Winning with Blue Bumpers: Underdog Strategies in Rapid React

FIRST General Forum
1

Rapid React has proven to be a difficult game for lower seeded alliances to pull off the upset. It’s been well documented that the top seeds are winning events an overwhelming majority of the time. This discussion shouldn’t center around why that is happening, but instead focus on the outliers… What led to the relatively few upsets that have occurred so far? What strategies have proven effective for pulling off the upset? I’d love to see more analysis done similar to the one I’ve laid out below.

This game is also about to drastically change as we get into District Championships, and an insanely competitive World Championship in Houston. The strategies will also evolve, and it’s possible that the field will even out with lower seeded alliances winning at a higher rate.

The upset I’m most familiar with is the 2-7 Matchup at GNR which I will detail below!

Match Preview:

Summary

In this sequence of matches, the 2 seed with 5913, 7028, and 2508 played the 7 seed with 4607, 3293, and 8188. 5913 was putting up 10+ Cargo in teleop with a sub 15 second Traversal, and is one of the best teams in MN. 7028 was putting up similar Cargo numbers with a fast Mid Climb and is also one of the best teams in MN (this event was stacked). 2508 had a solid ~40 second Traversal Climb. 4607 was averaging around 8 Cargo per match throughout the tournament, but averaged over 12 Cargo per match over their last 4 matches with a 15 second Traversal. 3293 was averaging 5 Cargo per match and had a ~50 second Traversal Climb, 8188 played lights out defense with an unproven Traversal climber and 2 ball auto that they finally got working right after quals ended.

QF Match 1:
2 Seed Auto+Climbs: 49 points
2 Seed Teleop: 20 points
7 Seed Auto+Climbs: 51 Points
7 Seed Teleop: 28 points

Tale of the Tape:

Summary

The strategy going into this match was for 3293 to play defense on 5913, and 8188 to play defense on 7028. 7028 had 7 High goals and 5913 had 2 + 2 low goals, but I understand their shooter was damaged part way through the match due to contact inside frame perimeter that wasn’t called. 2508 had a Roborio reboot which took them out of commission defending 4607 for about 30 seconds and allowed them to put up 14 Cargo. The margins were tight but the 7 Seed pulled this one out by 10 points.

QF Match 2:
2 Seed Auto+Climbs: 54 points
2 Seed Teleop: 26 points
7 Seed Auto+Climbs: 38 Points
7 Seed Teleop: 24 points

Tale of the Tape:

Summary

7 Seed ran with the same strategy this round… with less effectiveness now that the 2 Seed was fully functional. The 7 seed also had much worse auto luck this round with only 3 made instead of 6. 7028 made 5 Cargo before their shooter broke and 5913 made 8 under heavy defense. 4607 regressed a bit under heavier defense and scored 12 Cargo in teleop.

QF Match 3:
2 Seed Auto+Climbs: 47 points
2 Seed Teleop: 16 points
7 Seed Auto+Climbs: 57 Points
7 Seed Teleop: 22 points

Tape of the Tape:

Summary

A much better auto performance gave the 7 Seed a 32-22 lead out of auto. 7028’s shooter was broken going into this match but that wasn’t known by the 7 seed until well into the match. The 2 seed changed strategies to have 7028 play defense on 4607, and 2508 played counter defense on 3293. 8188 played counter defense on 7028… so it ultimately came down to a 1v1 with 4607 and 5913 as the only viable scorers.

Takeaways:

  • It’s evident from the scores that a lower seeded alliance must stay competitive in “undefendable” climb+auto points, this is where lower seeds lose most often
  • Slowing the game down with a lot of defense works in the favor of whoever has more undefendable points
  • A heavy defensive strategy tends to result in more robot carnage which leads to less predictable outcomes
  • Even a really competitive lower seeded alliance needs a lot to go right to pull off an upset

TL;DR and Moving Forward:

  • As fields get deeper and robots get more competitive, we’ll start seeing the undefendable point totals even up between all alliances
  • In Houston, I’d expect pretty much every playoff alliance to shoot 8 Cargo in every match - There’s likely going to be really high variance in what shots bounce out with 16 Cargo launched to the Upper Hub in 15 seconds
  • The auto variance noted above will be enough to generate upsets
  • I expect almost all playoff alliances to score around 40 points in the end game with some teams staying on the field and rapidly scoring, with most others climbing late to Traversal/High.
  • Triple offense likely becomes a lot more viable with 3 highly skilled and well driven teams sharing the field, this will help lower seeds keep up in teleop versus the top alliances, a well coordinated alliance that effectively plays hybrid offense and defense has insane potential in this game if executed well
  • I’ll predict that lower seed upsets get much more common as the season progresses for the reasons outlined above
21 Likes
2

I was watching some of the matches in that 4607 run to finals yesterday as one of the only examples of 1O 2D I can find. The other example I can find (that would have worked) is this one: MACMA 2022 F2-1 - YouTube, which was replayed due to a field fault (carpet ripped up in front of the blue hangar preventing 3205 from reaching the mid rung). But frankly watching it again, it only “works” because blue had a large advantage out of autonomous.

The only good example of Triple offense I’ve found is the 4 seed’s win at Waterloo, where they were able to get 1114 as a third robot (not happening at any ordinary district event!)

There’s a lot to take from the Great Northern and Waterloo alliances that demonstrate why these strategies worked in those specific cases and not in others

4 Likes
3

A good case study, although 1 ended up winning the series, would be 1v8 at the Macomb event. Incredibly close series despite it being a 2-0, with match 1 coming down to one blue robot not climbing. (I also believe the 1 alliance captain in match 1 had a non-working intake but match 2 was similarly close)

3 Likes
4

Hey Ryan, great write up!

A couple of things to note:

  • 5913’s limelight crashed during their 3rd teleop cycle in match one, which lead to them scoring zero cargo the rest of the match.

  • 7028 bent their indexer motor shaft with 1:06 remaining in match two, and did not get it repaired in time for match three. They could not shoot for the remainder of match two, and the entirety of match three.

While I think that the blue alliance’s defensive strategy was awesome and definitely played a massive role in the upset, match two is probably the best indicator of how it performs against a healthy alliance.

3 Likes
5

Right, it was 5913 who sustained damage inside the frame perimeter that was not called, as discussed here.

image

3 Likes
6

Additionally, If your going to send a robot on defense as the 6-7-8 seed, you might wanna take a good look at your whole alliances scoring power vs driving ability. A lot of the time you might end up with a captain or first pick who can play some shut down defense and is only 1-2 cycles faster than the second pick robot.

5 Likes
7

I’m from 7028, so I’m biased, but it’s also worth noting the advantage coming out of auto was with 7028 unable to shoot cargo. 7028 had previously scored 8 and 12 cargo points in matches 1 and 2.

Overall, defense does seem to be the unpredictable component in the finals that may not be reflected in the qualifier rankings. There is usually much less defense before the finals, so the driver’s ability to counteract it is less known. Additionally, durability of robots to stand up to repeated high-speed rams and smashes from an opponent may only be determined by pit scouting during qualifiers.

2 Likes
8

I believe 1O 2D works best when your alliance has the best offensive robot on the field, but overall your alliance would be outgunned in a purely offensive or standard single defense situation. In this case, both defenders have a high scoring robot to defend and it can really open up the field for their teammate.

This type of matchup doesn’t typically happen in elims, but can often happen in quals, and it’s a strategy we used in Qual 39 of FIN Columbus. We had this match marked our schedule as a possible loss. We were against teams that would eventually rank 3, 8, and 12 and partnered with teams who were struggling up to that point.

Not necessarily a “Blue Bumper Underdog” example, but definitely an example of turning the tides on a possible loss.

2 Likes
9

Mostly agreed. The trouble is if you assume each offensive robot is equal, and each defensive robot is equal, one alliance has one robot scoring, the other has 2 robots scoring, it’s basic math.

You need to alter the assumptions somehow, by taking advantage of some robots being better at defense or some robots being more easily defended, or by an overall strategy making the combination of your 2 defenders greater than the sum of their individual defensive contributions.

2 Likes
10

6421 won in blue at both Wisconsin and Seven Rivers.

Honestly both times came down to picking a better 3rd bot than the 1 seed. We make a list on Friday night of potential teams and sort them throughout the day on Saturday. Basically each mentor takes a robot each match to watch qualitatively to help us make informed decisions. We’ve been putting a ton of effort into increasing our scouting and evaluation processes and it was awesome to see that pay off in such a competitive season.

4 Likes
11

Would you care to share what qualities you were specifically looking for in your 3rd bot that were passed up by other alliances? What jumped off the page in your data, or what qualitative information did your mentors see that didn’t show up in the data?

12

Sure, so 1st pick is usually fairly straightforward, sort by average score in our data, make sure climbers are reliable, keep an eye on them Saturday to make sure nobody starts having major issues.

2nd pick this year we decided to prioritize a bot with decent defensive capabilities, with some form of climber and hopefully an auto. So to start we look at the reliable traversal climbers left and check how long they take, what else they do, ect. We work that down to high and mid bar climbers. While doing this, we are watching matches our scouts noted good defense in, and also checking drive trains (unfortunately mechanum this year has been resulting in automatic DNP for 2nd round, as well as designs prone to tipping and getting into opponents frame perimeter) .

From qualitative data, we talk to teams on our 2nd pick list if we’re in a captain position on Saturday to let them know that we’ll be keeping an eye out on their defense, and mentors are watching to see how bad they foul, how effective they are at interrupting shots, that kind of stuff.

Bear in mind that we were the 3rd and 2nd seeded alliances, for lower seedings another strategy may prove to be more effective.

13

We ran a triple offense as the 4 seed against the 1 seed at MACMA. We gave it a good run in SF 1-2. Our second pick (6333) was valuable for our alliance as a potential 2 ball auto, and 6-8 cargo per match, plus a mid bar climb. We had discussed a defense robot but knew we’d need a little more offense to keep up with the 1 alliance.

The ability of our second pick to “get in the way” on the opposite side while still scoring almost sealed it for us. The match could have gone either way, but it was nice proof that triple offense could be an effective strategy.

1 Like
14

This might warrant its own thread but would that teams should go for unconventional auto timings to decrease the chance of shot collision. Something like shooting the first ball immediately would reduce the chance of colliding with the teams who shoot after intaking the first ball

15

This topic was automatically closed 365 days after the last reply. New replies are no longer allowed.