Gettin’ Picky: A Guide to Alliance Selections (Part 1: The Picklist)

Alliance selection is hard, and there are not many resources out there about how to navigate it – we decided to create a comprehensive guide for alliance selection and everything that goes into it.

Picking is an art, not a science, so much of this advice is subjective and reflects our experiences. We would love to hear about other philosophies and methods in the comments.

 

About The Authors (a.k.a. “Why listen to us?”)

 

Katie Widen, self proclaimed Scouting Guru, has been making picklists since 2011. Some of her top hits include 1296’s Turing Division runs in 2017 and 2018. Her work has also been featured by 1675, 3928, and 253. She takes pride in her obsession with spreadsheets and seeing 20+ students working together to create the Ultimate Picklist. 

Brian Maher was a mentor on FRC 2791 from 2016-2019 and helped the team overhaul its scouting and strategy processes. Smart scouting and picking helped them take home their first ever regional win in 2017, followed by three other regional wins, six offseason wins, and getting picked in the first round in three championship divisions. He loves getting students excited about analyzing data and scheming up winning strategies. 

 

What is a picklist?

 

A picklist is a document a team creates to help them during alliance selections. 

• The rankings are not a good picklist!
  • Make your picklist using scouting data
    • If you don’t have any, ask other teams nicely. Someone will share.
  • If you absolutely have to go in blind, use OPR
    • Offensive Power Rating (OPR) is a mathematical estimate of individual robot scoring ability based on match scores. It’s not perfect, but it’s better than qualification rankings. The details of OPR is beyond the scope of this post, but there are plenty of resources about it on Chief Delphi.
    • OPR can be found in the TBA/FRC Spyder mobile apps.
• Your picklist should be 23 teams long, or 31 in a championship division
  • “Even if I’m the 8th seed?” Even if you’re the 8th seed. You never know where you’ll get picked, and there’s nothing worse than running out of teams on your list.
  • Why 23/31? The max amount of teams that will play in playoffs is 24/32 (including yourself). You want to make sure that even if every robot you liked got picked, you still have options.
• Your picklist should indicate which teams you want to pick and why
  • It can be useful to include short descriptions or relevant facts about the teams as quick refreshers.

 

When to make a picklist:

 

• When you are likely to rank in the top 15
• When you are likely to be a first round pick
  • If you can come up with a reasonable argument for why you could be a top 16 robot at the event, congrats, you should make a picklist. 
  • You never know when you’ll be picked earlier than expected.
  • There are always clueless captains – good picks can be the difference between winning the event and losing quarters in two matches. 
  • Even if your captain does have a picklist, you may notice things that they don’t.

How to make a picklist:

 

• Plan the meeting
  • Block out enough time for it in the evening
    • For smaller events, 2-3 hours is a good start.
    • For larger events, 4+ hours.
  • Find a location
    • Meeting in person is always better than tele-conference.
    • Brian’s Take: 
      • For our local regional, we would meet in the school after the bus returned. 
      • For travel events, we would seek a conference room in the hotel. Plan B would be a lobby or lounge, with using a student or mentor room as a last resort. 
    • Katie’s Take: 
      • Local regionals are weird. With 1296 we did everything at the venue because the kids got home so late – a small group did the preliminary sort Friday afternoon and we fine tuned Saturday. 253 has done discord calls. 
  • Have dinner before or during the meeting
    • Hangry scouts don’t make good picklists.
• Discuss your goals for the tournament
  • For district teams, this is generally maximizing the number of district points earned, with winning as a secondary goal
    • The district system encourages less risky picking: each victory in playoffs generates points, while regional teams who lost in semifinals get nothing.
  • For regional teams, this is often winning the event or earning a wildcard invite to the Championship 
    • Regional teams should discuss the difference between optimizing for a wildcard and optimizing for a win. 
• Before you start making a picklist, talk about what kind of alliance you want to build
  • Know what you need to do to win, talk about what top alliances will be
    • Who are the obvious top teams, both by rankings and stats? What will they try to do in the playoffs?
    • If you’re playing from the bottom half (seeds 5-8), ask: what are the long-shot strategies that can beat them?
  • Building the wrong kind of alliance can kill your chances before playoffs even start
  • Examples of poor alliance compositions from the past few years:
    • 2019: Two robots who can only score hatches
    • 2018: An alliance with only one scale bot
    • 2017: Alliances with no kPa potential
  • Watching early season competitions can provide great insight about what winning alliances look like
• Discuss your own strengths/weaknesses
  • You can’t build an alliance to complement yourself without evidence-based knowledge of your own skills.
  • Get a sense for the role you should ideally play on your alliance.
• Once you know what kind of alliance to build, identify needs and wants for picks
  • Needs are the things your alliance requires to execute your strategy of choice.
  • Wants are the things that are optional but preferred and will be a secondary sort to needs.
  • Your needs and wants do not have to be the same for captain/first pick and second pick.
    • Brian’s take:
      • If the requirements are vastly different (like 2019), I like to split up the “picklist” into two separate lists. For example, one list for offense (scoring) picks and one for defense picks. This is okay, and teams can appear in different orders on the two lists, as long as the total number of teams on the lists is 23 (regional/district/DCMP event) or 31 (CMP division).
    • Katie’s take:
      • I’m not a fan of the standard 2 Offense/1 Defense alliance line-up and prefer to have three robots that can all score. This usually leads to one continuous pick list that has a line drawn at the breaking point for a first pick. I don’t want to box our picks into two lists, if I can get two first-pick quality teams then I will.
• Once you have your needs/wants, use them to rank teams based on CART:
  • Compatibility: an ideal partner complements your strengths and helps mitigate your weaknesses. Unless your robot does everything amazingly, picking a carbon copy of yourself is rarely the best approach
    • Is their starting position the same as yours?
    • Are your autons complementary?
    • Do you work best on different sides/areas of the field?
    • Do you require the same feeder stations or scoring locations?
    • Have you played together before? Did it go smoothly?
  • Ability: What can a team do? How many game pieces can they score? These questions are usually answered by taking the _maximum_ of a collection of data points.
    • If it didn’t happen on a real field, it doesn’t exist (probably). Even if there is a video of something working on the practice field in their shop, if it has yet to work on a real field, it can’t be counted on to do so.
    • Counter point: if a team demonstrates everything that is required to do a task, they can probably do that task.
      • Example: A team that demonstrates skilled driving is probably a good candidate for defense.
  • Reliability: How often does the robot do what it can do? How many game pieces do they usually end up scoring? Do they break? These questions are usually answered by taking the average of a collection of data points.
  • Trend: Have they gotten better over time? Good. Have they gotten worse over time? Bad.
    • Recent matches are usually a better indicator of how a team will look in playoffs than early ones. Look at the trends.
    • The easiest way to visualize this is with a line graph.
    • If they got significantly worse over time, you can ask them why. If there was a single thing they had issues with and fixed it, that’s a good sign. If they’ve been chasing various gremlins the whole event, that’s a bad sign.
  • This leads to the consistency vs ability trade-off, i.e. should you prioritize high reliability or high ability? Given two possible picks with the same level of ability, you generally want the more consistent one. When you have two partners with the same level of consistency, you generally want the one that can do more. However, comparing two teams is rarely this clear-cut. 
    • You need to know your opponents: Do you need a reliable partner who won’t drop the ball or a wild card that might have an amazing match?
    • Generally, seeds 1/2/3 will have a captain who is pretty consistent, and thus a consistent first and second pick are usually required to minimize the risk of losing.
    • Generally, the 6/7/8 alliances will need to muster as much scoring potential as possible to have a chance against that consistent 1/2/3 alliance in quarterfinals. It’s okay to be risky here if that’s what it takes to win. 
    • An 8th alliance can be the most consistent alliance there, but if they can never outscore alliance #1, it is literally impossible to make it out of the quarterfinals.
  • If there are any questions that come up while making the list that you can’t answer (usually “why did team X break in match Y?”), write it down and ask/watch that team.  
  • Morning before selection: Look at the specific teams that you wanted more information about. 
    • Make clear what you want to know before you start scouting.
    • Update your list (if needed) based on your additional information.  

 

Common mistakes:

 

• Not choosing a strategy for the alliance
• Picking based on brand name/Not using your scouting data
  • This should be a tie-breaker at most
• Not having a pick-list agenda/organization
• Not having a long enough picklist
• Upping a team because of _one_ good match
• Lowering a team because of _one_ bad match
• Looking blindly at averages
  • You’ll only have 8-13 data points for each team. Look at all of them!

 

Red flags that you’re not making a good picklist:

 

• Everyone disagrees on who should be placed where 
  • You don’t have a cohesive strategy in place.
  • For the most part, how to order teams should follow logically from your needs and wants.
• The obvious best teams are not at/near the top. 
  • Your data is bad and/or your filtering is bad. 
  • This goes against the “don’t pick on brand name” item, so YMMV.
• The top robots are just like your robot… unless, of course, you’re perfect.
• Gut feel – If something feels wrong, it probably is.

 

Signs you that a list is/was a good one:

 

• The top robots have skills that complement yours.
• Alliance cohesiveness 
  • Is strategy a breeze? 
  • Do the teams simply play well together?
• Doing well competitively
  • Winning matches by your own merit – not opponents breaking
  • Pulling off an upset
  • Losing on a series taken to 3
  • Being out-played does not necessarily mean you made bad picks

What do you think? How do our opinions differ from yours? Have any questions? Let us know.

Next, we recommend reading Part 2: Alliance Selection and Part 3: Getting Picked

You can find the conversation on this article here.