Gettin' Picky, Part 2: Alliance Selection

One important thing imo is not adding on subjective, unproven data into the mix when you already have a plan. If you believe that a different defensive pick is better, you have to think about how well they showed in a defensive bout, whether thats pushing or being pushed, and not just “well their drivetrain sounded good when they told us about it on paper.” Another example is using actual scoring data instead of saying “well they are still getting used to their bot and might score better” just because you know them or they seemed better on paper with public data/data collected during pit scouting. Unless you can bring up some definite proof for the switch or its already extremely close, dont switch it up. For teams that dont use things like Slack or Discord, you cant really have a proper conversation and question all the points, and even if the team does, the rep may feel like they have to answer as soon as possible and might miss points in an online discussion.

That’s a good rule of thumb, but always be mindful, occasionally the second or third seed is the strongest alliance so you’d rather be on the 1/4/5/8 side of the bracket.

Another thing to note is that if you are on the strong side of the bracket and going for a win, or if you are the 7/8 captain, you should move the high-variance teams up on your picklist a fair bit. You’re probably not going to beat the strong alliances if you have “consistent” partners that are consistently worse than the opposing alliance’s robots. Spice it up and pick the team that had a few great matches and a few horrible matches. Then, if they happen to get hot at the right time you can beat the powerhouses.


This is huuuuuge. You have as much time as you need to make the pick. Relax, make sure the people that need to weigh in on the pick do so, and then make the pick.


You are 100% right and I realized that when we were doing our final review - @Brian_Maher and I have been struggling with these posts because there is so many different things that could be mentioned and we don’t want to publish novels.

Related to your interests - I’m interested in the concept of the “probabilities” of a scorched earth, do you think there would be any way to model that?


If i was willing to pay, would your opinion change?


No worries! Your posts are great. You’ll never be able to get 100% of the information in one post unless it were so long as to be unwieldy. I think you guys nail the most important 80% and then the rest of us can chime in on the thread and fill in any other side thoughts that we think are worth mentioning. :grin:

I’ve never done serious modeling of alliance selections. I’ll get there eventually though hopefully. What really bugs me is that I don’t have a good data source for declines, so modeling scorching is very difficult. I’d either have to scrape data off of the alliance selection threads, wait around until declines get added to the api, or build a model assuming declines don’t exist. My suspicion is that scorching could be relatively easily modeled by comparing the “strength” of the inviting team to the “best-case” team that the invitee could otherwise select if they declined, which is basically what you described above.


From this, you could also rank teams on the quality of their alliance selections and build a new metric called SPR (scouting power rating).

/s but only a little bit


Another, kind of tangential, thing to keep in mind that I want to bring up on this topic is making sure that the on field representative is not the target of blame if things don’t go as planned. If they listen to the list and make the picks they’re told to/supposed to, there is still a huge chance that due to numerous other factors things could go wrong and your alliance could be bounced in quarters.

Too many times I see the blame go directly to the field rep or head scout, especially in the case of an upset. In my time as a scouting mentor I personally had a head scout almost quit the team after they made the picks they were told to make and then our alliance was eliminated in an upset in quarterfinals. Some students can latch on to the field rep because they’re a visible person in the scenario and blame them for being the sole cause of a loss.

This may relate more to team culture than alliance selection in general, but it’s something to keep in mind. Make sure the chain of command is established and that people don’t rush to blame (and if someone has to be the “blame magnet” a mentor is probably more emotionally equipped to handle it than a high schooler) but use these situations as learning opportunities to improve the team as a whole.

Hope that makes sense thank you for coming to my TED talk.


We consider our field reps puppets, because we don’t want them to be making split second decisions under the pressure on the main stage. We give them almost no info and all picks are reviewed by lead scouters and strategy mentors and then transferred via phone. This makes it so that if we lose because of a bad pick, it was something missed by everyone and was probably a scouting oversight. We learned this after our rep went up with the spirit award ranking list instead of the alliance picklist.


In essence alliances, 1-4 have a good chance at winning a competition. So your team has a choice of being top 6 or being selected by Captian 1-4. Anything else is much harder to achieve a spot on alliances 1-4. “The surest way to play in the Championship” via game competition.

This is definitely a better chance of making it through the competition as a world championship participant. The stats don’t lie here.

So be top 6 potential top 4 captain -OR- be selectable in second or third rounds by TOP 4. To make it to Championships more reliably.

This is the goal, to have the easiest time getting that world championship entry as only the top 4 alliance teams reliably has the actual reliable firepower to succeed. So its hard be top 6, or last until picks 9-12 or be picked up in picks 20-24

The reason we go 28 deep is to have some choice at the back end.

One thing about declining: it’s not just a question of whether you can make a better alliance by declining, but also whether the alliance you make by declining can beat the alliance that your proposed captain will make if you decline.

So to use your first example, let’s say we’re playing DDS (and let’s ignore A and E to make things simpler). You and B are both hatch-only robots; B is a good hatch bot, but they’re low on your pick list because yours rightly prioritizes cargo bots. C is the best cargo robot at the event by far, and D is also a cargo bot but not as good. None of the other teams at the event are at the level of you, B, or C.

If you pair up with B you will be two hatch bots on an alliance, which isn’t ideal; you can put together a better alliance by pairing with D. But if you decline, B will pair with C and you’ll have a very powerful alliance to deal with. If you accept B, you may not be the strongest alliance possible but you’ll no longer have to deal with another very strong alliance.

This is an important factor to consider at events where there are a few teams at a level far above the rest of the pack.

1 Like

Attitudes like this is what leads to rules about time limits (and the various other rules that had to be written because teams exploited aspects of the tournament structure). Be ready for your pick when your time comes.

Plus it puts your alliance representative in a very uncomfortable position as they’re put on the spot as your team deliberates the correct pick. Personally, I like to avoid doing that to my students.


Obviously there’s more details needed about the rest of the captains, who would be available as a 2nd pick, etc. In that situation however, I would still be tempted to decline B, let them pair with C while we pair with D. In either situation, your alliance will not be the single strongest alliance so it will be hard to win, but I think pairing with D provides you the better chance of moving further in elims. You won’t have to face the BC alliance till finals so hopefully with your 2nd best cargo bot and yourself as the hatch bot you can make it there and somehow pull off the upset. If you paired with B as two hatch bots, you could easily just go down to a more well-rounded alliance in quarters (not considering who’s available as a 2nd round pick).

After I typed that I realized, if I was B in that situation, there’s no reason for B to pick you. If they did their scouting they know you would make two hatch bots and probably not be a good match so your pick list should account for that and plan for the most likely situation of them picking C and you getting D.

1 Like

If our strategy teams are still discussing the pick, having a field representative who knows they don’t need to rush the pick is important. Sure, I’d love to know exactly what pick we’re making instantly but sometimes there are debates and discussions occurring.

I don’t view that as exploiting the tournament structure at all, that’s just taking the time to make a good decision. Field representatives should know that there is no rule or time limit so they don’t feel rushed into making a bad pick quickly.


@Katie_UPS is 100% correct that there is no time limit for teams during alliance selections. As the emcee it is our responsibility to keep the show moving at a good pace and it can make for awkward situations but past that there’s currently no harm. Really our biggest motivation is to get to lunch earlier for all the field crew :wink: .

Honestly I think there should be some sort of draft clock of X minutes per pick (maybe 2-3) that way people know what pace to expect and it makes it make it less awkward and saves face for the person down on the field. If the draft clock were to expire they would be assigned the next highest ranked team.

(didn’t mean to reply to Kevin directly)


Even an informal draft clock may be helpful, with no real penalty, but maybe a better guideline to help reps not feel rushed. No one should ever be rushed into a bad pick.


QFT. The entire event does have at least a soft time limit (need to be out of the venue by X time or pay more), but that’s more affected by how playoffs go. The big thing is that the field crew NEED to be fed, and the earlier they can eat the better because that is their break time. Sure, food can be sent down to the field and consumed between matches–but time between matches can often be in semi-random supply depending on how the matches go.

Also: The more time you spend agonizing over a pick while you’re on the clock during selections, the less time you have to strategize with your partners before your first match!

Think about it this way: Normally, events budget about 30 minutes for the alliance selection. This is followed by 30-60 minutes for lunch (depending on event, schedule, and how behind the event is running). 30/8=<4. Assume that video, intro, and “team transit time” takes about 6 minutes (just to pick a number that coincidentally makes things nice and round) so now 24/8=3. But wait, each team has to pick twice! That means that theoretically, each team should have about 90 seconds to make their selection when their time comes up. Most teams will go within 20 seconds if they’re prepared (early first round and late second round) which does allow for other teams to deliberate a while.

Thought: If a pick clock were introduced, start it at 90 seconds and see how many teams go over–no penalty the first season it’s used.

1 Like

I have seen a fairly long discussion the last two years at champs. In both cases the emcee stayed well away and let the discussion play out. So I was under the impression that the emcees had been instructed to not rush alliance reps.

Of course there are a lot of other factors that go into making the decision. My point was just that when deciding to decline or not you need to consider not only which makes your alliance stronger, but also which gives you the best chance of defeating the other alliances.

1 Like

Completely agree with your point. Sorry i was just picking apart and delving way too far into the hypothetical. I love some good theorycrafting while I’m supposed to be working.