The current competition season marks 5 years since the last time an 8th seeded alliance won an official event. As a former mentor on one of the winning teams at Lincoln in 2019, I take a lot of pride in our remarkable accomplishment. However, it is high time for another 8th alliance to share in the excitement of pulling off FRC’s greatest possible upset.
In order to inspire other teams to follow in the treadmarks of 5853, 5530, and 7660, I will be undertaking an analysis of the strategies that enabled our alliance to succeed when so many others have fallen short. Why trust my analysis? While I am only an average strategist by the FRC community’s standards, I have spent years thinking about this event, the factors that contributed to our success and whether they were strategic or fortunate. My hope is that when teams find themselves on the 8th alliance this season, they have a little more reason to believe in their own potential. And, of course, I hope that at least one alliance this season can fulfill that potential and achieve the ultimate victory.
This series will be divided into three parts. In Parts 1 and 2, I will focus on the strategies that were employed in 2019 during alliance selection and playoff matches, respectively. In Part 3, I will apply the principles gleaned from our 2019 victory to the 2023 game and make some recommendations of how the 8th alliance can maximize its chance of success.
A few notes before I start:
- In no way do I claim to be providing advice that will guarantee victory for an 8th alliance; even with the best possible strategy, most 8th alliances will still be eliminated. Instead, my aim is to increase the odds that the right captains at the right events with the right robots available can take advantage of their circumstances and win an event.
- For reference, here is the 2019 Game Animation. This overview should be a sufficient introduction to or reminder of gameplay. I will highlight any relevant rules that go above and beyond these basics.
- By my assessment, there have been several major changes in FRC since the 2019 season: the proliferation of swerve drives and Open Alliance blogs, the accessibility of real time scouting data from sources like Statbotics, and the new playoff format. I will be holding off until Part 3 to directly address how these may affect the 8th alliance’s chances.
General Conclusions for the 8th Alliance’s Selection Strategy
Before diving back into the details of an event from 5 years ago, here is a brief overview of the conclusions I will be deriving from my analysis of the Alliance Selection strategy implemented at Lincoln.
- Events with a deep field and fewer elite robots produce the best environment for evenly matched alliances and therefore big upsets.
- Autonomous and End Game scoring should be slightly prioritized over Teleop scoring given their higher point values and restriction on defense
- Look for teams that excelled in their last few qualification matches, even if they have low average metrics
- One selection should be a higher risk robot with a high ceiling for scoring
- The other selection should be a consistent performer in a facet of the game where the other robots are weak
Lincoln 2019 Alliance Selection Overview
If you are like me and want to assess how an alliance pulled off a playoff upset, then your first stop is the Alliances tab for an event on Statbotics. Let’s pull up Lincoln 2019 and take a look at those bar charts (using Expected Points Added or EPA as a rough metric for robot ability).
Evenly matched alliances are the first ingredient in the upset special recipe, and the parity at Lincoln is immediately apparent. This level of competitiveness requires a deep field of robots from which to pick during both rounds of alliance selection. The data supports this: Lincoln featured the 10th highest average EPA and the 12th highest Top 24 EPA out of all district and regional events in 2019.
This means the 8th alliance captain, 5853, FEAR the Termigators, had plenty of strong robots available when it was their turn. They ended up selecting 5530, The Greenhills Lawnmowers, and 7660, The Byting Irish. Let’s switch to the Figures tab to get a sense of 5530’s and 7660’s relative ability.
Even considering Lincoln’s competitiveness, it is probably shocking that the 6th and 8th best robots by EPA were still available for the 8th captain. Let’s explore the contributing factors.
Since EPA was not a widely available metric in 2019, OPR (Offensive Power Rating) was the best option for teams without robust scouting operations to use as a holistic metric of robot performance. 5530 and 7660 had the 11th and 13th best OPR at Lincoln respectively, making their availability at the end of the first round a bit more understandable.
However, switching to OPR raises just as many questions about the draft order. For example, some of the teams with the best OPR were not picked until the middle of the second round. Conversely, the 3rd overall pick barely cracked the top 20 for OPR. Clearly, OPR (and EPA by extension) have serious limitations when it comes to explaining the selections made by alliance captains. This indicates specific capabilities were valued higher than overall performance.
Sure enough, the field of teams at Lincoln featured a wide variety of robot archetypes with few all-around exceptional teams. Robots could largely be sorted into a couple categories: “Climbers” who ranked high but struggled with speed or game piece manipulation, “Hatch” or “Cargo Specialists” who focused on 1 game piece, and “Everybot” class robots who scored fast, but were limited in where they could collect and score game pieces. knowing their own strengths and weakenesses, alliance captains often needed to pick a robots that offset their own robot’s deficiencies.
A good visualization of the different robot capabilities can be found in the graphs produced that year by Alicia Bay using Caleb Sykes scouting database. Between the two graphs, you can get a sense of the strong Cargo or Hatch bots (Game Element Analysis) and the strong Climbers (Overall Earned Points Analysis). Once you filter down to FIM Lincoln on the graphs, make sure to also set the 2 filters in the bottom right corner to show “All”, otherwise you’ll only see the robots that qualified for the World Championships.
The 3rd alliance from Lincoln is a good case study of how this affected alliance selection. The captain, 3656, was among the strongest climbers and a consistent cargo scorer, hence their high rank. However, they had a limited ability to score hatches, leading them to pick 6570, the best Hatch specialist at the competition who could score at all locations. In the second round, they snatched the best overall game piece scorer available in 5067. While you could quibble that such and such robot would have been a better selection, the 3rd alliance’s Finalist medals vindicate 3656’s alliance selection strategy.
The 8th alliance
When the first round of Alliance Selection reached the 8th alliance, 5853 found themselves in the captain spot. Average at best (22nd in EPA, 28th in OPR), they needed smart selections to remain competitive. The first order of business was to ensure their alliance featured a robot with a Level 3 climb capability.
A Level 3 climb was worth the equivalent of 4 cargo cycles, so alliances lacking a robot with that ability would be at a massive disadvantage during playoffs. In fact, only one alliance at Lincoln won a playoff match without a level 3 climb (Semi 2 – Match 1) and plenty of alliances lost matches because their Level 3 climb failed.
This observation can be extrapolated to a general selection principle for 8th alliances: place a high value on ensuring your alliance’s Autonomous and End Game scoring keeps pace with the competition. Autonomous and End Game tasks are typically worth more points than those in Teleop. Without these points, your alliance either starts with a massive disadvantage early in a match or has to fight to build a large enough lead to offset the influx of points coming at the end.
Additionally, it is important to consider the effect of defense. Defensive strategies are more common during playoffs and this can dampen the scoring output of many teams during Teleop. However, defense is usually restricted during Autonomous and End Game. That means teams’ Autonomous and End Game scoring output during qualifying is likely identical to what teams will produce during playoffs. Prioritizing these facets of game play sets a scoring foundation for the 8th alliance and gives them more freedom for developing a Teleop strategy without the burden of needing to chase the match.
In terms of 5853’s options, none of the remaining teams had demonstrated a consistent climb. Nevertheless, there was a clear best option hidden in the data.
5530 – The Greenhills Lawnmowers
The choice of 5530 may seem pretty obvious in hindsight; of the robots left with a demonstrated Level 3 climbers, they had the highest game piece scoring average. However, it’s essential to note that the provided graphs and metrics all incorporate playoff performances. Remove the 6 playoff matches from the dataset and 5530 was a riskier pick then this hindsight suggests. Let’s switch to what scouts would have actually seen.
5530 Level 3 Climb Performance
- Match 1 – Fail (climb awarded by penalty)
- Match 9 - Fail
- Match 15 - Fail
- Match 24 – Dramatic Fail (fell over)
- Match 33 – Success!
- Match 39 – Defer to alliance partner
- Match 42 – Dramatic Fail
- Match 52 – Fail
- Match 59 - Success!
- Match 68 – Died on field
- Match 74 – Success!
- Match 80 - Defer to alliance partner
This is not a table that suggests playoff glory awaits. Considering many teams make major alliance selection decisions following Day 1 of qualification (after Match 60 in this case), 5530 would not just have been a risky pick for a top alliance; they would have been a foolish pick.
Nevertheless, the data indicates 5530’s climb performance was improving over time. Remove their early matches and matches when they deferred the Level 3 climb, and their success rate jumps to 50%. More importantly, 5530 successfully climbed during the last two matches in which they attempted it. This is exactly what makes the ideal pick for the 8th alliance: too inconsistent to be picked already, but trending in the right direction. Needless to say, the pick worked out: 5530 successfully climbed to Level 3 in every single playoff match.
It’s a well-established principle in FRC that the captains on higher numbered alliances ought to take the risk of picking inconsistent but high potential robots to improve their chances. However, some risky picks are smarter than others, and 5530 is a shining example of a smart risk. As alliance selection progresses, the importance of average performance shrinks and the importance of performance trends increases. In fact, there’s an argument to be made that the 8th alliance captain should only consider performance data from Day 2 of qualification, as robot capabilities can often change drastically over the course of Day 1.
Regardless of how teams conduct their data analysis, the implications of 5530’s performance are clear: the 8th captain should select one robot with strong scoring performances in later qualification matches, even if its overall performance is inconsistent.
7660 – The Byting Irish
One of the key challenges of the 8th captain is handling the pressure of making consecutive picks. To mitigate this, it is essential for the captain to have an alliance architecture already in mind. Fortunately, 5853 and 5530 agreed on the selection of a very different type of robot from what 5530 offered.
7660 were the opposite of 5530: low capability but consistent. The robot did not possess mechanisms for climbing, ground collection, or lifting game pieces. This meant it was limited to collecting game pieces from the loading stations and scoring at the cargo ship and lower level of the rocket. However, it made up for its deficiencies in versatility by running smooth, consistent cycles, leading to a high game piece scoring volume. The result was the 3rd highest Teleop EPA and, according to component OPR, the 3rd highest average number of game pieces moved.
The most fascinating aspect of this pick is that, in addition to 7660, the 8th alliance had several robots with similar scoring profiles available, including 66, 6618, and 5067. So what set 7660 apart? Let’s look at a table that combines Teleop EPA from Statbotics with component OPR data from Alicia Bay and Caleb Sykes.
| Team | Teleop EPA at Alliance Selection | Avg # of Game Pieces Scored | Avg Hatch Panel Pts | Avg Cargo Pts |
|---|---|---|---|---|
| 66 | 16.9 | 7.5 | 5.2 | 14.5 |
| 5067 | 13.6 | 6.1 | 4.2 | 12.2 |
| 5530 | 14.5 | 4.8 | 3.4 | 9.2 |
| 6618 | 16.0 | 6.5 | 5.1 | 11.8 |
| 7660 | 17.3 | 7.1 | 5.6 | 12.8 |
As mentioned earlier, the last three columns incorporate playoff performance, so Teleop EPA is the most reliable barometer of robot performance during qualifications. The metric suggests that 7660 was the highest scoring Teleop robot left for selection. Furthermore, 7660’s performance trend was also positive; in their last qualification match, they scored 8 cargo during Teleop for 24 total points, far outstripping their average.
However, it is not that simple. 66’s performance is not too far behind, with the component OPR data actually giving that robot a slight edge. Additionally, the 66 robot featured both a ground intake and an arm that could reach to higher levels of the rocket; that means more pickup and scoring opportunities. Using the logic of the 8th alliance’s 5530 pick, would it not have made more sense to take the risk of selecting the robot with the higher scoring potential (66) versus the robot with the slightly better qualification statistics (7660)?
It was a close call and 66 would have been an excellent pick regardless, but the answer comes down to building a balanced, complimentary alliance. 5530’s strengths lay in a high scoring potential during end game and versatility during Teleop. However, since 5530 had not demonstrated consistency and Teleop scoring volume, the other pick needed a track record of high game piece scores to ensure the alliance always had a baseline of points to keep matches competitive. This is especially important given the increased competitiveness and defense during playoffs makes Teleop performances less predictable. Sure, two riskier, higher potential picks could line up to produce an incredible match or two, but usually the inconsistency catches up. (Remember the goal is to win the event, not just pull off an upset or two).
Too much consistency without potential and you put a ceiling on the alliance’s performance; too much potential without consistency and the floor can drop out.
Therefore, the 8th alliance selected the best available scorer in 7660 without considering if they had a higher upside.
A Final Consideration
While it was not a major factor at Lincoln, there is one more consideration the 8th alliance should make, especially when deciding between similar robots in the later stages of alliance selection - pick a team with whom you work well. Compatibility, comradery, and trust go a long way as your alliance negotiates strategy and handles adversity, even if it requires slight sacrifices in robot performance.
That does it for Part 1. Thanks for reading to the end and look out for Part 2 on the 8th alliances Match Strategy and Part 3 on how to apply these principles to 2024.