Recapping the 2022 Opus Magnum Tournament

Recapping the 2022 Opus Magnum Tournament

There are a bunch of posts I have been meaning to make. This post will serve both as a springboard for some of those posts, as well as an excuse for not having made them earlier! Instead of writing, I have spent much of the last 3 months competing in the 2022 Opus Magnum Tournament. Let’s take a look at how it all went down.

Tournament Format

The 2022 Opus Magnum tournament officially began January 8th, with a schedule lasting just under 3 months. The face of the tournament was twitch/discord/reddit user “OMGITSABIST,” or Bist for short. A self-described cat-girl from Taiwan, Bist competed in the 2021 tournament, finishing in 12th. She was inspired by the experience to build several puzzles and host this year’s tournament.

There were 9 puzzles in total, and usually two weeks to work on each (with some overlapping time when you could work on either of two puzzles). All puzzles arrived via the shiny new website http://events.critelli.technology, built by “panic,” who you may recognize from the computer post. For each puzzle, players submit two solutions. At the puzzle’s deadline, Bist started up a stream on twitch, showed all solutions, and everyone earned points. The better the solution, the more points, with a maximum of 10.

Helpers

A few more important people made this all happen. Shadowcluster performed well in the very first tournament and is widely regarded as the top area optimizer. Instead of competing, he preferred to help playtest and advise Bist. RP0, host of the first tournament back in 2019, took a similar role. And Grimmy, who only competed in a couple of puzzles in tournament history but is a top player across several Zachtronics games, advised as well. Between Bist, panic, Shadowcluster, RP0 and Grimmy, the tournament was in wonderfully capable hands.

Players

Dozens of people sent in submissions over the course of the tournament. 72 people submitted for at least one puzzle. 29 of them submitted for every single puzzle. Many of these people were participating in the tournament for their first time. Before I get into the medals race, I’ll highlight several players who joined this year for the first time, and submitted to all or nearly all of the puzzles.

Glass Lady

Glass Lady is a perfect example of someone in it for the love of the game. She submitted to every puzzle except the (far more difficult) final one, and made solutions that she liked. Her sum-focused solution for week 2 “Soothing Salve” is shown below.

A solution by Glass Lady

Sum is a nice metric for players who don’t enjoy the process of chasing any individual metric to its extreme. Instead of going for low cost, or low cycles, or low area, this solution (as well as the other 43 that were sent in for this puzzle), attempts to balance between all 3. This leads to a very aesthetically pleasing solution. Glass Lady took 41st overall in the tournament.

bsmith

New to tournaments as of 2022, bsmith stuck out to the end and submitted to every puzzle. Their throughput-optimal solution for week 6 “Fulmination” is shown below.

A solution by bsmith

You may recognize throughput as the metric described in my first post about this game. In week 6, players were asked to solve this nasty puzzle for throughput, and bsmith was one of 33 players to do so optimally. bsmith placed 30th overall in the tournament.

Tulare

In addition to submitting for every puzzle, Tulare added the “cherries emoji” to the common lexicon of the Opus Magnum discord user. His attitude and wholesome take on personal growth and advancement are refreshing, in a community that sometimes gets too carried away in chasing perfection. The cherries emoji became his personal “I like what you have to say” reaction, and many others have adopted it since then. Below is Tulare’s sum solution for the week 7 puzzle “Potent Painkillers.”

A solution by Tulare

Tulare finished 29th in the tournament. He is now looking to start a series called “Alchademy” to help others share their improvements and concepts in Opus Magnum in a collaborative environment.

Cadspen

Certainly a contender for “most improved,” Cadspen beat the game in December, and decided to hop right into the tournament. She reached competitive level at the game very quickly! In week 1, she already had the following cycles solution for Rust Removal. Note that this puzzle has a particularly nasty method required, in which the 12 copper come from different mechanisms and everything barely completes at the end. The gif will loop poorly for that reason, but the 13 cycle score is legitimate either way.

A solution by Cadspen

Cadspen took 28th overall, with submissions to every single puzzle.

Buttermilk Doughnut

According to a discord post, Buttermilk Doughnut came into the tournament with “some hubris.” They had a great time and learned a lot, and may be writing up a “lessons learned” post somewhere. I invite them to link in the comments of this blog post if it does go up.

Now, that hubris was not entirely misplaced! They had a few excellent performances, including this compact cycle-optimal solution for the week 8 puzzle “Radio Receivers.”

A solution by Buttermilk Doughnut

Buttermilk Doughnut took 24th in the overall standings.

geco

geco, also called corliss, found the tournament through old YouTube vods. With some impressive performances of their own, they definitely made a strong debut! Their solution for Fulmination, optimized for latency, is shown below.

A solution by geco

This tournament was the first to include an explicit “Latency” challenge – where the only score that matters is time to first output. At 12 cycles to first output, this solution earned geco 5th place for the puzzle. Overall, geco took 19th place.

Coocoo52

I can’t find any record of Coocoo52 submitting to an official tournament puzzle before 2022, although they have been a member of the greater community for quite some time. They are most well known for their extreme challenge solutions within the game, including “every arm shares the same track loop” and “only allowed instructions are grab, retract, drop”. Seriously, check these out.

As for the tournament, we got to see Coocoo52’s far more “ordinary” solutions to competition puzzles. Of course, I say that, and their highest scoring finish of the tournament is on the wonky 12-copper infinite. It’s Rust Removal cycles again, but now with extra programming to make it “loop” afterward. Good sense of humor, since everything after cycle 11 is not necessary here.

A solution by Coocoo52

Overall, Coocoo52 took 18th.

fiesta0618

Another wholesome discord presence, fiesta0618 improved steadily over the course of the tournament. By week 3 they poked into the top 10 for the first time. In week 6, Fulmination, their throughput solve earned them 6th place. That solution is shown below.

A solution by fiesta0618

Notice that this and bsmith’s solution from before are both optimal throughput. This is because no fire is wasted, and the fire input comes out at top speed. However, they reach the perfect throughput of 3 outputs per 16 cycles (average) in very different ways! bsmith employs 1/8 + 1/16 = 3/16, while fiesta0618 employs 1/6 + 1/48 = 3/16.

In breaking ties on throughput, the cheaper solution would win. So fiesta’s solution at 945g placed very well. This approach of 1/6 + 1/48 tended to produce far cheaper solutions.

In the end, fiesta0618 took 16th place in their first tournament.

notgreat

A joke made far too many times throughout the results streams, was that “notgreat” seemed to be the opposite of their name. Coming in 13th overall, several of notgreat’s solutions landed in the top 10. I’ll be making a separate post later about this year’s computation puzzle, which is one of their strongest results. In short, they discovered an algorithm that only 2 other people found, which led to exceptionally fast cycle counts. Additionally, they built an alternate interface for the game that allows rewinding and programming during execution. While this wasn’t production ready during this tournament until near the end, I expect it to help immensely on future area and cost solutions.

To keep the trend of including a solution for every debutant, here is notgreat’s 11 cycle solution to Rust Removal.

A solution by notgreat

ebonnov

Arriving during last year’s post-tournament weekly challenges, ebonnov showed promise as a medal contender. 2022 was their first official tournament, and they took 8th place overall. This included one overall win on Soothing Salve sum. That beautifully clean solve is shown below:

ebonnov’s winning sum solution to Soothing Salve

Velvacaine

Finally, we have our highest placing new entrant. Velvacaine was the only newcomer to average over 8 points out of 10 across every submission. No individual puzzle gave them a win, they instead proved to be very well rounded. Rather than embed their strongest finish, which was minimum area for Potent Painkillers with a cycle count over 2000, I’ll share their latency solve for Fulmination.

A solution by Velvacaine

This was the cheapest of all solutions to drop the first output on cycle 12. The reason it wasn’t a winner, was because 3 people managed to find a path to 11. That said, the timing around getting the 6-pronged arm to serve all 3 quicksilver, sets it apart from the other 12s.

Velvacaine took home 6th place in the tournament.

Fatigue

This is a competitive event lasting 3 months. At the same time, it is a video game. There is no prize beyond bragging rights. Everyone playing is a human being with a life outside of the game. As much as I lauded the new players who stuck it out, there is also something to be said about the players who got their solutions in when it was exciting for them, and dropped out once it was not. I want to give a particular mention to Zorflax, who debuted this tournament but set a very high self standard. By the halfway point of the tournament, he seemed to be forcing himself to play rather than enjoying it, and vowed to take a break and check out the puzzles later with less pressure. Good on you for figuring out what you needed to do.

Whether born of fatigue or other commitments, Rolamni, Specht98, and jinyou also opted to not hold themselves to their highest standard. All three are serious medal contenders, so without them the landscape changes a bit. Of these 3, Rolamni did continue to submit every week with far less time committed. The others, along with many other names, submitted to only certain puzzles.

Who’s left?

Of last year’s top 10, myself and PentaPig were clear favorites. We had each won before, me in 2019 and 2021, and PentaPig in 2020. Certainly though, you would not count out mr_puzzel (5th place 2021), Goodbye Galaxy (6th place 2021), Noeuchar (7th place 2021), F43nd1r (3rd place 2020), or any other player who might have been massively improving in the last year

The Tournament Puzzles

I highly recommend clicking through the 9 official puzzles on the tournament website. Weeks -1 and 0 were there for infrastructure testing and hype, official weeks begin at 1. They all have some amount of flavor text, and a description of what you are optimizing for. You can get a significant sense of the puzzle without ever even opening the game!

In hindsight, a few things stand out about this tournament.

Harder area puzzles

With Shadowcluster advising, the area puzzles for 2022 were calibrated to be interesting to the highest skilled players. As a result, they were hard! First was Soothing Salve.

Soothing Salve

Soothing Salve – parts and presentation

The center of the output is Vitae, which can only be formed at the same time as Mors. Mors doesn’t appear in the output, so you might expect to use the glyph of disposal to get rid of it. But no! The best strategy for area, was to build 6 outputs while keeping the Mors around, since it takes up less space than the glyph of disposal.

As an added challenge, the tiebreaker for this puzzle was instructions. Keeping 6 Mors around is hard enough, but doing so using an instruction tape that only builds one output? That means you have to have conditional handling for Mors that might be there or might not.

This was my lowest scoring week. I traditionally struggle with area optimization anyway. While I felt proud of what I did accomplish, I learned just how much better was possible. Shown below is the winning solve by Specht98.

An early loop of Specht98’s winning Soothing Salve solution. The metric was area, ties broken by instructions.

If it’s difficult to follow the grabs-at-nothing, I don’t blame you. The gif won’t show it properly because Mors disappears on loop, but this instruction tape successfully builds 6 outputs one at a time. All the grabs-at-nothing handle possible places that it might have stored Mors. For completeness, here is the same solution but making its 6th output.

A different loop of Specht98’s winning Soothing Salve solution. Now with more Mors on the board to show what the handling is about

Film Crystal

Film Crystal – parts and presentation

Much more of a traditional area challenge, we had to build a big output with a hole in the middle.

Theoretically, you could pick which inputs you wanted to use. If you placed the salt-air, then you got silver by purifying metal, and there was no reason to use iron when copper is strictly better. If you placed the quicksilver-air, then to have a clean balance between atoms in the output, you wanted to place iron only. But which route was better for area?

I was one of several people to build a 33 area solution using purification. 32 never seemed within reach, if it was even possible. Others, including Rolamni, mentioned spending hours upon hours looking for a working 32 only to have it fail in some convoluted obscure way. It may have been this endeavor that led Rolamni to eventually promise less of his time.

My layout at 33 area. Many people found 33, but there were hints that it might not be minimal.

This was my layout, taking 33 hexes in total. Programming is left as an exercise to the reader (it is 1811 cycles). One worthwhile insight is that you can pivot on the salt in the input, and then put a copper in place of the air. Then the glyph won’t spawn a new salt-air, and the tile that once held salt can be used for moving the partial product to make its bonds.

However, none of us 33-makers were prepared for what Noeuchar had in store.

Noeuchar’s winning layout at 32. Solution takes nearly 5000 cycles and can’t be embedded.

Yes, that is 32. And the iron input is in the middle of the output. This takes almost 5000 cycles, so I can’t reasonably take a gif of it, but I encourage you to watch the results video here, to see how the heck this design manages to make outputs. It can’t put metal directly on projection because no arm is able to reach that tile, so it bonds things together to make silver, then unbonds them later. The dance with iron to keep it from crashing into the product is sublime. Definitely a highlight of the tournament.

An aside..

It’s worth noting that both Specht98’s winning solution for Soothing Salve and Noeuchar’s winning solution for Film Crystal were upstaged by Shadowcluster’s playtest solutions, but in both cases only barely. Shadowcluster was able to find a 32 which didn’t put the iron in the center, allowing it to use only one arm but still 32 hexes of space in total. These made up 2 of the 3 cases where Shadowcluster managed to hold onto a playtest solution better than anyone submitted.

Heavier focus on instructions

Another feature of this tournament was the repeated use of instructions as a metric. I will go into this more in another post. The restricted space puzzles have typically used instructions as a metric in place of area, and each tournament has included one such puzzle.

Aether Reactor

Aether Reactor – parts and cabinet display

These “cabinet” puzzles are always a polarizing feature of the tournaments. Some people hate them, others thrive with the challenge. Part of the challenge comes from instructions. The difference between middle-of-the-pack and winner in many puzzles is rarely more than a few cycles, gold, or area out of dozens. But instructions tend to separate the field a lot more. In 2021, PentaPig won the instructions puzzle with 55 instructions, where 10th place had 73. The ratio there impacts how many points are awarded, so players need to be competitive in instructions to stay at the top.

For Aether Reactor, PentaPig once again won the instructions metric. Click through his imgur album to see how he got to the 32 pictured.

PentaPig’s winning instructions solution for Aether Reactor.

I was a competitive 3rd at 37, and 10th was 44. Once again there were big gains to be had for PentaPig’s tournament aspirations.

More instructions puzzles?

But in 2022, Aether Reactor was not the only instructions puzzle! I addition to its use as a tiebreaker in Soothing Salve area, we had to solve Film Crystal and Radio Receivers for instructions. Sort of.

Because of prior work by Rolamni, we know instructions in unlimited space have a forced solution style. You put down a really really long track, and you make it so that an arm with at most 4 instructions on that track, eventually solves the puzzle. Every puzzle in the campaign can be solved in 4 or 3 instructions by this method.

The approach I took in 2020 was to pose “instructions + cost/5g” as the free space instructions metric, so that each segment of track penalized you the same as one instruction. The approach Bist took in 2022, was to directly pose “tracks + instructions” as the free space instructions metric. And she used it not once, but twice.

I completely dominated tracks + instructions. Untied primary victories on both puzzles, which left me regaining several points vs PentaPig that I lost in area and other metrics. You can watch the tournament recap videos if this teaser leaves you wanting instant gratification, but hold out for another post later where I talk about those victories.

As a result of the repeated use of instructions for scoring, PentaPig and I were far ahead of the rest of the field. The two of us seemed like certain 1st and 2nd, with an outside chance Rolamni could put in effort and rise to compete. However, I would like to highlight a few nice performances from others first.

New first-time winners

As mentioned above, ebonnov got a win in their first ever tournament. Here’s a look at other players who played in previous tournaments, but won for the first time in 2022.

Specht98

I called out Specht98’s win on Soothing Salve as part of week 2. Specht98 had competed in the 2021 tournament but never got a win. That said, Soothing Salve was not their first win! They arrived in week 1 with a winning solve as well, for the cost metric. If you thought that cycles solves to Rust Removal were a little janky, the problem is even more apparent at min cost.

The winning cost solution for Rust Removal, by Specht98. Framecount reduced for embed, sorry for the less smooth motion!

Specht98 had a 50% win rate by the end of week 2. They also won the throughput metric for Fulmination, using only 845g. First time winner and 3 wins during a tournament, a very powerful showing. Skipping weeks put Specht98 out of the medal contention, however, leading to a 20th place finish overall.

While Rust Removal was a well calculated puzzle, and simple enough to be a viable week 1, I think the choice to have a nonlooping infinite be the first puzzle of the tournament might drive some players away. The game likes looping. I’ll take this opportunity to once again plug Tulare’s Alchademy, which is being organized on the discord. I believe it will focus more on puzzles that feel good to solve, rather than ones that reward tricks like this.

Haxton

In 2020 and 2021, Haxton only submitted to the computation puzzle. This year, they submitted to every puzzle, and one of them was a victory! The puzzle in question was the very symmetrical 3rd puzzle, Lubricating Solvents.

Lubricating Solvents – parts and presentation

The metrics for Lubricating Solvents were cost, and the bizarrely-named “tarcles.” Tarcles is a portmanteau reflecting the calculation “(tracks + arms) * cycles.” By far the most experimental metric in this tournament, it turned out extremely well for Haxton. There was a tie for 2nd between several players, at 4 arms 0 track 28 cycles. This scored 112 by “tarcles”. But Haxton had better, the only person to find the following solution:

Haxton’s winning “tarcles” solution for Lubricating Solvents. 52 cycles with 2 arms and no track, made a tarcles score of 104.

At 2 arms, no track, and 52 cycles, this was the winning tarcles solution. The looping behavior is a little nonstandard, since there are extra inputs on the board when the 6th output drops, but that happens after the solution completes.

One of the most powerful insights for this puzzle was that two of each input, bonded together in the proper way, had all the atoms in the right place for outputs. This was referred to as “METHOD” during the stream due to how prevalent it was as a solution approach. Even with METHOD though, Haxton stood alone in how they implemented it so quickly in only 2 arms.

Finally submitting to a full tournament, Haxton took home 11th place overall.

morraconda

morraconda had participated in previous tournaments under the name monsterracer. In 2021, she got 19th place, with consistent participation but no individual top 10s. This year, for the cabinet puzzle Aether Reactor, she was very confident in her cycles solution. Cabinet cycles has a distinct challenge when compared to sprawling freespace cycles. The chambers fundamentally limit how quickly your solution can operate, and morraconda thrived with these careful, crowded pipelines.

I was very confident in my own cycles solution. At one point during the results stream, two people’s cycles solves had yet to be shown on stream, morraconda and me. One of us won, and the other took 2nd. When 2nd place was revealed to be me, everyone was super excited. morraconda had gotten her first tournament puzzle win.

Morraconda’s winning cycles solution for Aether Reactor.

For what it’s worth, I felt like my solution at 68 cycles was at most 2 cycles slower than as fast as it could go. morraconda beat me outright with those exact 2 cycles. Ultimately, morraconda finished in 12th place, a massive improvement from last year. Her performance was marked by thriving in a handful of metrics but still struggling whenever the metric was instructions.

mr_puzzel

For all his consistently high placements in tournaments prior, mr_puzzel had never won a tournament puzzle. That is, until week 7 of 2022. Potent Painkillers, the vitae and mors rings shown previously, came down to another dramatic top 2 reveal. All but 2 sum solutions had been shown, with morraconda and mr_puzzel yet to be called. He was on the voice call as well, as “isaac.wass (allegedly)” so you can hear his reaction to the reveal. morraconda took 2nd, with 439 sum. mr_puzzel took 1st, with 438.

The triumph came using the following solution. The assembly of each ring is particularly beautiful, as sum records often tend to be.

mr_puzzel’s winning sum solution for Potent Painkillers.

Ultimately, mr_puzzel would earn 5th place in the 2022 tournament.

F43nd1r

F43nd1r (pronounced as “faendir”) had participated in the 2019 and 2020 tournaments, getting a 2nd place but never an outright win. In 2021 he didn’t manage to participate in the tournament. He submitted to everything in 2022 though, and in week 8 he broke through.

Radio Receivers was a cycles puzzle, but to break ties, we had to optimize for area. The winning solution would be whoever got 81 cycles the smallest. 19 of the 35 submissions were at 81 cycles, so by the halfway point of the stream we were just counting down smaller and smaller implementations of the 81 cycle algorithms. Seeing Buttermilk Doughnut’s solution far earlier in this post at 59 area, one might think this was close to the limit. But F43nd1r definitely thought otherwise. This was his winning solve, at 81 cycles and 44 area.

F43nd1r’s winning cycles solution for Radio Receivers. Ties in cycles were broken by area.

He described his progression in an imgur album that can be found here. A remarkable feat, given that the next smallest was 7 entire hexes bigger at 51. F43nd1r finished the tournament in 7th place.

brookieoz

Brookieoz is a rather unique case. He hosted the 2021 tournament for Opus Magnum, having submitted to half of the 2020 puzzles and then dropped out. His main Zachtronics focus has traditionally been Spacechem. But there are definitely some transferrable skills.

Brookieoz submitted to exactly one puzzle in the 2022 tournament, the computation puzzle. I will have yet another post in the near future, about how these newfangled computer machines work. But suffice to say, after he unleashed the subtraction puzzle on the world in 2021, he knew what a good solution would look like. And so he made it.

brookieoz’s winning solution to Transmutation CX, which will be covered in more detail in a later post

If you can tell what this is computing from the gif alone, you’re a wizard who should be signing up for the next Opus Magnum events as soon as possible. For the rest of us, you will need to wait for the eventual post where I discuss “Transmutation CX,” the final puzzle of 2022.

One funny consequence of brookieoz submitting exactly one time this entire tournament, and winning, is that he has an untouchable average score. That isn’t how placement is decided, so he finished firmly in 46th, but it’s neat to have a perfect record.

That is, not counting Shadowcluster, who yet again had a playtest solve better than all submitters. More on this in the future post as well..!

Other winners

PentaPig had 4 wins, and I had 3. Noeuchar had the win in the 32 area puzzle, which was a repeat feat after winning an area puzzle last year as well. For the sake of showing more pretty gifs of winning solutions, here is PentaPig’s Rust Removal cycles solution:

PentaPig’s winning cycles solution to Rust Removal.

And here is my Fulmination latency solve. I was one of the 3 who got first output on 11, the others being Noeuchar and ebonnov. Mine was the cheapest, and so took first place.

My winning latency solution to Fulmination.

Final Standings

10th place went to Goodbye Galaxy, who won a tight battle with morraconda and Haxton.

9th place went to Username Void, a large improvement over last year’s 25th (though that was with one missed week)

8th place went to ebonnov, in their first official tournament.

7th place went to F43nd1r, “barely” beating ebonnov. The story here is that Bist awarded ebonnov bonus points for submitting a solution with 420 cost, 420 cycles and 69 instructions for Aether Reactor. These points would be redeemed at the conclusion of the tournament, however many it took to catch but not pass the person ahead. This leads to F43nd1r and ebonnov being separated by nearly nothing, instead of the more genuine 2 point gap.

6th place went to Velvacaine, now actually only fractions of a point ahead of F43nd1r.

5th place went to mr_puzzel, still in a close race.

4th place went to Noeuchar, who in addition to the wonderful 32 area solution, got 2nd place behind brookieoz in the final puzzle.

3rd place went to Rolamni. Yeah, even his low-time-commitment effort is incredibly strong. The gap from 3rd to 4th was over 10 points.

2nd place went to me. I fought for it, but PentaPig pulled away early, and cemented his victory during week 7. This breaks my technical “undefeated” streak, since the only year I had not won previously, was because I was host.

And finally, 1st place, with an average score of 9.385 points, was PentaPig. A dominant performance from the true “Opus Magnum main” of Opus Magnum.

Look at this graph, every time I do it makes me.. no longer be undefeated in Opus Magnum tournaments!

As a visual treat, Bist made the graph here. Every color corresponds to one of the submissions scored. The tallest bars are 10 points. The “Bonus” is 10 points for everyone who completed the final puzzle, with an extra few for ebonnov for memes. There’s more detailed information in the results spreadsheet, but I do love the clarity of the graph.

And that’s it!

Thanks for reading! I will be sure to follow this with a detailed look at instructions and trackless instructions, as well as a summary of the ideas involved in the 2022 computation puzzle. Additionally, look out for another post about silly things people have used Opus Magnum computers to do.