Defense robots as first round picks

We all know that defense played a significantly larger role in this year’s game than most. A few times this year, we saw an unusual kind of pick made: a first round picks focused primarily on defense. In most years, this is insane, but sometimes, good defenders were a scarcer resource than good offense bots. I saw it a few times this season:

  • FMA Champs: 225 declined the #1 seed, 303, to pick 747, a defense and level 3 climb bot who was the best defender there by a mile. They picked up 5401, a top 10 or so cycler at the event as their second pick and went undefeated in playoffs.

  • Tech Valley: 145, the sixth captain who was great at scoring cargo in the cargo ship, needed to do something crazy to have a shot against #1/#2, so they picked 250, one of the two truly standout defenders at the event, and the field was just deep enough to give them a good low hatch robot to round it off. They upset their way to semis and took it to three against #2. Frankly, I was more worried facing alliance #6 than alliance #2 in finals, because their defensive chops made them a scarier wildcard.

  • Battlecry: 195 picked up their Einstein buddies 1073 (the best defenders in the building) from the 7th spot. While the gambit didn’t quite pay off, the alliance was a contender and I don’t think it was an unreasonable move, though it’s not the pick I would have personally made in that position.

  • IRI: 2767 and 1073 were picked first round and primarily played defense. I think it’s really interesting that this happened twice at the highest level of play, but it makes sense: the pool of offense bots for second picks is just that much deeper, and 2481 2767 would have an early second pick with IRI rules, so a first round defense pick was far less of an offensive penalty.

Where else did first round defense picks happen this year? Were they the right (or close enough to right) pick? I think this phenomenon is a really interesting quirk of this year’s game.

Bonus question: did anyone ever try anything really wacky like 1 offense bot + 1 defender + 1 counterdefender in playoffs?


One I know of from Ontario is during the Durham college event (week one) 4476 as the 6 alliance captain picking 4525 and advanced to the finals of said event. For the rational for this choice I would let 4476 speak to that.

Team 461 picked 3494 at the Indiana State Champs for their defense and got an offensive bot as the second pick. I think it was the right choice because we made it to semis over the #3 alliance with stronger offence (in my opinion) .


5943 was a first pick at Tech Valley for defense as well. We think it was the right choice (not to be biased or anything), but our alliance had technical issues so we weren’t able to make it past Quarters

1 Like

At Central Illinois Regional the 5th alliance captain 3695 took 5690 with their first round pick.
They got to the semifinals before succumbing to a very tough number 1 seed in a tie breaker match.
It seemed like that was a right pick.

At the 10K regional 2502 took 5464 as a first round defense pick.
5464 suffered a fatal break in their drive train in the first quarterfinal match so the results weren’t good.

A consideration with taking a defense robot in the first round at a district/regional event was the pre-determined alliance station placement. A first round defense pick would be across from the other alliance’s second round pick. That would effect the sight line situation usually in favor of the alliance with the offense first pick.

135 not getting their level 3 climb two matches in a row also helped a little too :slight_smile: Still well-earned though.

I’d be curious if your scouting data showed 3494 denying more points than 4103 or 1501 would score, who were still available at that point. It’s an interesting choice to have to make for sure, as you’re running out of top offense rocket-capable bots. All the other alliances took defense-capable robots in the 3rd round at Indiana state champs - 3559, 3176, 1018, 2197, 1741, 4926, 1024. I’d just love to hear the reasoning that drove 461 to make that call earlier than everyone else, it’s always cool to hear how other teams think / operate.

1 Like

747 was the best overall robot that frequently played defense in FMA, by a mile. When looking at pure defensive play, there were others that were close. 2191 could play effective lockout defense with their swerve drive. 1279 was an aggressive defender that could both zone out areas and pursue teams. 708 could be an absolute bulldog on defense, as demonstrated in their finals run from the 8 seed at Bensalem, but opted to play offense roles during qualifications at FMA Champs.

But none of these teams presented the total package that 747 did. 2191 could neither climb beyond nor drive off of any level higher than level 1, and couldn’t score hatches (in sandstorm or otherwise). 1279 presented some moderate offensive ability, but was easily victimized by defense themselves and left points on the table in the end game. 708 could score at any level and had a HAB3 climb, but was inconsistent in their sandstorm and end game scoring, and would often take a while to place their designated hatches at the beginning of tele-op (they would set up the level 2 rocket hatches for 1807). 747 could drive off level 2 in sandstorm, had proven hatch scoring on both the front and side of the cargo ship in sandstorm, and had one of the quickest and most reliable HAB3 climbs in FMA. In addition to being a good defender, 747 did everything else you’d want from your defensive robot better than all the other top flight defensive candidates. That’s a big part of what separated them from the pack.

5401 thrived on being overlooked in the draft. They got three banners as a second round pick in 2019, including the aforementioned FMA champs and the Tesla division. The other two instances were the result of a defense-and-climb oriented alliance captain (4454 at Hatboro-Horsham, 346 on Tesla). Having a defender in the first two picks opened up these captains to look for the great offensive teams that had slipped through the cracks, and 5401 kept slipping through the cracks.

1 Like

The 3494 pick was due to seeing how much of a hard counter their play style was to ours. Based on how they completely shut us down in the finals of the previous event, it was determined that if we were targeted we would be contributing very little to our alliance’s score. Both picks at state were the “pick who you don’t want to play against” type.


747 also picked us first as the #1 seed captain at Montgomery, and we picked them first overall at MKM. We’re 0-5 against them, that’s 22% of our losses this season.

I’ve watched all these defense-only robots play and played against many of them. 747 is the best IMO, and I’d put 6443 at #2.


More evidence that the best way to get picked as a defender is to shut down an alliance captain.


Yeah this tends to work pretty well

It’s not exactly the same, but 6443 was the captain of the fifth alliance on Carver and played almost exclusively defense after sandstorm. They were finalists and only lost to the powerhouse Carver alliance on Einstein.


Team 1251 was a 1st pick by the 5th seed aliance at south Florida. They played incredible defence and only had a level 3 climb. Our alliance captain on Newton ,5015, had themselves play defense as well.

2491, North Star Regional


What about defensive robots as alliance captain? 6443 were captains in 4 of 6 tournaments so far and ranked 2,3,3,7 (Houston) in those. Won once, finalists twice (including Houston) and semifinalists once. We were a broken vacuum pump away from finalist at IRI as 3rd pick of second round.


Plus that climb is super fun to watch!

1 Like

How do teams confirm the best defenders?

In an ideal world, I would have an expected score for an alliance and then use the difference between the expected and actual across matches to come up with a defensive delta, but in the real world that is math that I don’t know how to do and I’m unsure of its validity. Thoughts? Suggestions?

inb4 “qualitative data” - I think its easy to over/under-estimate a defender’s effectiveness. Also thinking about how to apply complex math is exciting, even I don’t actually know the math involved.

If you want to calculate OPR and similar stats, take linear algebra. I didn’t and I have some regrets.

EDIT: I mave have just described DPR (Did I get that right?)

DPR IIRC is tracking the difference between scores when you play against them, so it sounds similar.

I do know its not a great metric though.

Those ZEBRA tags plus linear algebra seemed to do a pretty good job when compared to our scouting. Maybe mandate those at all events?

Until then we’re probably just comparing score differential for teams scoring normally vs against that specific defense bot as our best objective metric, and that is super noisy data

1 Like

If there was an easy answer to determining the effectiveness of a defensive player, sports would have figured it out a long time ago. Yet, it’s always remainded on the fringes of sports analytics, and most defensive analytics tend to have a dececnt amount of skepticism and inaccuracy in most professional team sports.