I have been trying to make an algorithem based on solely scouting data to kind of simulate how will the alliance selection go.
This is ofcourse just a base to the actual thinking of the alliance selection.
In 2023, (haven’t used it in a comp yet)
I had took a few competitions scouting data, ran it, and the results were pretty good.
Last year if someone would do cubes and score really well, they would be a great match to someone who does cones and scores really well,
and defense was not a major factor in picking a certain team.
But this year, as I view it, it is a different story that I am not sure how to approach.
Offense: I think that a good alliance would be like 2 speaker, 1 amp, and 1 of the speaker scoring would do amp optionally.
So lets say I have an alliance leader who does speaker really good, they would probably rather team up with one of the highest scorers that is able to do both amp and speaker but better in scoring into the speaker.
Then second pick would be like a nice amp scorer.
That can work, but then we add another big factor…
Defense; I think that meanwhile some teams will have insane shooters that are really hard to defend, defense would still be a major factor here.
But now if defense is a major factor, it also means that now my offense would change drastically.
So I am curious what do you guys think will go into the selection, especially which factors will go into it.
No models can catch the nuance, let alone map that to a limited selection of teams
Collected data almost always has data substantial errors with no sanity checks.
The biggest thing is to approach the process from multiple angles and be open to crafting something that is not the status quo.
Second picks are often made (by teams that really know what they are doing) based off some metric that isn’t number of game pieces scored. Often it boils down to “competent”, solid field awareness, and demonstrated reliable endgame in the last 5 matches.
I get what you are saying with the second pick, but I am not sure this season endgame playoffs matter that much especially for 2nd round picks.
I am pretty confident no 2nd round pick would be able to do the trap, which is the major part of the endgame (by points)
Although it is an interesting point for the climb, since harmony does not seem to hard, so a robot that is consistent with the climb can def help with the endgame points.
But I think 2nd round picks would be good for defense and amp, doing the “dirty work” while the better robots would be the main scorers.
Like, I don’t want to pick a robot that can consistenly climb over a robot that has great or even okay amp cycles, or good defense
This is all very difficult to say until we see some actual gameplay, and even then things change from week to week.
When building an alliance you need to focus in on letting the alliance work together as a whole, that often means your second pick isn’t the best remaining in any one metric, and may be chosen off of a single flash of brilliance that was noted sometime during quals.
Point is those “flashes of brilliance” or “all around solid choice” or “would work well with the other two bots” are qualitative things that are really hard to model.
In addition to what others have said about looking for different criteria about the alliance as a whole working together, one thing we’ve been struggling to model:
How do you handle the increasing need to gamble as you work down to the lower alliance captain spots? That is, generally speaking the 1 seed will pick on consistency, while the 8 seed may need to pick on peak performance, then hope to get lucky. Everyone else is somewhere in between. That is also a significant factor, and depends on earlier picks and likely competition.
Again, I want to state that this is ofcourse not 100% nor close to it, just good base to start thinking about the alliance selection
After some thought, and seeing some points said here, I decided to go with this logic:
First pick:
1 of the top 5 scorers
1 out of the top 3 speaker scorers
Then, check for amp, if does amp, then add more points.
Who gets picked would be a top speaker scorer, possibly does amp.
Second pick:
Best amp scorer
BUT…
If there is no “good” amp scorers (will be discussed what is a good amp scorer)
It will go for the 3 best defenders.
And the one who has the best endgame, gets picked.
Who gets picked would be either a top amp scorer or a top defender, that will also have a nice endgame.
I think that gives a general view of how the alliance selection will go, but since there were no competitions I am not sure (bummer we are the first competition in the world lol)