00:44.640 brewing at Snubby Land creative media. 00:00:44.640 --> 00:00:49.059 In late 2007, Snubby spent several months piecing together a project that would be very 00:00:49.059 --> 00:00:52.210 different from anything that he had created before. 00:00:52.210 --> 00:00:57.079 He wanted to make a flash game so difficult that calling it “The World’s Hardest Game” 00:00:57.079 --> 00:00:59.129 wouldn’t be seen as a joke. 00:00:59.129 --> 00:01:00.740 The objective was simple. 00:01:00.740 --> 00:01:06.510 For 30 levels, one needed to maneuver a red square from one green zone to another, dodging 00:01:06.510 --> 00:01:11.700 the blue, purple, and red enemies, while collecting all the yellow coins along the way. 00:01:11.700 --> 00:01:16.229 Upon death, the player got sent back to either the beginning of the level or the last checkpoint 00:01:16.229 --> 00:01:18.799 with all collected coins wiped away. 00:01:18.799 --> 00:01:22.869 The game wasn’t too difficult at the start, allowing most people to quickly clear the 00:01:22.869 --> 00:01:28.130 first few levels, but soon after, players would realize exactly what kind of monster 00:01:28.130 --> 00:01:29.590 they were up against. 00:01:29.590 --> 00:01:34.630 By and large, the levels only get harder as the game progresses, and the game must essentially 00:01:34.630 --> 00:01:40.189 be completed in one sitting, as it does not save people’s progress; closing or reloading 00:01:40.189 --> 00:01:44.180 the browser tab means starting everything over from the beginning. 00:01:44.180 --> 00:01:48.180 If someone gave up trying to complete the last level, they would have to slog through 00:01:48.180 --> 00:01:51.990 the first 29 levels again just to get another shot at it. 00:01:51.990 --> 00:01:57.180 Upon completing each level, the game taunts the player with a new, snarky message, seeking 00:01:57.180 --> 00:02:01.299 to eat away at their confidence, and as the death counter in the top right creeps higher 00:02:01.299 --> 00:02:06.430 and higher, the immense frustration felt after failing a level just before the end is only 00:02:06.430 --> 00:02:07.430 compounded. 00:02:07.430 --> 00:02:09.940 The game truly pulls no punches. 00:02:09.940 --> 00:02:15.261 Upon release, it truly may have been the most difficult, but still completable, flash game 00:02:15.261 --> 00:02:16.320 ever created. 00:02:16.320 --> 00:02:21.970 So, that raises the question: Who was the first player to conquer “The World’s Hardest 00:02:21.970 --> 00:02:23.250 Game?” 00:02:23.250 --> 00:02:26.980 Upon the game’s release, Snubby Land was a website of pretty modest size. 00:02:26.980 --> 00:02:31.670 There were about 50 registered members who had contributed around 1,000 posts to the 00:02:31.670 --> 00:02:37.150 forums, around 400 of which had come from Snubby himself, and about 250 that had come 00:02:37.150 --> 00:02:42.730 from someone named mexifry895, the website’s second most-active contributor. 00:02:42.730 --> 00:02:47.420 He would be the first to leave a comment about the game on the 22nd of December, one week 00:02:47.420 --> 00:02:52.200 after its release, venting his frustration with level 10, a definite choke point that 00:02:52.200 --> 00:02:55.360 is noticeably harder than all the previous levels. 00:02:55.360 --> 00:02:59.290 The next day, someone named Ziggy claimed to make it to level 15. 00:02:59.290 --> 00:03:04.371 Mexifry gave more updates about his progress, and after struggling with level 22, Snubby 00:03:04.371 --> 00:03:05.879 jumped in to lend a hand: 00:03:05.879 --> 00:03:07.300 “Just time it right. 00:03:07.300 --> 00:03:08.739 Move between the boxes. 00:03:08.739 --> 00:03:13.780 You move at the exact same speed as the enemies so just wait for the opening and go through.” 00:03:13.780 --> 00:03:14.780 - Snubby 00:03:14.780 --> 00:03:20.549 He got stuck on level 23 for three days, and after finally clearing it, he swiftly labeled 00:03:20.549 --> 00:03:23.860 level 24 as being “impossible.” 00:03:23.860 --> 00:03:29.110 It appeared as though the rest of the game was going to be a complete slog, but, by comparison, 00:03:29.110 --> 00:03:31.390 it ended up being pretty smooth sailing. 00:03:31.390 --> 00:03:37.500 On December 28th, with a final attempt session that lasted about 4 hours and recorded 689 00:03:37.500 --> 00:03:42.680 deaths, mexifry895 became the first victor of The World’s Hardest Game. 00:03:42.680 --> 00:03:47.470 However, the leaderboard showed that the developer was still quite a bit ahead when it came to 00:03:47.470 --> 00:03:48.470 skill. 00:03:48.470 --> 00:03:53.430 Mexifry urged Snubby to post the game on Newgrounds so it could get more attention; the game was 00:03:53.430 --> 00:03:58.430 a clear hit with the few people that had the opportunity to try it out, and it certainly 00:03:58.430 --> 00:04:01.340 had the potential to become a smashing success. 00:04:01.340 --> 00:04:06.960 By March of 2008, the game had around 10 different victors, and someone named terrence had managed 00:04:06.960 --> 00:04:11.370 to beat the game in just 15 deaths, 4 less than Snubby’s best. 00:04:11.370 --> 00:04:16.820 An obvious question began to loom; was it possible for someone to conquer this titan 00:04:16.820 --> 00:04:19.320 without dying a single time? 00:04:19.320 --> 00:04:24.320 On March 21st, the game was published to its sponsoring website, armorgames.com, where 00:04:24.320 --> 00:04:27.290 it was finally accessible to a broader audience. 00:04:27.290 --> 00:04:33.410 In just 3 days, the game reached 60,000 plays, and it quickly began spreading to other prominent 00:04:33.410 --> 00:04:36.310 websites like Newgrounds and Addicting Games. 00:04:36.310 --> 00:04:40.009 By the end of the year, that play count was over 3,000,000. 00:04:40.009 --> 00:04:45.620 The game’s easy-to-pick-up but tough-as-nails nature was able to swiftly hook countless 00:04:45.620 --> 00:04:51.060 players who wanted to see how difficult The World’s “Hardest” Game truly was. 00:04:51.060 --> 00:04:57.199 However, what happens when you mix a game that 99.9% of players give up trying to beat 00:04:57.199 --> 00:05:02.070 with a platform that makes it trivially easy for people to access all of a game’s assets 00:05:02.070 --> 00:05:17.030 and source code? 00:05:17.030 --> 00:05:21.980 People hacking popular flash games and slapping them on shady websites wasn’t exactly a 00:05:21.980 --> 00:05:27.160 new concept back in 2008, but there is no question that The World’s Hardest Game had 00:05:27.160 --> 00:05:30.289 it worse than anything that had come before it. 00:05:30.289 --> 00:05:34.101 Players were able to vent their frustration with the game’s difficulty by skipping straight 00:05:34.101 --> 00:05:36.590 to the end of each level completely unharmed. 00:05:36.590 --> 00:05:41.699 Countless people effortlessly skipped to the end and uploaded a score of “0 deaths” 00:05:41.699 --> 00:05:43.730 to the game’s official leaderboard. 00:05:43.730 --> 00:05:48.260 Indeed, within a week of the game being shared with the general public, the leaderboard was 00:05:48.260 --> 00:05:49.819 completely ruined. 00:05:49.819 --> 00:05:53.840 If someone wanted to prove that they got a 0 death completion of the game, they were 00:05:53.840 --> 00:05:56.150 going to need to get it on video. 00:05:56.150 --> 00:06:01.000 As I do with pretty much every single one of my videos nowadays, I spent hours searching 00:06:01.000 --> 00:06:06.190 YouTube to try and find the earliest recordings that I could of full completions of The World’s 00:06:06.190 --> 00:06:07.190 Hardest Game. 00:06:07.190 --> 00:06:12.000 To probably nobody’s surprise, I found a boatload of early videos claiming to complete 00:06:12.000 --> 00:06:16.950 the game in “0 deaths,” but were just recordings of that weird hacked version with 00:06:16.950 --> 00:06:19.970 smiley faces drawn on all of the green areas. 00:06:19.970 --> 00:06:21.320 Gotta love old YouTube. 00:06:21.320 --> 00:06:26.150 There was a lot of garbage to sift through, but I was able to find what I believe to be 00:06:26.150 --> 00:06:31.419 the first full playthrough of the game uploaded to YouTube, a 24 death completion performed 00:06:31.419 --> 00:06:35.530 by someone named HatredCopt3r on August 2nd, 2008. 00:06:35.530 --> 00:06:39.720 Because of time limit restrictions that existed on YouTube back in the day, Copt3r had to 00:06:39.720 --> 00:06:44.650 speed up their completion to get it uploaded, cranking it to “2x speed” as indicated 00:06:44.650 --> 00:06:45.949 by the title. 00:06:45.949 --> 00:06:50.430 However, upon analyzing the video, it became pretty clear that it was actually sped up 00:06:50.430 --> 00:06:55.770 by a factor of 1.5, making the completion a lot faster than some may have thought. 00:06:55.770 --> 00:07:00.480 Though Copt3r hadn’t completed the game in the least amount of deaths accurately recorded, 00:07:00.480 --> 00:07:05.410 they did unintentionally claim the record for the fastest completion with video evidence, 00:07:05.410 --> 00:07:08.750 a time clocking in at 13:13. 00:07:08.750 --> 00:07:12.940 The next day, they were able to cut their deaths personal best in half, completing the 00:07:12.940 --> 00:07:21.509 game with just 12: one in level 12, one in 19, one in 23, one in 25, two in 26, one in 28, 00:07:21.509 --> 00:07:24.160 two in 29, and two in 30. 00:07:24.160 --> 00:07:29.230 This run was also sped up 1.5x, and slowing it back down to the proper speed yields a 00:07:29.230 --> 00:07:32.820 final time of 12:23, a new world record. 00:07:32.820 --> 00:07:37.290 Copt3r’s completions of the game were certainly impressive, but it was obvious that there 00:07:37.290 --> 00:07:40.570 was still some room to improve, both death-wise and time-wise. 00:07:40.570 --> 00:07:46.840 Thankfully, on October 18th, 2008, the game would finally get the completion that it deserved, 00:07:46.840 --> 00:07:51.290 as someone named Hachimen became the first to record a 0 death completion of the World’s 00:07:51.290 --> 00:08:26.920 Hardest Game. 00:08:26.920 --> 00:08:31.379 Surpassing millions of other players, Hachimen had become the master of the World’s Hardest 00:08:31.379 --> 00:08:35.839 Game, completing the game with 0 deaths 10 months after its release. 00:08:35.839 --> 00:08:40.150 If you have ever seen a video of this game on YouTube before, you have almost certainly 00:08:40.150 --> 00:08:45.140 watched either this playthrough or another one that we will get to shortly, as it currently 00:08:45.140 --> 00:08:50.580 holds the record for the most viewed video of the game with over 17 million well-deserved 00:08:50.580 --> 00:08:51.580 views. 00:08:51.580 --> 00:08:56.180 Hachimen eased into each level with composure, executing mostly pre-determined movements 00:08:56.180 --> 00:08:59.649 with surgical precision without going too overboard. 00:08:59.649 --> 00:09:03.930 Time was somewhat of the essence, but speed was not the ultimate goal; keeping that death 00:09:03.930 --> 00:09:09.510 counter unchanged was all that Hachimen needed to do to qualify the playthrough as perfection. 00:09:09.510 --> 00:09:13.000 If someone wanted to prove that they were the new champion of The World’s Hardest 00:09:13.000 --> 00:09:18.040 Game, simply completing the game with 0 deaths was no longer going to cut it. 00:09:18.040 --> 00:09:22.740 Hachimen’s completion just happened to clock in at 8 minutes and 37 seconds. 00:09:22.740 --> 00:09:27.019 Now, it was time to see how much faster the game could get. 00:09:27.019 --> 00:09:30.880 In July of 2008, Snubby released the next installment in the WHG series, The World’s 00:09:30.880 --> 00:09:35.980 Hardest Game 2, featuring 50 total levels, some new mechanics, a heightened level of 00:09:35.980 --> 00:09:40.520 difficulty, and music that was composed by none other than the first installment’s 00:09:40.520 --> 00:09:43.440 first victor, mexifry895. 00:09:43.440 --> 00:09:49.850 The game is so challenging that, to this day, nobody has managed to complete it in 0 deaths, 00:09:49.850 --> 00:09:55.130 with the current Glitchless world record of 12:28 by Danicker still having 7 of them. 00:09:55.130 --> 00:10:00.220 While both the sequel and the original continued to receive massive amounts of attention going 00:10:00.220 --> 00:10:06.430 into 2009, it seemed like nobody was able to top Hachimen’s record of 8:37. 00:10:06.430 --> 00:10:09.750 People may have thought that there wasn’t much left to accomplish with the game if the 00:10:09.750 --> 00:10:13.990 death floor of 0 had already been reached, since people didn’t really care about how 00:10:13.990 --> 00:10:15.980 fast they were able to complete the game. 00:10:15.980 --> 00:10:22.160 However, on September 28th, 2009, someone named Shiroma would prove everyone wrong, 00:10:22.160 --> 00:10:27.250 as he uploaded a 0 death completion of the game that clocked in two seconds faster than 00:10:27.250 --> 00:10:29.610 Hachimen’s record: 8:35. 00:10:29.610 --> 00:10:34.570 Shiroma just started as an average World’s Hardest Game player, initially giving up trying 00:10:34.570 --> 00:10:37.180 to beat the game after struggling immensely. 00:10:37.180 --> 00:10:42.790 However, he was eventually able to complete the game in 450 deaths, whereafter, he started 00:10:42.790 --> 00:10:46.839 beating it over and over again to see how much better he could get at it. 00:10:46.839 --> 00:10:52.730 Each level got easier and easier, and eventually, he was able to piece together his own 0 death 00:10:52.730 --> 00:10:53.730 completion. 00:10:53.730 --> 00:10:58.040 With no built-in goal left to accomplish, he decided to try speedrunning the game. 00:10:58.040 --> 00:11:02.310 He found the video of Hachimen’s 0 death completion and used it as a reference for 00:11:02.310 --> 00:11:05.089 how he could more efficiently complete the game. 00:11:05.089 --> 00:11:10.710 After some practice and attempts, he was eventually able to produce the 8:35 as his first run. 00:11:10.710 --> 00:11:16.200 Expectedly, it was very similar to Hachimen’s 8:37, and though they took somewhat different 00:11:16.200 --> 00:11:21.010 approaches in a lot of the levels, the overall theme of taking extra time to increase the 00:11:21.010 --> 00:11:23.470 odds of success was readily apparent. 00:11:23.470 --> 00:11:28.420 The best example of this is probably the long wait at the start of level 23 to get a more 00:11:28.420 --> 00:11:30.389 favorable cycle on the spinner. 00:11:30.389 --> 00:11:35.459 By including a timer with the video and labeling the completion as a “speedrun” in the 00:11:35.459 --> 00:11:40.340 title, it was clear that Shiroma wanted to fully prioritize how fast he could complete 00:11:40.340 --> 00:11:42.080 “The World’s Hardest Game”. 00:11:42.080 --> 00:11:46.970 He had gotten another 0 death completion and the world record, but he was certainly not 00:11:46.970 --> 00:11:48.540 yet done with the game. 00:11:48.540 --> 00:11:52.779 For two months, Shiroma sought to improve his approach to each and every level until 00:11:52.779 --> 00:11:55.520 it was as fast as it could possibly be. 00:11:55.520 --> 00:11:59.770 Practicing the later levels was arduous, as there was no way for him to navigate to a 00:11:59.770 --> 00:12:04.760 specific level; if he wanted to get another stab at something, he had to go back to the 00:12:04.760 --> 00:12:08.889 main menu and play through every single level that came before it. 00:12:08.889 --> 00:12:10.860 But he didn’t let that deter him. 00:12:10.860 --> 00:12:16.850 He wanted the next world record to be a huge step up, not just another 1-2 second improvement. 00:12:16.850 --> 00:12:22.290 Finally, on November 30th, 2009, he had pieced together the completion that he was looking 00:12:22.290 --> 00:12:23.430 for. 00:12:23.430 --> 00:12:27.279 Starting around level 9, the gameplay was noticeably better. 00:12:27.279 --> 00:12:32.320 Instead of waiting before each rapidly moving obstacle, Shiroma nearly bulldozed through 00:12:32.320 --> 00:12:34.790 the entire level without slowing down. 00:12:34.790 --> 00:12:39.620 In level 12, he skipped an entire cycle at the start, eventually equating to a level 00:12:39.620 --> 00:12:42.480 completion that was nearly twice as fast. 00:12:42.480 --> 00:12:48.170 At the start of level 14, there was absolutely no waiting, Shiroma just dug right in. 00:12:48.170 --> 00:12:52.660 In level 17, he saved about 5 seconds by skipping a cycle at the end. 00:12:52.660 --> 00:12:58.680 And in level 23, he didn’t wait for an easier cycle; Shiroma cut the completion time for 00:12:58.680 --> 00:13:01.199 yet another level in half. 00:13:01.199 --> 00:13:06.481 The strategy for the end of level 24 was much improved, the inaccuracies in level 26 were 00:13:06.481 --> 00:13:12.190 cleaned up, level 27 had a much faster approach, the lines taken in level 28 were far more 00:13:12.190 --> 00:13:18.630 optimal, level 29 was played with absolutely no fear, and level 30 was optimal as always. 00:13:18.630 --> 00:13:20.230 The final time? 00:13:20.230 --> 00:13:21.370 6:41. 00:13:21.370 --> 00:13:25.449 Shiroma had improved his record by nearly two whole minutes. 00:13:25.449 --> 00:13:30.470 With this run, Shiroma proved that there was nobody else who was even close to his skill 00:13:30.470 --> 00:13:31.940 level with the game. 00:13:31.940 --> 00:13:36.589 He cut through each level like butter, performing movements that probably looked impossible 00:13:36.589 --> 00:13:39.589 to anyone viewing back in 2009. 00:13:39.589 --> 00:13:44.350 If he quit now, it would almost certainly be years before anybody could hope to topple 00:13:44.350 --> 00:13:45.350 his record. 00:13:45.350 --> 00:13:47.950 However, he still had more to do. 00:13:47.950 --> 00:13:53.519 On December 11th, 2009, Shiroma claimed the next record in his conquest to completely 00:13:53.519 --> 00:13:55.700 dominate The World’s Hardest Game. 00:13:55.700 --> 00:14:00.720 Looking at the run, there was not a single facet that could be pointed out as being suboptimal. 00:14:00.720 --> 00:14:07.459 The new path taken in level 24 was just ridiculous, and all of the strategies used in his 6:41 00:14:07.459 --> 00:14:10.199 had seemingly been executed to perfection. 00:14:10.199 --> 00:14:15.220 Up to this point, Shiroma was only uploading his runs to the Japanese video-sharing site 00:14:15.220 --> 00:14:16.750 Niconico, where they saw some moderate success. 00:14:16.750 --> 00:14:20.870 However, this time, he decided to upload his achievement to YouTube so he could share it 00:14:20.870 --> 00:14:25.190 with a wider audience, a decision that would eventually pay off. 00:14:25.190 --> 00:14:30.089 It was a steady performer at first, but after a few years of simmering, the view count just 00:14:30.089 --> 00:14:34.620 started soaring, reaching over 15 million views and counting. 00:14:34.620 --> 00:14:40.250 So yes, this is the other highly-viewed World’s Hardest Game video that was alluded to earlier, 00:14:40.250 --> 00:14:43.639 nearly matched with Hachimen’s 17 million view completion. 00:14:43.639 --> 00:14:48.310 It’s pretty amazing that two of the most viewed speedruns ever performed were of a 00:14:48.310 --> 00:14:54.370 flash game made by a 16-year-old in his spare time; the internet truly does work in some 00:14:54.370 --> 00:14:56.160 pretty incredible ways. 00:14:56.160 --> 00:15:01.820 On a personal note, Shiroma’s 6:29 was the very first speedrun I ever remember watching 00:15:01.820 --> 00:15:07.509 back in the early 2010s, and I recall just sitting there in complete disbelief. 00:15:07.509 --> 00:15:12.029 This was a game where I was taking hundreds of deaths to get past just the first chunk 00:15:12.029 --> 00:15:17.009 of levels, and this guy was able to complete the whole thing in six and a half minutes 00:15:17.009 --> 00:15:19.460 without dying a single time. 00:15:19.460 --> 00:15:24.070 This is the run that initially built my curiosity when it came to speedrunning, as I learned 00:15:24.070 --> 00:15:28.660 that the simple flash games that I enjoyed playing had so much potential just waiting 00:15:28.660 --> 00:15:30.180 to be unlocked. 00:15:30.180 --> 00:15:34.480 Without this run, I don’t know if I’d be in the same spot in the speedrunning community 00:15:34.480 --> 00:15:36.350 that I am today. 00:15:36.350 --> 00:15:41.600 It seemed like Shiroma had finally perfected his craft, but he knew there was still one 00:15:41.600 --> 00:15:44.320 level in particular that needed to be improved. 00:15:44.320 --> 00:15:45.850 The target? 00:15:45.850 --> 00:15:47.089 Level 18. 00:15:47.089 --> 00:15:51.410 In all of his previous records, the strategy had always been the same. 00:15:51.410 --> 00:15:56.190 Clear three rows of coins going to the right, clear one row going back to the left, and 00:15:56.190 --> 00:16:00.180 clear the remaining three moving right again towards the goal. 00:16:00.180 --> 00:16:05.250 Even if all the rows were able to be cleared in two trips across the level instead of three, 00:16:05.250 --> 00:16:09.779 a final movement to the right would still be required to reach the goal, meaning that 00:16:09.779 --> 00:16:12.230 the strat couldn’t be done any more optimally. 00:16:12.230 --> 00:16:18.390 However, on January 11th, 2010, he got to show off a brand new method that effectively 00:16:18.390 --> 00:16:31.149 cut the number of trips down to one. 00:16:31.149 --> 00:16:36.890 Being nearly tied with his 6:29 going into level 18, he left 8 seconds ahead. 00:16:36.890 --> 00:16:41.930 Shiroma figured out that by buffering up and right at the start of level 19, the player 00:16:41.930 --> 00:16:47.930 would just not hit any of the enemies, allowing him to complete it without ever slowing down. 00:16:47.930 --> 00:16:52.410 Combined with slightly more optimal movement in the last 10 levels, Shiroma grew his lead 00:16:52.410 --> 00:16:55.440 to 12 seconds and scored himself a 6:17. 00:16:55.440 --> 00:16:57.750 Now, this had to be it. 00:16:57.750 --> 00:17:01.639 2nd place was 2 minutes and 20 seconds behind him. 00:17:01.639 --> 00:17:04.350 Each level looked like it had been perfected. 00:17:04.350 --> 00:17:29.350 What else could possibly be done to push the time any lower? 00:17:29.350 --> 00:17:53.770 14 more seconds erased. 00:17:53.770 --> 00:17:59.220 Shiroma had somehow figured out yet another assortment of new strategies all on his own, 00:17:59.220 --> 00:18:01.510 each one being some sort of cycle skip. 00:18:01.510 --> 00:18:06.799 In level 3, instead of waiting after collecting the coin as intended, Shiroma was able to 00:18:06.799 --> 00:18:11.830 barely maneuver back to the goal in the same cycle, barely dodging the enemies on both 00:18:11.830 --> 00:18:14.260 the way in and on the way out. 00:18:14.260 --> 00:18:19.690 In level 17, he figured out that it was barely possible to move ahead of the enemy by buffering 00:18:19.690 --> 00:18:24.180 down and right at the start of the level, which doesn’t equate to any timesave because 00:18:24.180 --> 00:18:29.140 of how the cycles work out later in the level but does make for a pretty funny swag strat. 00:18:29.140 --> 00:18:34.030 In level 25, he completely cut out all of the waiting by buffering down at the start 00:18:34.030 --> 00:18:36.980 and barely squeezing through the spinner a cycle early. 00:18:36.980 --> 00:18:42.300 Finally, in 27, he figured out that it was possible to reach the goal from the top corner 00:18:42.300 --> 00:18:46.970 with the bar still in front of it, skipping having to wait for it to move back up. 00:18:46.970 --> 00:18:52.440 With tighter execution in some of the other levels, Shiroma set a new record of 6:03 on 00:18:52.440 --> 00:18:54.690 March 31st, 2010. 00:18:54.690 --> 00:18:57.690 This was finally the run that he had been looking for. 00:18:57.690 --> 00:19:03.299 2 minutes and 34 seconds ahead of 2nd place, he finally decided to take a rest atop his 00:19:03.299 --> 00:19:04.299 throne. 00:19:04.299 --> 00:19:10.320 He had completely dismantled “The World’s Hardest Game” almost entirely on his own, 00:19:10.320 --> 00:19:15.000 pushing the game to heights that were inconceivable for many back in 2010. 00:19:15.000 --> 00:19:20.060 Back then, it seemed obvious that this record was going to stand for a long while, but would 00:19:20.060 --> 00:19:40.230 it truly be able to stand the test of time? 00:19:40.230 --> 00:20:02.450 7 years passed by, and nobody had even come close. 00:20:02.450 --> 00:20:08.700 It was the real deal; Shiroma’s record was truly almost a decade ahead of its time. 00:20:08.700 --> 00:20:13.679 There were multiple 0 death completions uploaded to YouTube during this span, but in terms 00:20:13.679 --> 00:20:21.230 of speed, Pzula’s 7:09, performed on March 21st, 2017, was the closest anyone had gotten, 00:20:21.230 --> 00:20:24.799 a run which unfortunately appears to have been lost to time. 00:20:24.799 --> 00:20:29.299 However, thanks to this passage of time, Pzula had access to the coolmath version of the 00:20:29.299 --> 00:20:35.150 game, released in 2013, which had a built-in level selector, allowing him to practice the 00:20:35.150 --> 00:20:38.230 game much more efficiently than Shiroma ever could. 00:20:38.230 --> 00:20:42.169 For two weeks, Pzula poured in attempts, inching closer to the record. 00:20:42.169 --> 00:20:46.690 By the 15th of April, he had a 6:10, just seven seconds away. 00:20:46.690 --> 00:20:51.909 He got onto a run that started off a bit slower than Shiroma, but after implementing a more 00:20:51.909 --> 00:20:55.410 optimal route in level 8, he was able to pull a bit ahead. 00:20:55.410 --> 00:21:00.530 Pzula saved a bit more time after implementing a new death defying strategy in the middle 00:21:00.530 --> 00:21:01.620 of level 10. 00:21:01.620 --> 00:21:06.110 The lead expanded further after he cut straight towards the goal after collecting the last 00:21:06.110 --> 00:21:09.600 coin in level 16 instead of moving around the enemy. 00:21:09.600 --> 00:21:15.030 In level 18, Pzula used a slightly different path that was a bit easier but about the same 00:21:15.030 --> 00:21:16.030 speed, and in level 21, he narrowly avoided death between these two enemies to collect 00:21:16.030 --> 00:21:17.030 the coin a bit faster. 00:21:17.030 --> 00:21:21.820 He took a cleaner path through level 24, and had a 4 second lead going into level 26, putting 00:21:21.820 --> 00:21:27.340 him on pace for the first sub 6 in the game’s history. 00:21:27.340 --> 00:21:32.280 And then… the lead was wiped away. 00:21:32.280 --> 00:21:35.400 Pzula exited the level a second behind. 00:21:35.400 --> 00:21:38.850 However, he had one final trick up his sleeve. 00:21:38.850 --> 00:21:43.940 Going through the middle of level 27 was certainly a lot more difficult, but it could cut out 00:21:43.940 --> 00:21:47.760 a lot of waiting and unnecessary movement up to the coin. 00:21:47.760 --> 00:21:52.669 Pzula executed the new route almost flawlessly and pulled two and a half seconds ahead of 00:21:52.669 --> 00:21:53.740 Shiroma. 00:21:53.740 --> 00:21:58.460 He matched Shiroma’s death defying movement in level 28, and took a more optimal path 00:21:58.460 --> 00:22:02.350 in 29, which grew the lead back to 4 seconds. 00:22:02.350 --> 00:22:07.640 Pzula, once again, had the potential to break what was almost certainly the game’s final 00:22:07.640 --> 00:22:08.970 minute barrier. 00:22:08.970 --> 00:22:13.090 All he had to do was hug the wall to the finish. 00:22:13.090 --> 00:22:14.750 And then… 00:22:14.750 --> 00:22:20.000 Staring death in the face, he decided not to risk it and moved around the enemy, claiming 00:22:20.000 --> 00:22:22.760 a new record of 6 minutes flat. 00:22:22.760 --> 00:22:29.159 After 2572 days, Shiroma’s record had finally been beaten. 00:22:29.159 --> 00:22:35.140 Unfortunately, for Pzula, a deathless run and a sub 6 were going to have to wait. 00:22:35.140 --> 00:22:39.510 After taking a short break, Pzula went back to doing attempts with some new strategies 00:22:39.510 --> 00:22:40.730 under his belt. 00:22:40.730 --> 00:22:46.030 He got onto a run that had a strong first 5 levels but lost around 2 seconds in level 00:22:46.030 --> 00:22:47.429 6 from a quick death. 00:22:47.429 --> 00:22:52.490 However, he pulled right back even after implementing a new approach in level 7. 00:22:52.490 --> 00:22:54.049 But, hold on. 00:22:54.049 --> 00:22:58.659 How was he able to just phase through the enemies while moving around at the bottom 00:22:58.659 --> 00:22:59.820 of the level? 00:22:59.820 --> 00:23:04.030 If you’ve ever played this game before, you may know about a weird glitch in level 00:23:04.030 --> 00:23:08.500 7 where you can seemingly stay invincible if you just stay at the same y level that 00:23:08.500 --> 00:23:09.990 the game starts you out at. 00:23:09.990 --> 00:23:14.270 If you move up or down, even a little bit, the enemies can hit you once again. 00:23:14.270 --> 00:23:18.559 You still have to collect the coins, meaning that you can’t just cheese the entire level 00:23:18.559 --> 00:23:22.980 by holding down the right button, but it is still a pretty fun thing to mess around with. 00:23:22.980 --> 00:23:28.390 The position Pzula gets to by tapping up once from the bottom of the level just so happens 00:23:28.390 --> 00:23:32.529 to be another one of these seemingly “random” invincibility zones. 00:23:32.529 --> 00:23:37.890 So, here’s the flaw with the game’s collision detection that allows this to happen: 00:23:37.890 --> 00:23:42.750 The player has five hit detection points arranged in a cross pattern, one in the middle of each 00:23:42.750 --> 00:23:45.211 edge, and one in the center. 00:23:45.211 --> 00:23:49.240 While coin collision is determined with a simple bounding box check, for the game to 00:23:49.240 --> 00:23:55.340 count an interaction with an enemy as a hit, the enemy must overlap one of these 5 test 00:23:55.340 --> 00:23:56.340 points. 00:23:56.340 --> 00:24:00.610 This means that the player can move quite a bit inside an enemy from the corner without 00:24:00.610 --> 00:24:05.780 taking a hit, while going in from the edge spells instant doom, facets that were likely 00:24:05.780 --> 00:24:07.560 intended by the developer. 00:24:07.560 --> 00:24:12.650 However, this system starts to break down because of flaws that exist in Flash Player 00:24:12.650 --> 00:24:13.750 itself. 00:24:13.750 --> 00:24:18.840 As it turns out, for a hit to be registered, an enemy must essentially overlap the same 00:24:18.840 --> 00:24:24.150 one of these five points for two frames in a row, a phenomenon that is very similar to 00:24:24.150 --> 00:24:26.510 how spike glitch in Red Ball works. 00:24:26.510 --> 00:24:31.240 In flash player, point to pixel checks are judged based on whatever the previously rendered 00:24:31.240 --> 00:24:32.240 frame was. 00:24:32.240 --> 00:24:37.260 Let’s say we are in a scenario where an enemy moves into the player’s right contact 00:24:37.260 --> 00:24:38.260 point. 00:24:38.260 --> 00:24:42.500 The game first checks if the point is within the bounding box of the enemy, as this check 00:24:42.500 --> 00:24:47.660 is orders of magnitude faster than performing a full point to pixel check on every single 00:24:47.660 --> 00:24:51.890 frame regardless of how far apart the player and enemy are. 00:24:51.890 --> 00:24:56.540 Since the point is inside the bounding box, the game goes ahead and performs the full 00:24:56.540 --> 00:24:59.159 point to pixel check with the enemy. 00:24:59.159 --> 00:25:04.441 However, because the previously rendered frame still looks like this, the check fails, as 00:25:04.441 --> 00:25:08.380 no pixels from the rendered enemy are overlapping the point. 00:25:08.380 --> 00:25:12.840 On the next frame, the enemy moves off of the player’s right contact point. 00:25:12.840 --> 00:25:17.630 The game first performs the bounding box check, but it expectedly fails. 00:25:17.630 --> 00:25:22.600 Therefore, the game doesn't run the full point to pixel check on the previously rendered 00:25:22.600 --> 00:25:25.970 frame that would have found the two to be overlapping. 00:25:25.970 --> 00:25:29.460 Gotta love those janky built-in Flash Player functions. 00:25:29.460 --> 00:25:34.910 Essentially, the only zones where an enemy can be lethal are ones where it overlaps itself 00:25:34.910 --> 00:25:37.150 for two consecutive frames. 00:25:37.150 --> 00:25:41.980 Because the enemies in level 7 are moving so fast, the regions of enemy overlap are 00:25:41.980 --> 00:25:46.960 small, meaning that as long as all five of the character's collision points are not in 00:25:46.960 --> 00:25:49.200 these areas, it cannot be hit. 00:25:49.200 --> 00:25:50.330 Extrapolating that to provide a better visual for level 7, as long as the character stays 00:25:50.330 --> 00:25:51.330 completely within these zones, it is safe. 00:25:51.330 --> 00:25:55.370 Unfortunately, because of how unusual this looked and how ubiquitous noclipped “completions” 00:25:55.370 --> 00:26:00.890 of the game were on YouTube, Pzula had plenty of people in the comments calling him a “hacker” 00:26:00.890 --> 00:26:06.560 and a “cheater,” a trend that unfortunately would not die out with this video. 00:26:06.560 --> 00:26:11.940 Pzula got a bit behind going into level 20, but he had fortunately figured out a way to 00:26:11.940 --> 00:26:14.520 save a boatload of time in the level. 00:26:14.520 --> 00:26:19.380 After collecting the first coin, it was just barely possible to squeeze behind this enemy 00:26:19.380 --> 00:26:25.100 without waiting for the next cycle, leading to about 7 seconds of time save, mostly thanks 00:26:25.100 --> 00:26:29.220 to the cycles later in the level lining up more favorably as a result. 00:26:29.220 --> 00:26:33.820 Going into level 21, Pzula was easily on pace for a sub 6. 00:26:33.820 --> 00:26:38.740 He maintained the same 5 and a half second lead going into level 26, and after completing 00:26:38.740 --> 00:26:42.780 the level without dying, he was now over 10 seconds ahead. 00:26:42.780 --> 00:26:49.309 This run was on pace to skip an entire ten second barrier, having the potential for a 00:26:49.309 --> 00:26:50.309 5:49. 00:26:50.309 --> 00:27:00.179 Unfortunately, in level 28, disaster struck. 00:27:00.179 --> 00:27:05.150 One death right at the start, and one death in the middle of the second section. 00:27:05.150 --> 00:27:08.049 That lead shrunk to just 4.4 seconds. 00:27:08.049 --> 00:27:12.529 Thankfully, the rest of the run was smooth sailing, and Pzula slipped under the last 00:27:12.529 --> 00:27:20.650 enemy in level 30 to claim a new record of 5:55, the first sub 6 in the game’s history. 00:27:20.650 --> 00:27:26.020 Pzula decided to take yet another break after attaining this milestone, but that didn’t 00:27:26.020 --> 00:27:28.880 mean the game as a whole was cooling off. 00:27:28.880 --> 00:27:33.880 Competition on the leaderboard was stronger than ever, as Pzula had inspired runners like 00:27:33.880 --> 00:27:37.120 hsblue to push their times lower and lower. 00:27:37.120 --> 00:27:44.490 Finally, on February 28th, 2018, Pzula came back with a 5:46, with almost all of the timesave 00:27:44.490 --> 00:27:47.710 coming from him cleaning up all three of his deaths. 00:27:47.710 --> 00:27:53.490 Indeed, Pzula had finally managed to complete his first ever deathless run of the game. 00:27:53.490 --> 00:27:58.760 Like had been the case with Shiroma’s 6:03, there was once again a run in 1st place that 00:27:58.760 --> 00:28:04.250 was nearly “perfect” for all of the strategies known at the time, with him only lacking the 00:28:04.250 --> 00:28:08.380 enemy weave in level 10 that he had pulled off in his 6 flat. 00:28:08.380 --> 00:28:11.660 This one had the potential to stand for quite a while. 00:28:11.660 --> 00:28:17.029 However, third place runner hsblue would soon discover something that could blow the game 00:28:17.029 --> 00:28:19.260 wide open once again. 00:28:19.260 --> 00:28:25.940 On March 7th, hsblue uploaded a new PB of 6:15 that had a pretty strange anomaly in 00:28:25.940 --> 00:28:26.940 level 17. 00:28:26.940 --> 00:28:31.690 He was able to partially clip into this wall while moving down towards the exit. 00:28:31.690 --> 00:28:34.269 So, how on earth does that work? 00:28:34.269 --> 00:28:39.250 Essentially, the game uses the same edge points to check for collisions with walls that it 00:28:39.250 --> 00:28:41.740 does to check for collisions with enemies. 00:28:41.740 --> 00:28:46.769 Expectedly, this means that it's possible to clip slightly into the corner of each wall, 00:28:46.769 --> 00:28:49.160 since none of the collision points are being hit. 00:28:49.160 --> 00:28:53.399 If the player moves into a position that puts one of the collision points out of bounds, 00:28:53.399 --> 00:28:58.059 the game will move the player at 3 pixels per frame, the game’s normal speed cap for 00:28:58.059 --> 00:29:01.390 movement, in the opposite direction until it is back in bounds. 00:29:01.390 --> 00:29:05.909 However, if the player starts holding the direction key opposite this force, they will 00:29:05.909 --> 00:29:11.049 be able to stay inside of the wall, as the two opposing 3 pixel per frame velocities 00:29:11.049 --> 00:29:12.500 will cancel out. 00:29:12.500 --> 00:29:17.590 This is exactly what happened in hsblue’s run; on this frame he was holding down and 00:29:17.590 --> 00:29:21.799 left, which caused him to clip slightly inside of the wall, and on this frame, he switched 00:29:21.799 --> 00:29:27.269 from holding left to holding right, allowing him to stay inside of it as he moved down. 00:29:27.269 --> 00:29:32.590 In this scenario, this maneuver was frame perfect, though he had done it by complete 00:29:32.590 --> 00:29:35.190 accident, and it didn’t really accomplish anything. 00:29:35.190 --> 00:29:40.340 However, the implications of this unusual mechanic soon became obvious. 00:29:40.340 --> 00:29:46.049 On August 22nd, hsblue was able to implement wall clipping in a useful fashion, as he claimed 00:29:46.049 --> 00:29:53.179 a perfect level 1 time of 3.867 seconds, a run which has now been tied by 21 other people. 00:29:53.179 --> 00:29:58.250 In the run, hsblue first clips into this corner by holding down and right, frame perfectly 00:29:58.250 --> 00:30:02.760 releases the down button to stay inside the wall, and then frame perfectly hits the up 00:30:02.760 --> 00:30:06.570 button to keep the player inside the wall while moving to the right. 00:30:06.570 --> 00:30:11.460 On this frame, the player moves three pixels to the right as usual but now also moves up 00:30:11.460 --> 00:30:15.679 three pixels, as the top collision point is no longer being activated. 00:30:15.679 --> 00:30:19.621 The game then detects that the player’s left collision point is touching a wall, so 00:30:19.621 --> 00:30:23.160 it moves the player 3 more pixels toward the right to force it out. 00:30:23.160 --> 00:30:29.060 Because of wall clipping, hsblue moves 3 pixels up and 6 pixels to the right in a single frame, 00:30:29.060 --> 00:30:34.870 double the game’s intended speed cap, which, in this instance, saves a single frame from 00:30:34.870 --> 00:30:36.669 two frame perfect inputs! 00:30:36.669 --> 00:30:42.250 Alright, so wall clipping wasn’t exactly a revolutionary new strategy yet, but with 00:30:42.250 --> 00:30:46.559 enough time to simmer, there was hope that someone would actually figure out how to save 00:30:46.559 --> 00:30:48.230 some decent time with it. 00:30:48.230 --> 00:30:52.940 However, while people were still searching for more applications for this new technique, 00:30:52.940 --> 00:30:57.290 Pzula decided to come back and completely shock the community. 00:30:57.290 --> 00:30:59.640 August 28, 2018. 00:30:59.640 --> 00:31:05.029 Pzula streams world record attempts of The World’s Hardest Game to a large audience 00:31:05.029 --> 00:31:06.510 of 0 viewers. 00:31:06.510 --> 00:31:12.960 His splits show that his current PB is a 5:31, 15 seconds faster than his leaderboard world 00:31:12.960 --> 00:31:16.320 record, though he never decided to upload the run. 00:31:16.320 --> 00:31:22.480 He is laser-focused on his mission of completing the game in less than 5 minutes and 30 seconds. 00:31:22.480 --> 00:31:26.760 He is playing on a version of the game that he has modified to include an in-game timer 00:31:26.760 --> 00:31:29.070 for each level and the whole run. 00:31:29.070 --> 00:31:34.360 With built-in timing, resetting, and level navigation functionality, this mod has been 00:31:34.360 --> 00:31:39.590 instrumental in helping him discover and practice the multitude of new strategies that he is 00:31:39.590 --> 00:31:41.210 seeking to implement. 00:31:41.210 --> 00:31:45.830 In level 3, he uses a position-based setup to save a single frame. 00:31:45.830 --> 00:31:51.029 In level 5, he does more moving and less waiting, as he allows one of the enemies to clip him 00:31:51.029 --> 00:31:52.070 on the corner. 00:31:52.070 --> 00:31:57.440 He hits a much faster cycle while collecting the last coin in level 7, and in level 10… 00:31:57.440 --> 00:32:00.030 he hits all three possible weaves. 00:32:00.030 --> 00:32:04.139 He sneaks under this enemy elevator in level 12 to save a bit of waiting. 00:32:04.139 --> 00:32:08.019 He bulldozes through level 13 without stopping a single time. 00:32:08.019 --> 00:32:13.480 He takes a significantly faster route in level 16 that involves a death defying third coin 00:32:13.480 --> 00:32:15.929 grab to save a hefty chunk of time. 00:32:15.929 --> 00:32:20.580 He collects all of the coins in the first two thirds of level 18 without stopping a 00:32:20.580 --> 00:32:25.110 single time, swiftly passing through the enemies as they move up and down. 00:32:25.110 --> 00:32:30.571 He cuts through these enemies in level 22 without stopping and completes level 24 without 00:32:30.571 --> 00:32:35.040 letting go of the right key until the very end, utilizing a new route that he discovered. 00:32:35.040 --> 00:32:40.120 He goes to the right first in level 26, which makes the cycles line up better for the rest 00:32:40.120 --> 00:32:41.120 of the level. 00:32:41.120 --> 00:32:43.630 The best possible time is 5:28. 00:32:43.630 --> 00:32:49.260 All he needs to do is complete the last 4 levels without dying, just like he has done 00:32:49.260 --> 00:32:51.256 countless times before. 00:33:23.014 --> 00:33:26.700 5:29. There it was. 00:33:26.700 --> 00:33:32.530 Shiroma’s 6:29, the most influential World’s Hardest Game speedrun ever performed, had 00:33:32.530 --> 00:33:35.159 just been toppled by over a minute. 00:33:35.159 --> 00:33:41.010 Again, the game had a world record where almost every single detail went exactly according 00:33:41.010 --> 00:33:42.010 to plan. 00:33:42.010 --> 00:33:46.549 The video received a good amount of traction, and many viewers and fellow runners offered 00:33:46.549 --> 00:33:49.140 their congratulations in the comments. 00:33:49.140 --> 00:33:54.620 However, there was also expectedly a tsunami of detractors that instantly called the run 00:33:54.620 --> 00:33:59.159 cheated once they saw Pzula getting hit by the enemies in level 7 without dying. 00:33:59.159 --> 00:34:05.730 I guess some people just aren’t able to fathom that a Flash game programmed by a 16-year-old 00:34:05.730 --> 00:34:10.650 back in 2007 may not have every single bolt tightened to spec. 00:34:10.650 --> 00:34:16.440 Regardless, this impressive show of dominance placed Pzula 31 seconds ahead of 2nd place, 00:34:16.440 --> 00:34:19.980 meaning that this one certainly had the chance to stand for a while. 00:34:19.980 --> 00:34:24.980 However, it didn’t end up staying on top for as long as most people may have expected, 00:34:24.980 --> 00:34:28.470 as a new runner named Aaron would quickly make himself known. 00:34:28.470 --> 00:34:33.840 Joining the community in September of 2018 with a youtube channel called “The_IL_Guy 00:34:33.840 --> 00:34:39.310 runs,” Aaron made his first mark with, well, Individual Level speedruns. 00:34:39.310 --> 00:34:45.560 He beat or tied the world record in every level aside from 3, 18, and 30, a very impressive 00:34:45.560 --> 00:34:46.560 feat. 00:34:46.560 --> 00:34:52.650 However, on November 18th, 2018, he dropped something even more substantial, a 2nd place 00:34:52.650 --> 00:34:55.740 full-game personal best of 5:46. 00:34:55.740 --> 00:34:59.570 The_IL_Guy was no longer just The_IL_Guy. 00:34:59.570 --> 00:35:02.890 He was now The_FG_Guy as well. 00:35:02.890 --> 00:35:10.770 Aaron kept grinding, improving his PB to a 5:39, then a 5:38, then a 5:30, and finally, 00:35:10.770 --> 00:35:13.660 a new world record of 5:26. 00:35:13.660 --> 00:35:18.340 In the run, Aaron took the other available path in level 9 because it turned out that 00:35:18.340 --> 00:35:20.480 it was actually slightly faster. 00:35:20.480 --> 00:35:26.090 In level 18, he used Pzula’s no stop strategy the whole way through, and in level 21, he 00:35:26.090 --> 00:35:30.500 maneuvered around the enemies after collecting the third coin instead of getting sandwiched 00:35:30.500 --> 00:35:31.710 in between them. 00:35:31.710 --> 00:35:37.050 He used a faster return route in level 27, and he pulled off a frame perfect collision 00:35:37.050 --> 00:35:40.640 cheese in level 29 to squeeze through these enemies early. 00:35:40.640 --> 00:35:46.081 He didn’t do the quick exit in level 7 or the risky third coin grab in level 16, but 00:35:46.081 --> 00:35:50.720 all of the other crazy stuff that he implemented allowed him to beat Pzula’s monumental record 00:35:50.720 --> 00:35:51.780 by three seconds. 00:35:51.780 --> 00:35:56.630 It’s worth reiterating that essentially none of these strategies could have been found 00:35:56.630 --> 00:36:00.230 and perfected without the help of Pzula’s speedrun mod. 00:36:00.230 --> 00:36:05.339 Back when Shiroma pulled off his 6:03, the game didn’t even have a level selector; 00:36:05.339 --> 00:36:10.020 it was pretty much impossible to practice anything outside of just doing runs over and 00:36:10.020 --> 00:36:11.150 over again. 00:36:11.150 --> 00:36:15.900 Things had come quite a long way, and Aaron’s implementation of this ridiculous strategy 00:36:15.900 --> 00:36:20.930 in 29 to save a fraction of a second was pretty emblematic of that. 00:36:20.930 --> 00:36:25.829 For over a year now, hsblue had slowly been climbing up the leaderboard, inching closer 00:36:25.829 --> 00:36:30.079 and closer to the world record, something that he very much desired. 00:36:30.079 --> 00:36:35.490 Just when he felt like he could finally chase it down, Aaron leaped into the 1st place spot. 00:36:35.490 --> 00:36:38.390 However, he wasn’t going to let that deter him. 00:36:38.390 --> 00:36:45.040 He came back and started setting massive Personal Bests, first a 5:45, then a 5:42, then a 5:29, 00:36:45.040 --> 00:36:53.160 then a 5:27, and then a 5:26, tying Aaron’s record to the second, but actually clocking 00:36:53.160 --> 00:36:55.599 in just 4 frames slower. 00:36:55.599 --> 00:37:01.119 Still, this run did have some impressive new tech, namely a double coin grab at the end 00:37:01.119 --> 00:37:06.410 of level 6, and a much faster cycle in level 8 that required a frame perfect wall clip 00:37:06.410 --> 00:37:08.319 to avoid hitting the enemy. 00:37:08.319 --> 00:37:12.560 Finally, wall clipping had a substantial use in full game runs! 00:37:12.560 --> 00:37:18.079 Hsblue omitted the third weave in level 10 and took a safer approach in level 16 that 00:37:18.079 --> 00:37:23.900 didn’t lose any time, but he was still very much on pace for a sizable new world record. 00:37:23.900 --> 00:37:29.609 Unfortunately, disaster struck in level 18, and he had to revert to Pzula’s old method 00:37:29.609 --> 00:37:33.560 of completing the level, which lost him about three seconds. 00:37:33.560 --> 00:37:39.109 Regardless, he pressed on; no matter how long it took him, he was going to keep playing 00:37:39.109 --> 00:37:42.030 until he got the record outright. 00:37:42.030 --> 00:37:47.230 Unfortunately for him, that very same day, Aaron moved the goal posts once again. 00:37:47.230 --> 00:37:52.450 Aaron uploaded a video that began with him resetting at the start of level 1. 00:37:52.450 --> 00:37:58.440 Yes, he was actually going for the double frame-perfect IL wall boost to save a single 00:37:58.440 --> 00:38:02.400 frame, something which he thankfully hit on the next attempt. 00:38:02.400 --> 00:38:06.250 He didn’t go for the wall clip in level 8 because he couldn’t find the right window 00:38:06.250 --> 00:38:12.240 size to get it to work, oh, I guess I forgot to mention that every wall clip in the game 00:38:12.240 --> 00:38:17.690 is dependent on the game’s window size, thanks flash, but he did hit a corner cut 00:38:17.690 --> 00:38:21.950 in level 10 to save a cycle, also hitting all three of the weaves. 00:38:21.950 --> 00:38:28.440 In level 12 he got past the final enemy obstacle a cycle early, but in level 30, he sadly missed 00:38:28.440 --> 00:38:33.700 collecting this coin on his first go, and he got to the end too late to be able to slide 00:38:33.700 --> 00:38:39.640 under the last enemy; it was as if the ghost of Pzula’s 6 flat had come to haunt him. 00:38:39.640 --> 00:38:44.710 Aaron lost around a second, ending with a 5:24 instead of a 5:23. 00:38:44.710 --> 00:38:50.160 However, since he still improved the record and wasn’t on the verge of a milestone, 00:38:50.160 --> 00:38:53.650 this mistake probably didn’t matter that much. 00:38:53.650 --> 00:38:56.650 Hsblue’s attempt counter continued to rack up. 00:38:56.650 --> 00:39:00.520 There was no time better than the present for him to go for the world record. 00:39:00.520 --> 00:39:05.190 Aaron would likely continue to push his times further and further, making hsblue’s goal 00:39:05.190 --> 00:39:08.359 more and more difficult to attain with each improvement. 00:39:08.359 --> 00:39:12.750 For 4 years, he had been playing The World’s Hardest Game, hoping that one day he could 00:39:12.750 --> 00:39:14.829 have the record to himself. 00:39:14.829 --> 00:39:20.950 All he needed to do was match his 5:26 without messing up level 18, and the title would be 00:39:20.950 --> 00:39:21.950 his. 00:39:21.950 --> 00:39:26.600 On May 31st, 2019, he got onto a run that started as planned. 00:39:26.600 --> 00:39:30.710 He hit the wall clip in 8 but didn’t go for the clip or weave in level 10. 00:39:30.710 --> 00:39:36.140 He didn’t do the fast ending for level 12 because there was simply no need, he was still 00:39:36.140 --> 00:39:38.060 easily on pace without it. 00:39:38.060 --> 00:39:43.470 He messed up the last coin grab in level 16, but it didn’t end up costing him very much. 00:39:43.470 --> 00:39:48.980 He approached level 18, the same level that had essentially killed his world record chances 00:39:48.980 --> 00:40:03.330 during his 5:26, but this time… disaster struck again. 00:40:03.330 --> 00:40:08.810 He didn’t lose a full three seconds this time, but his world record prospects had still 00:40:08.810 --> 00:40:11.200 been flushed down the drain. 00:40:11.200 --> 00:40:26.760 However, there was still one hail mary strategy that could potentially save it. 00:40:26.760 --> 00:40:48.920 TAS 23. 00:40:48.920 --> 00:41:00.000 5:23. 00:41:00.000 --> 00:41:11.870 Miraculously, he had done it. 00:41:11.870 --> 00:41:16.120 It’s hard to overstate how much this run meant to hsblue. 00:41:16.120 --> 00:41:20.660 He had put so much time and energy towards this grind, he had dreamed about having this 00:41:20.660 --> 00:41:24.140 record for years, and now, he had it. 00:41:24.140 --> 00:41:28.460 The first sentence of the run’s description puts it best: “If you put your heart and 00:41:28.460 --> 00:41:34.310 mind to it, you can do anything” Funnily enough, without Aaron’s mistake in level 00:41:34.310 --> 00:41:40.599 30 of his 5:24, essentially the last mistake that one can possibly make in a World’s 00:41:40.599 --> 00:41:46.130 Hardest Game speedrun, hsblue wouldn’t have become the undisputed, full second champion 00:41:46.130 --> 00:41:47.880 of the game with this run. 00:41:47.880 --> 00:41:52.290 Whether he would have later claimed the record outright or not, I guess we’ll never know, 00:41:52.290 --> 00:41:57.740 but it is funny to look back and see how perfectly the stars aligned for this moment. 00:41:57.740 --> 00:42:03.330 Indeed, Hsblue seemed to have achieved his end goal with the game, but Aaron certainly 00:42:03.330 --> 00:42:04.579 had not. 00:42:04.579 --> 00:42:08.541 Aaron knew that there were still plenty of things that he could do to push the record 00:42:08.541 --> 00:42:12.109 to a new 10-second barrier, 5:1x. 00:42:12.109 --> 00:42:18.390 His 5:24 didn’t have the quick exit in level 7, wall clip in 8, TAS 23, and some other 00:42:18.390 --> 00:42:22.569 individual level strategies that could potentially be implemented. 00:42:22.569 --> 00:42:28.280 And of course, there was that pesky mistake in level 30 that had cost him an entire second. 00:42:28.280 --> 00:42:33.200 The whole summer passed without anything new, but finally, he managed to piece together 00:42:33.200 --> 00:42:35.360 something pretty incredible. 00:42:35.360 --> 00:42:40.221 Levels 1-7 started out perfect, but he still couldn’t go for the wall clip in 8 because 00:42:40.221 --> 00:42:43.510 he still couldn’t find the right window size to make it work. 00:42:43.510 --> 00:42:45.589 So… what was going on? 00:42:45.589 --> 00:42:49.950 At this point, pretty much every single World’s Hardest Game runner was using the official 00:42:49.950 --> 00:42:54.720 standalone adobe flash player for their runs, as it was way more convenient, allowed for 00:42:54.720 --> 00:42:59.890 better performance, and allowed the game to be played at any window size, including fullscreen. 00:42:59.890 --> 00:43:03.790 However, Aarons’ platform of choice was… 00:43:03.790 --> 00:43:04.960 Internet Explorer. 00:43:04.960 --> 00:43:10.849 For whatever reason, a web browser that had been discontinued and replaced by Microsoft 00:43:10.849 --> 00:43:16.720 because of how bad it was, actually ran the game better on his dusty old computer. 00:43:16.720 --> 00:43:21.210 Aaron’s interesting platform of choice made it far more difficult for him to find a window 00:43:21.210 --> 00:43:27.040 size where the 8 wall clip and the other strategies he was going for all worked, so he just decided 00:43:27.040 --> 00:43:28.640 not to bother with it. 00:43:28.640 --> 00:43:33.690 Despite this unfortunate handicap, Aaron knew that he could still shatter the game’s next 00:43:33.690 --> 00:43:35.460 10 second barrier. 00:43:35.460 --> 00:43:39.280 He pressed on, nailing the wall clip and weaves in level 10. 00:43:39.280 --> 00:43:44.369 He essentially stayed perfect going into level 23, where he went for the TAS strat first 00:43:44.369 --> 00:43:52.100 implemented by hsblue. 00:43:52.100 --> 00:43:53.150 He got it. 00:43:53.150 --> 00:43:58.900 Granted that he didn’t mess up level 30, the run was now on pace for a 5:20 flat, just 00:43:58.900 --> 00:44:00.790 a bit shy of his goal. 00:44:00.790 --> 00:44:05.290 However, he had a plan that could just barely nudge it over the edge. 00:44:05.290 --> 00:44:13.100 He entered level 26, and proceeded to do this on the right side. 00:44:13.100 --> 00:44:14.829 Another death-defying weave. 00:44:14.829 --> 00:44:18.599 He was now on pace to break into a new 10 second barrier. 00:44:18.599 --> 00:44:40.710 All he had to do was clutch the ending. 00:44:40.710 --> 00:44:42.880 There it was. 00:44:42.880 --> 00:44:44.839 5:19.633. 00:44:44.839 --> 00:44:50.740 Plenty of geniuses in the YouTube comments completely ignored the pinned message explaining 00:44:50.740 --> 00:44:56.359 the game’s collision mechanics… to call it out for being fake, but it was absolutely 00:44:56.359 --> 00:44:57.359 real. 00:44:57.359 --> 00:45:01.920 Aaron took a well-deserved break after this run, as it was unlikely that anyone would 00:45:01.920 --> 00:45:06.839 even get close to challenging his throne, but he still knew in the back of his mind 00:45:06.839 --> 00:45:09.000 that more could be done. 00:45:09.000 --> 00:45:14.670 Going into 2020, it became apparent that there were more TAS or IL-only strategies that could 00:45:14.670 --> 00:45:17.540 potentially be brought over to full game runs. 00:45:17.540 --> 00:45:22.170 First, there was a wall clip in level 6 to collect the last coin without having to revolve 00:45:22.170 --> 00:45:24.839 with the spinner, saving over a second. 00:45:24.839 --> 00:45:29.390 Next, there was a frame-perfect turnaround here in level 7 that allows the player to 00:45:29.390 --> 00:45:32.600 phase through this first enemy, saving around half a second. 00:45:32.600 --> 00:45:37.700 Additionally, there was a cycle skip in level 12 that required a frame perfect elevator 00:45:37.700 --> 00:45:43.500 entry, perfect character alignment to squeeze under these enemies, and absolutely no slowdown 00:45:43.500 --> 00:45:49.150 for the rest of the level, allowing for a ridiculous two and a half seconds to be saved. 00:45:49.150 --> 00:45:53.720 Finally there was a frame perfect entry in level 26 where the player can go under the 00:45:53.720 --> 00:45:58.829 dots at the top right instead of going over them, ultimately saving around a second. 00:45:58.829 --> 00:46:05.329 Of these 4, Aaron recognized that the strategies in 7 and 26 were the most realistic to implement. 00:46:05.329 --> 00:46:10.500 And, at last, he had found a way to hit the wall clip in level 8, which meant that he 00:46:10.500 --> 00:46:13.380 could save an additional two and a half seconds. 00:46:13.380 --> 00:46:18.880 If everything went according to plan, a 5:14 was not out of the question. 00:46:18.880 --> 00:46:35.500 Now, he just had to piece everything together. 00:47:29.789 --> 00:47:32.237 What. A. Run. 00:47:32.253 --> 00:47:38.190 Aside from a somewhat suboptimal level 26, Every single thing went exactly according 00:47:38.190 --> 00:47:43.190 to plan, and the expected potential for these strategies was matched exactly. 00:47:43.190 --> 00:47:48.990 This run is, absolutely, one of the greatest flash game speedruns ever performed. 00:47:48.990 --> 00:47:54.430 The time and dedication that Aaron has put towards perfecting his craft is truly inspiring, 00:47:54.430 --> 00:47:58.960 and there is very little chance of this record being beaten any time soon. 00:47:58.960 --> 00:48:06.850 He stands in first by over 9 seconds, as hsblue still sits in a distant 2nd with his 5:23. 00:48:06.850 --> 00:48:13.020 As you may have guessed, this run, a true cornerstone of flash game speedrunning, was 00:48:13.020 --> 00:48:15.710 performed on Internet Explorer. 00:48:15.710 --> 00:48:18.010 You really can’t make this stuff up. 00:48:18.010 --> 00:48:23.829 Aaron may come back again, but as it stands, he is the undisputed champion of the World’s 00:48:23.829 --> 00:48:25.180 Hardest Game. 00:48:25.180 --> 00:48:51.500 However, that doesn’t mean it’s bound to stay that way forever. 00:48:51.500 --> 00:48:57.040 In December of 2007, a high school student released a game that completely changed the 00:48:57.040 --> 00:48:58.040 internet. 00:48:58.040 --> 00:49:03.080 The game challenged players to not just overcome it, but to truly conquer it by achieving a 00:49:03.080 --> 00:49:06.530 completion with the fewest number of deaths possible. 00:49:06.530 --> 00:49:11.470 After a player finally managed to achieve a perfect score of 0 deaths, many shifted 00:49:11.470 --> 00:49:16.340 their attention towards a new challenge, seeing how fast they could complete the game. 00:49:16.340 --> 00:49:21.750 When Aaron set his current world record of 5:14, many believed that the human limit for 00:49:21.750 --> 00:49:27.490 the game was around a 5:12, as the TAS 12 cycle was deemed to be too difficult to implement 00:49:27.490 --> 00:49:28.620 into runs. 00:49:28.620 --> 00:49:34.000 However, with Aaron recently building much greater consistency with the strategy in practice, 00:49:34.000 --> 00:49:38.680 many now believe that a sub 5:10 could potentially be attained one day. 00:49:38.680 --> 00:49:45.750 The most recent full-game TAS is a 4:57.467 authored by Steev0, which pulls off ridiculous 00:49:45.750 --> 00:49:50.109 wall clip and hitbox abuse strategies in almost every single level. 00:49:50.109 --> 00:49:55.120 Published back in April of 2019, this is the video that truly brought the game into the 00:49:55.120 --> 00:50:00.910 era of frame-perfect optimization, making Steev0 responsible for almost all of the strategies 00:50:00.910 --> 00:50:04.619 that are used in top-level World’s Hardest Game speedruns today. 00:50:04.619 --> 00:50:09.720 Expectedly, this TAS is a bit outdated, with there being many levels where small improvements 00:50:09.720 --> 00:50:13.802 could be made, but it’s still funny to see that humans have been able to at least match 00:50:13.802 --> 00:50:20.670 this TAS in 25/30 of the levels, actually pulling ahead by a frame in four of those. 00:50:20.670 --> 00:50:26.359 The human SOB currently sits at a 4:59.600, the majority of the difference coming from 00:50:26.359 --> 00:50:32.329 level 20, where the TAS saves a hefty 55 frames by hitting a cycle that is deemed humanly 00:50:32.329 --> 00:50:33.349 unviable. 00:50:33.349 --> 00:50:38.570 There are simply way too many frame perfect inputs required, nearly triple digits, for 00:50:38.570 --> 00:50:44.490 a sub 5 to ever have a chance of being pulled off in a full game run, but man, it is pretty 00:50:44.490 --> 00:50:48.490 amazing to think about how far things could theoretically go. 00:50:48.490 --> 00:50:53.020 But looking back, it’s crazy to see how far the World’s Hardest Game has come in 00:50:53.020 --> 00:50:54.710 the past 15 years. 00:50:54.710 --> 00:51:00.490 Nowadays, it is the third most played game of all time on both addicting games and armor 00:51:00.490 --> 00:51:06.440 games, having a combined 95 million plays on those two websites alone. 00:51:06.440 --> 00:51:11.040 After releasing the second game in the series, Snubby spent three years making a game called 00:51:11.040 --> 00:51:16.770 Straight Bullyism, a highly ambitious RPG with tons of animated cutscenes that is hard 00:51:16.770 --> 00:51:20.230 to believe was made entirely by a single developer. 00:51:20.230 --> 00:51:25.930 Eventually, he came back to The World’s Hardest Game series, releasing TWHG 3 in January 00:51:25.930 --> 00:51:30.000 2014 and TWHG 4 in June 2016. 00:51:30.000 --> 00:51:34.849 With the death of flash on the web looming in the late 2010s, he was sponsored by Coolmath 00:51:34.849 --> 00:51:39.960 Games to port the first installment to HTML5 so that it could continue to be enjoyed by 00:51:39.960 --> 00:51:44.849 future generations, with almost all of the levels being reworked in some way. 00:51:44.849 --> 00:51:49.810 The remaining 3 installments were ported by coolmath themselves to a much lesser degree 00:51:49.810 --> 00:51:55.410 of quality; 2 is impossible to beat thanks to there being game-breaking glitches in multiple 00:51:55.410 --> 00:52:00.819 levels and 4 was never fully finished and got scrapped before it was ever released. 00:52:00.819 --> 00:52:05.950 The original versions of the first two installments run great on the flash emulator Ruffle, so 00:52:05.950 --> 00:52:09.490 they have continued to have life on other websites as well. 00:52:09.490 --> 00:52:14.160 As Snubby moved further into adulthood, he eventually started going by Stephen instead 00:52:14.160 --> 00:52:19.349 of Stevie, which is why his name is different on all of the newer installments and ports. 00:52:19.349 --> 00:52:24.430 His web game development days seem to be behind him, with him pulling the plug on Snubby Land 00:52:24.430 --> 00:52:29.600 and all of his other websites, but the legacy that he has left behind on the internet is 00:52:29.600 --> 00:52:31.460 nothing short of incredible. 00:52:31.460 --> 00:52:37.020 He created one of the most popular and memorable series of games ever released on the web, 00:52:37.020 --> 00:52:41.750 with the first game in particular truly being a lightning-in-a-bottle creation. 00:52:41.750 --> 00:52:46.579 Its formula inspired people to keep playing the game over and over again until they could 00:52:46.579 --> 00:52:51.810 truly master it, and their efforts to push the game closer and closer to its limit have 00:52:51.810 --> 00:52:55.329 been viewed in awe by tens of millions of people. 00:52:55.329 --> 00:52:59.819 Each record holder shaped the game in their own unique way: HatredCopt3r helped lay the 00:52:59.819 --> 00:53:05.770 foundation for low-death completions, Hachimen was the first to verifiably achieve the game’s 00:53:05.770 --> 00:53:11.619 definition of perfection, Shiroma unlocked the game’s boundless potential, Pzula revolutionized 00:53:11.619 --> 00:53:17.079 the game with his practice mod and innovative discoveries, hsblue broke the game wide open 00:53:17.079 --> 00:53:22.590 with wall clipping, while also showcasing the power of never giving up, and Aaron pushed 00:53:22.590 --> 00:53:28.020 the game near its absolute limit through his massive, shocking records that always seemed 00:53:28.020 --> 00:53:29.480 to come out of nowhere. 00:53:29.480 --> 00:53:36.490 Over 15 years, this game went from being beaten, to being “perfected”, to being completely 00:53:36.490 --> 00:53:39.100 dismantled and dominated. 00:53:39.100 --> 00:53:43.650 Speedrunners are the ones who realized that to truly conquer The World’s Hardest Game, 00:53:43.650 --> 00:53:48.809 it was necessary to completely redefine what “perfection” meant. 00:53:48.809 --> 00:53:50.780 Thanks for watching. 00:53:50.780 --> 00:53:57.790 Enormous thank you to Aaron, Pzula, Danicker, Fraims, hsblue, iBall, and Shiroma for making 00:53:57.790 --> 00:53:59.299 this project possible. 00:53:59.299 --> 00:54:04.069 I’m sorry that this one took quite a while to put together, but I hope that it was worth 00:54:04.069 --> 00:54:05.069 the wait. 00:54:05.069 --> 00:54:10.609 It’s hard to believe that we are this close to a hundred thousand subscribers, and I can’t 00:54:10.609 --> 00:54:12.570 thank you all enough for the support. 00:54:12.570 --> 00:54:16.940 If you enjoyed watching this video and want to be a part of reaching that milestone, then 00:54:16.940 --> 00:54:18.599 consider subscribing. 00:54:18.599 --> 00:54:24.030 Also, definitely go watch all of Aaron’s 5:14, linked in the description, and thanks 00:54:24.030 --> 00:54:28.310 to all my supporters on Patreon, which you can join for just $1 per video. 00:54:28.310 --> 00:54:33.900 I hope you all have a great rest of your day, and I’ll see you all… again in less than 00:54:33.900 --> 00:54:40.460 three months, I promise.