Lessons learned from the 2016 Zelda II All Keys Tournament

On December 30, 2016, in General, by Neil Stevens

We had the 2016 Zelda II 100% All Keys Tournament over the last 7 and a half weeks. It was the second annual event, and both of them were won by Simpoldood. I think it was a great success, but we still have more to learn on how to do these well.

Structural Change

The big change from last year was the new tournament structure. Instead of one giant double-elimination bracket, we used a World Cup-style two-stage tournament. Groups of five players played a round-robin, and the top two players from each group moved on to a single-elimination bracket, which I was calling the Knockout Round (in the tradition of my insisting on calling the loser’s bracket the Bonus Bracket).

To make the fairest possible first round, I tried to make all the groups even. I did this by dividing all the players into seven ‘tiers’, from the seven top personal best times in tier 1, to the seven bottom personal best times in tier 5. Group A was given to 2015 champion Simpoldood, Group B to 2015 runner-up Error72, Group C to the third place Pro_JN, and the remaining four slots were given to the other four members of Tier 1, in order of PB.

To fill in the remaining four slots in each group, I randomly chose one member from each tier, to join each group. This way nobody knew who would be in each group until the announcement stream, but at the same time all the groups would be as fair as we could make them.

The group stage addition had advantages. First, everyone got to play a minimum of four games, against players of a wide range of skill levels. This was a huge success, based on the number of players who spoke of enjoying the format. Second, it meant players had four races to schedule concurrently, which removed some bottlenecks in race scheduling, allowing us to eliminate 60% of the players rapidly. This was also a huge success, in that we finished the tournament on December 30 this time, compared with February 8 last time. The first tournament took 50% longer than this one!

I think the one-and-done KO bracket also worked well. It had the virtues of being fast, easy to follow, and loading every race with importance.


Short sign-up period

We did have problems, the first being the short sign-up period. This was on me. I didn’t realize it was time to run the tournament until very near the start time we used last year. This is because last year we waited for the Legend of Zelda tournament to be far enough along, that we weren’t in the way of our NES Zelda buddies. But this year Lackattack24 didn’t have any interest in playing in another tournament, and nobody wanted to run one with out him, so they didn’t have one. Therefore we had no clear reminder to start ours until the very last minute.

Fortunately we did get an amazing 35 entries on short notice for a great tournament.

No Wild Cards

Some of the middle-of-the-road round robin matches could have been important, except that we didn’t have any sort of Wild Card entries into the KO bracket. We will definitely use Wild Card entries next time, to balance the bracket, and to add importance to yet more matches. We’ll likely send into the KO bracket those players who didn’t make the top two in their groups, who had the best average times in the tournament.

We included average times in this tournament as a tiebreak option for the round robin stage. It ended up not mattering, but it could have, and that in theory gave players reason not to forfeit. Adding a Wild Card system will only increase that importance of every second in every race in the group stage.

No Third Place Game

We also didn’t setup a third place game. We should do that next year.

Challonge is Buggy

Challonge really let us down this year I think. Its rigid, inflexible style confused players who wanted to enter race results, as the system only prompted them to enter one race. It also gave us a problem with respect to the KO bracket: I couldn’t let people start scheduling KO round races while predictions were going on. I was forced to have an artificial prediction cut off, far before the beginning of the first race.

One issue was technical: Challonge seemed to be glitchy with respect to the two-stage tournament, and wouldn’t let me edit any scores from the first round. We had one round robin match that ran into the KO round games, and I would like to have entered that score, but I couldn’t. I will solve these issues next year by using spreadsheets and Google Forms (as we already do for the race scheduling), abandoning Challonge entirely.

With Challonge out, I will probably also make Discord entry mandatory next year, so that players will have a guaranteed means of communicating with one another. I’m not a fan of Discord, but having one central location is great. I will also ask for Twitter IDs, since that’s another great, immediate means of contacting a player for scheduling. Discord also helped a lot for organizing races off of SRL, and for doing restream and commentary coordination.

Not everyone could get restreams

Restreams were sometimes an issue. Two main issues held them up. First, races were scheduled at all hours of the day and all days of the week. A race at 1pm eastern time on a weekday was not likely to get a restream because most restreamers are in North America and were at work or school. However times like that were popular for races because of the large number of Swedish runners in the race, as well as for people with non-traditional school or work schedules. These folks often couldn’t get restreams, and that was disappointing, but unavoidable.

The other reason restreams would not happen sometimes, is that the races were scheduled on the fly, with no notice. Obviously these runners would know that by doing this, they would not get a restream. So i don’t consider this a problem with the tournament. Rather, the ability to race on the fly is one of the great strengths of the round robin format!

Rules nitpick

The final issue we ran into this year was a rules issue, actually. For years we’ve used the Speed Runs Live “warpless” rule, which as of the tournament was as follows:

To prevent players from warping by intentional game-over: If you game-over anywhere on the West Continent (i.e. before you use the raft) you must return to the place where you game-overed. If it was in a wandering monster encounter, you must return to the overworld location where that encounter had occurred. If it was in a fixed encounter (e.g. a cave or bridge) you must re-enter that fixed encounter. If the encounter has multiple sections, you must return to the section where you game-overed. You may re-enter the encounter/section from a different direction if desired. If you game-over in a palace you must return to that palace, but you do not need to return to the room where you died.

What this rule does not specify is the route you must take to return to where you got a game over. This usually was a friendly hand to people who got a game over coming out of Spectacle Rock with the hammer. The player would be allowed to return there by the bridge from the graveyard, instead of going all the way through Death Mountain again. The idea is that it’s still a massive time loss, and never an advantage to take an intentional game over, as it’s never a useful warp.

One player in the tournament took it further though, after a game over at Spectacle Rock. That runner got the Fairy spell and Downstab before going back to Spectacle Rock. By the rules as written, this was perfectly legal, and I okayed it as the main tournament organizer. However this ruling sparked a discussion in the community, and may lead to a changing of that rule on SRL. If the rule changes on SRL as a result of community decision making, I will have next year’s tournament use that new rule.


Every tournament teaches the community something. In 2015 we (Svenne1138 and I) ran the first tournament with lessons from the 2015 Legend of Zelda Any% No Up+A tournament fresh on our minds, as we waited until that tournament was in its late stages before starting ours. This year we had lessons from that first tournament, as well as an entire year of other tournaments, to guide us. That’s what led to the new structure. These giant double elimination brackets in my opinion are too much, take too long, and have too many bottlenecks. They create pressure on the Bonus Bracket to hurry up, and lead to a massive stall in the Winner’s Bracket.

2017’s tournament therefore will be very similar to this one I expect. We’ll just use Google Forms exclusively in place of Challonge, give people more time to enter, more time to fill out prediction brackets, and finally use a number of Wild Card entries to balance the KO round bracket. It should be fun.

Stuff I forgot

I’m going to edit into the bottom of this document things I forgot to add. Adirondackrick points out that we should add Livestreamer to the list of resources we give restreamers, because it’s great for synchronizing the runners’ streams, making it clear how even they are. I used it in my latter restreams, and it worked really well for just that purpose.


Comments are closed.