Usually, a team’s ideal scouting spot is the center of the field. We have realized that teams cannot accurately scout whether the balls go into the inner goal or the outer goal, and better teams may have slower cycles because they are consistently making the 3pt shot-- which subjectively is better. One solution is to position 2 superscouts to run back and forth, however, they’d only get 2/6 robots, or overlook the smaller teams who may have improved over build.
Use your practice matches to test. I’d suggest you may want to go look at going to the other side of the field if possible for a better angle. What did people do in 2017? My first year doing FRC was 2019. That would have been really hard.
I’m pretty sure we need to have a scouting sheet and system in place before hand- but yeah, definitely
2017 had less emphasis on the balls, my first year was 2018. But it was clear when a robot was shooting steam into the boiler (i think that’s the terminology…), the confusion here is more centered on how can you tell when someone makes a 2 point shot versus a 3 point shot.
I’m pretty sure he means that the ones that are making enough 3 point shots to be statistically significant will not be available for you to pick (they’ll be seeded above you).
My team still believes having all the data is useful. We tend to rank even the top teams the night before alliance selection to have a ‘mock draft’ of sorts. What’s more important are the above average teams that sometimes have really good years and other years slip and fall. Knowing is better than being unaware of how each team ranks and stats. (it’s a fun point to have a celebration when you can beat a top seeded team)
Also, what about midtier robots that can occasionally aim and make a 3 point shot? There is higher marketability than robots who simply cannot.
I agree that collecting the data could be useful. (what if your seeded 1st? Wouldn’t you rather have a robot that can consistently shoot the inner goal than a comparable one in all other aspects) It’s also just fun to figure out who’s best.
That being said, I can’t currently think of a really good way to insure this. The back of the ports are clear so maybe someone can stand behind and watch, but I think the divider between the two (inner and outer) is also clear so you will have to have really good eyesight. And many events the whole area is blocked behind the driver’s and you can’t stand there.
The blue Alliance wil probably have a breakdown of the stats, but not individual robots. You may end up just having to understand that you will have flawed data in that area and adjust decisions you make off of it accordingly. Also later in the season as everyone thinks about the game a little more you may get better suggestions.
I think that live scoring will definitely be helpful with this as long as the scout understands the point values and multiple teams on the same alliance aren’t scoring at once.
Standing at the back wouldn’t allow you to see which robot had scored. But thanks, I think my team will work on developing a way to count cycles over total points score (kinda like how OPR is calculated) this does mean teams that can be carried by partnering with God bots though, which is what scouting is used for.
I’d have to watch some matches before deciding what to tell the scouts, but likely I’d say to swap to watching the screen if they are about to score and are the only one from their alliance scoring at the moment.
Big fan of a single scout watching a specific bot only we want to watch (partners/foes) in that match (focus) then on higher performing scores , go deeper as counting behind the inner port counts for performing teams with a stopwatch. lessen distraction get better data. Paper is faster for observational notes than a tablet… have one scout each match look overall for performers out of the six.
Blue alliance has plenty of data from FMS…watching closely tells you who is good. Pen , highlighter excel with watch lists. Answer the how.
Our current plan is to count the high goal shots without worrying if they are the 3pt shot.
We may revise this after doing some live stream scouting. We usually have one scout watch one robot and we think we could track the 3 pt shot by keeping an eye on the scoreboard.
We try to keep scouting as simple as possible. As of right now, we think that knowing whether a team can consistently hit the high goal is enough and the extra work to find out if those goals are 3pts is going to be too much trouble for the benefit that having that info will give us.
In that case, would you divide the scouting team, 3 for blue and 3 for red and to come back together after? That might be a little disorganized, but yeah, i currently can’t think of any better solutions.
Karthik used that name well before I did, so if there are ownership rights he should get them.
Relatedly, I made a scouting assistant tool in 2017 to solve a similar problem where it would be nearly impossible for human scouts to accurately count fuel scores. I don’t think hardly anyone actually used it though since fuel scoring was so rare. If there is interest I could build a similar tool for 2020, but I don’t want to bother unless I get requests.
Another option would be to guess at the 3 point scores based on 2 point accuracy. Like, find the relationship between misses and 3pt shots. For example, a team that misses the outer goal 20% of the time might only hit the inner goal 10% of the time. While a team that only misses the outer goal 5% of the time might be hitting the inner goal 30% of the time. Not sure exactly what that function would be, but a good empirical derivation could be made from 1 event worth of data.
Our scouts have split up in the past (voluntarily), although that was at an offseason event with significantly more available seating.
Splitting up the team isn’t an ideal scenario, obviously - team unity/volume of team chants suffer, as well as communication. I’d prefer a different solution that still allows accurate scouting of inner vs outer goal shots, but I haven’t seen a better one yet (I’m sure there is one, just haven’t seen it).
Depending on your particular scouting system, it might be easier or harder to split up - I would imagine that this solution wouldn’t be viable systems utilizing ethernet for data transfer on a per-match basis, since the scouts need to be centralized.
On the other side of the spectrum would be digital systems that don’t require cables for data transfer (or even those that do, but transfer data much less frequently) - scouts could conceivably camp on their respective sides and not have to move at all.
Somewhere in the middle is paper scouting - paper definitely needs to be collected in a centralized location eventually, but a good stockpile of forms could reduce how often that occurs.
That said, I’ve only operated a paper scouting-esque system, so there may be some nuances about digital scouting that I’m missing.
I might end up doing something like this. If anyone wants the details for their own implementation (I’ll probably end up writing something in Google Sheets/Javascript to do the math), I believe Calculated Contributions work similar to OPR. The largest issue I could see is that about half of non-high goal robots will see negative inner goal values, as their teammates generally happened to be less accurate with them, as well as that similar teams could be credited for more inner goals than they shot high goals. My math instincts tell me that the proper correction for this is to replace the 1s in the input matrix with the number of high goals made, which should a) return a nice % of high goals made inner and b) prevent teams for being credited with inner goals in matches where they didn’t shoot high. Essentially, instead of solving for average high goals made, it’s solving for average accuracy. I feel like this could get quite close to correct values, maybe closer than scouters trying to see exactly where the ball went, as scouters may be significantly biased towards either crediting the team with an inner goal or not.
