SC2SEA.com - Starcraft 2 SEA eSports Community Site

Tal'darim Time Trial Post Mortem by Dingobloo

I hope by now you've at least read about the map, or at best played it otherwise none of the rest of this blog is going to make much sense.

The idea for the contest started with nirvAnA approaching the news writing crew about making a puzzle or some other form of competition for SC2SEA's second birthday celebration, he had in mind the kind of URL guessing game that he had done for the first contest, I mentioned the idea of doing a custom map, but the idea was shot down largely because of the effort involved, but I promised myself that If I had something workable by the end of the weekend I would at the very least finish it, worst case I could scrap it and make a picture puzzle with the trivia I had collected.


I wish I had saved more WIP versions

So I began working on a map that originally started as more a "winged" design where each section would branch off and after some challenge would award you with that wing being complete and each successive day I would give you the password for the next wing, so people could play along and be gated each day, but someone who was a contender would complete all 3 challenges and be the first to the end.

Problems quickly became clear, while this might have been a fun way to get people to come back each day, it was also really difficult to decide a clear first place winner in any way other than first past the post. I quickly re-thought the design, and as someone who enjoys watching speed runs, I thought it would be an interesting avenue for competition, especially given that it was comparatively easy to judge the times given Starcraft 2's replay feature.

I'm not entirely convinced it was the best option ultimately. Once the first early times were posted I think plenty of people realized they had no idea how to beat them and decided not to submit or to stop attempting, an unfortunate side effect of the format. Thankfully the number of contests we had running during the month meant I could spend enough time on it to make it a little bit better, here are my thoughts on each part of it.

The Trivia


Setting up the trivia

The pool of questions is only 15 questions or so deep as I wanted a good variety of answers including players, cities, countries which were simple enough to answer but without much repetition. Originally the questions were 100% random, but the problem is with such a small pool of questions I often had repeated questions on the same run, to the point where I couldn't ignore it. Randomizing with no duplicates without being able to have a sorted list is a little bit more involved, I ended up picking a random starting position, then created a random offset between 1 and 3 to get the next question, wrapping if we hit the end of the list of questions, this then meant I had to pad it out using the unit upgrade questions to ensure there were 5 x 3 max offsets = 15 quesitons. Another change I made after the feedback to nirvAnA's puzzle was to make sure you could type answers in any case, which was simple enough.

The Warp Arena

I struggled making this one truly interesting enough, on the one hand I like that your first few times through it was nice to have an easier environment to familiarize yourself with answering a question while doing something else, the players with the best times clearly understood how little a threat the enemies actually posed. For a map that was meant to be speed run I think it needs to have that dual depth to it where you can complete it sub-optimally but someone attempting to do it fast has additionally difficulties. If I had more time I think I would have revisited it, potentially offering the question after you've killed all 4 warp prisms or some similar constraint before the question was asked.

The Lightning Round


Where the High Templar storm

It's clear to see from the original screenshot why I might naturally go towards a gauntlet for this bridge-like section. I had at first used waves of Banelings but the game play ended up being a bit too close to the arena section for my liking. Additionally, the idea to use all Protoss units as the enemy was too appealing to ignore as a theme.

Swapping over to High Templar was actually a bit more of a programming challenge, not only did I need to have a valid point to storm on and to make it random, it also had to tread that fine line between fair and dangerous. If I only let it choose 2 points out of the possible 9 It was possible to avoid a few of the more dangerous situations where it would just blanket a large area, but it also became extremely unlikely to get stormed at all, part of the compromise I made throughout tweaking this was to reduce the storm damage by up to half, both for the initial hit and the periodic amount. Unfortunately several things made this more difficult than intended.

You originally had 3 marauders at the beginning of the map that were replaced with 2 marines later on in development largely because I was able to consistently win with 3 marauders even if I completely messed up the lightning round section. Another factor that meant that it was more challenging than intended was that most of the testing was done within the map editor's Test Map function, which runs in single player, with no lag, it's extremely noticeable going between the two and plenty of people had problems in this section.

The Tug Of War


An overview of the Tug of War Section

This originally started as two sections, the area just outside the base was originally planned to be a Baneling minefield that you would have to navigate to reach pickups as well as the question barrier and the base itself was originally planned as an encounter with an Ultralisk. Unfortunately without any method of healing or a way to reproduce units I think it was playing out too overly cautious. The layout reminded me a bit of a DOTA lane with the Protoss at the altar, and the ramp up to a base of your own, so I was able to re-tool it with that in mind. The allied "creep" let you get a little more aggressive with your limited units and actually fight some stuff which was a big improvement.

I added the two barriers as a way to ensure that the trivia was as central to this portion as it was to the rest, and I was luckily able to disable collisions from the protoss units on forcefields so they could get through (it was really up in the air for a while and a lot hinged on the units being able to get through). Unfortunately given a canon's usual DPS and Health the friendly creep were just pushing through to kill the canon once the questions were answered and I had to modify their stats. I didn't put too much thought into it, but by boosting the shield's armor and regeneration rate I was able to find a decent sweet spot where the player was able to intervene and provide enough dps with their marauders to push it over the edge for your team.

I also knew I wanted include something other than unit control, so handing over a neutral building and letting you build from it seemed the obvious choice. Ensuring people didn't do anything funky with them was a bit more challenging, as you notice in the final map, you're not allowed to build with your scvs, basically because cherry picking the exact units that might cause problems was a huge pain in the ass, and the lift-off capabilities of the buildings also had their problems. I was basically left with only a few mechanics to play with, the 3 energy abilities of the orbital command, building scv's and upgrade.

Integrating the upgrades through collecting gas while creeping I thought had a nice feel to it and it also solved one of the other problems I was having, which was how to make the push to kill the photon canons more timing based. By adding in the DT's, the neutral camps, and the gas I had hoped that the ideal time to push would be a combination of reaching the supply cap, an upgrade finishing, and you having enough scan to kill any dt's, but once you ran out of scv's or scans you would need to pull back, and your best plan of attack was to collect more gas and set up another upgrade/scv timing.

Unfortunately by reducing the canon's fire rate later on in an effort to make it a bit easier, combined with a few problems that I will highlight later meant that it wasn't ultimately the best way to play out that section.

The ghost was another late addition that actually came while I was working on the final cutscene and needed to test to make sure the ghost could easily cast a nuke. I loved that his abilities combined well with the timing nature of the third portion, while he had abilities that were tempting to use during the other aspects of the map. EMP doing full damage to the high armour shields of the canons I thought was neat and his ability to reveal the DT's while suboptimal was useful. He was put in instead of 2 marines later on which at least gave your set of units some survivability but I was concerned because a large amount of tuning had already happened at this point.

The Altar

I wanted the climax to be a bit of a cool-down from the previous event but ultimately I think trying to find the balance between a newer player and one who knew all the questions was too difficult and it needed something else. I had considered increasing the speed of the colossus and letting them beat on you for a bit, so if you entered the final part with too little health you had a risk of dying and while this might have solved some problems I did not have time to adequately test if it was fair. There were numerous exploit's and weirdness around the triggering of this event, the scariest of which was just scanning and throwing a mule straight into the centre. Some of these were solved by simply killing every unit outside of the top 2 levels of the altar as soon as it triggered, so if all of your initial units were outside the altar they would die and you would lose. This also had a benefit of providing you just a small pocket of vision for the final event which I liked.

The Final Cutscene


Cut scenes in the trigger editor are good fun

I originally planned to use the new cut scene editor that came with patch 1.5 for this, unfortunately despite learning how to use it a few months a go I wasn't able to get it to do what I needed (mainly to trigger events for the load-up and the nuke call-down) it seems poorly integrated with the trigger editor in general, and seems to rely on your using the animations instead of the built in abilities which is inconvenient. I ended up just doing it how it used to be done, a bunch of interpolates between camera positions and trigger waits and I was happy with the result, except when it broke the replays at the end because the camera can't recover by itself.

The Replays

First, Second and Third by Pinder, Delete #2 and Synx

The top 3 unfortunately ended up rather similar just because the critical path ended up pretty much the same once they understood the what they could get away with. Pinder's was the first submitted and ultimately the most refined, he gets what I can only call a perfect first and second event running straight past the zealots and through the lightning round without taking a hit. He hesitates only once in the third stage before taking out the first canon and running straight past the dual canons, not even stopping to kill the final pylon. Delete's is only slightly behind for no obvious reason and Synx does stop to kill the last pylon costing him a small amount of time. It also resulted in my favourite bit of chat when someone complained about the high health canons and Pinder just said "just finish the game".


100% accurate depiction of Pinder during his run

2 Canons 1 Pylon by Delete #1

Delete's entry (prior to the one that was too late to accept) I thought was excellent, I placed the pylon for the 2 final canons knowing full well the factoid that even if the canons were normal canons, attacking the pylon was the best option. It didn't occur to me that at the time I increased the stats of the canons that I also should have done a similar adjustment to the pylon, and it was a reminder that painting with a broad brush can be your undoing if you don't get the details right. Watching him play this extremely thin "bum rush the pylon" strategy also resulted in some of the randomness changes, originally I didn't think it was going to be as random, just because I thought people would need to set into at least 1 round of pickups before they could finish it so getting hit with 1 or 2 storms wasn't a big deal as long as you kept the majority of your marines but this turned out to not be the case.


The details matter

My Bad by SmirkToT

SmirkToT's entry was legitimately good unit control that allowed him to push straight through, only to be foiled by me breaking the final event while trying to fix the cutscene. I had edited out the line that accepted the answer so I could trigger it on command to make sure it was working, and didn't realize until the next morning. It wasn't the winning entry but sorry all the same.


Poor regression testing

Creep by Radiohead/nirvAnA

This was actually a strategy I assumed was going to be used and wanted to fix but didn't know how. Setting the alliance to "Pushable" only solved the problem partly as using hold position still turned you into an immovable object, I ended up removing hold position from all of the players units control cards an hour after it went live because I felt it was too obvious to anyone who has played DOTA that this would work and there was really no margin to screw it up.


Sploits

The Half Windsor by Dox

Unfortunately Dox's placement doesn't accurately reflect how good his run is relative to other people that played the map, he's one of the few ones that got submitted that played it pretty close to how it was designed but just did it in a very refined way including good unit control when clearing the camps that eventually resulted in my changing them so they wouldn't accidentally spawn on top of you. A lot of my testing was done with me kiting away from the camps then grabbing with the other upgrades with a single unit, while Dox edged his way closer so he could grab them quickly resulting in him getting spawned on more often then I did in testing. I feel like he might have been a bit higher if I'd been able to do my due diligence to stop some of the larger exploits from happening.


The only submitted replay with the neutral area in it

So thank you all for playing, and I hoped it was at least enjoyable because if nothing else it was educational, I'm sure none of you will forget the winner of the first cheese shield, nor how to spell pneumatized carapace.