Same sort of thread as last year, but I think scouting this year will end up a bit different from usual, with several ways to play defense and score points. So, besides the obvious raw data (balls scored, hanging, etc), what match or pit data do you think will be useful to collect for the coming season (shooting positions? 3pt percentage?)? Any useful ways of measuring a team’s efficiency? Any tips for scouting and data analysis are helpful.
Early going, but what I’m suspecting:
- Make/miss high/low. I’m not sure I can expect a scout to catch inner goal shots, so I expect we’ll scrape TBA’s API for the match data, determine what percentage of the high shots were inner goal shots, and then prorate those extra points.
- Climb (and maybe climb time?)
- Control panel:
- Did it come available? (I guess we will scrape TBA for this.)
- Did they attempt/complete rotation control?
- Did they attempt/complete rotation control?
- Did they do a trench run? I don’t think it’s essential, but in an alliance going for the control panel tasks I think there’s value in a cycler that has protection.
- Defensive moves (you know it when you see it)
Pit scouting will actually be a thing this year:
- Wet weight
- Wheel choices
4 Likes
As an addition, is there any meaningful objective way to gauge defensive/general driver skill? So far, the only way we’ve had is a subjective 1-10 rating on a match by match basis, and an iffy average points scored against metric.
we are going to be recording scores and attempts for each goal to get a rough accuracy percentage. we will also be recording…
-where they shot from
-where they collected from
-time to score
-time to collect
-full cycle time
for the climb we will be recording time from entering the generator till a legal climb(and yes or no if balanced)
the control panel will be if it could have been used, then how long it took to manipulate if used
then your normal general comment section for comments on stuff like defense, etc
for pit scouting we will be asking stuff like what they prefer to do in a match, what goal do they prefer, what shooting location do they prefer. then your more normal pit scouting stuff like weight and drive train specifications.
edit: it would be interesting to have a extra scout on each side of the field whos job is to just count 3pt goals…
Weight, and we’re calculating accuracy for each target
We’ve adopted SPAM’s “defensive moves” stat, defined as “you know it when you see it”. Things like stealing game pieces, delivering a hit, forcing an opponent to slow down, etc etc etc.
I could be wrong, and maybe your team has this tight, but this feels like a way to burn scouts out. Two mitigations:
- To the extent FIRST makes data available, please do try to scrape TBA’s API for this. I’m sure they’ll have climbing data (not necessarily times, but whether it’s up/balanced), and whether the control panel was even an option. Eliminating starting and ending hab levels from the forms in 2019 was a shocking time-saver for us.
- Total match stats are a thing too. Matches are the same length for everyone, so does it matter if one robot is slow to acquire but fast to score and one is average at both if one of them puts more balls in the goal?
1 Like
I.e. number of defensive moves per match?
i dont think we would be recording perfect times. we would likely be doing options a scout selects (fast af, fast, medium, slow, slow af) for collecting and shooting, then a rough cycle time in like 2-5 second increments.
for where they shot, a scout would check off a list of each spot the team shot from during that match, same for collection.
for climb we would have them start counting in their head when they crossed the berm and stop when they are off the ground and enter that.
control panel will just be a yes or no
We’re currently collecting primarily what the robot can do that scores points in a match and plan to encourage points such as ‘fits under trench’ or ‘can’t go over the ledge in the middle easily’ in the notes section so the interface is simple to use but important information that doesn’t change substantially match to match is still collected.
Accuracy is a statistic that doesn’t show as much for the effort to calculate it. The fact is that you should worry about what they can do in a match because a team with 50 percent accuracy but double the cycles is the same as 100 percent for the other team.
We utilize the idea of Super Scouting that 1678 brings up in their scouting resources: https://www.citruscircuits.org/scouting.html
From there, we centralize all the data of qualitative metrics by only have them attributed by two individuals who are our lead scouts that have been around the most, and directly work with myself as the strategy mentor and drive coach.
Otherwise, we will be obtaining quantitative metrics in general scouting which is currently being assessed as the following:
Score Location (miss/make)
Climb (location, time, level)
Auton (cycling location and accuracy)
And some other similar things. We plan to change things once we better understand our “perfect partners” abtrjbutes this week, and finalize Saturday.
How much do you all think recording where a robot shoots from is worth it? Would you draw dots on a map, or somehow break the field up?
Disclaimer: I scouted a shooting game once, and that was during an offseason competition.
I would recommend breaking up the field into sectors a grid; drawing dots on a map would be pretty difficult to record, unless you’re using a digital input system.
Since most teams probably will shoot from the same location multiple times, I think that it’s worth it in order to figure out match strategy - not only how to best defend them, but also ensuring that you don’t pick two shooters that would occupy the same location.
Depending on the analysis software you use, having this data would also allow you create a heatmap for each team of their shooting locations. I know Tableau supports this - not sure about other programs.
Until we do some practice matches, we’re looking at the following:
- Move from initiation line
- Scoring in ports during both AUTO & TELEOP
- Control Panel Manipulation (Turning & Aligning)
- Climb, Climb time, & ability to adjust position on switch
- Possibly pickup location of power cell and shooting location.
Some of this will change as we define what we want in an alliance partner.
We will use super-scouts to observe defensive actions - since those actions are they are a bit difficult to quantify. Limiting those observations to a few specially trained scouts will (hopefully) remove some of the ambiguity.
This topic was automatically closed 365 days after the last reply. New replies are no longer allowed.