What do you think when you hear “Tablet Scouting” or “Scouting App”

Definitely there is going to be an issue with TBA delay. For the data in general, we collect TBA and match data separately and store them in separate databases. Afterwards when we have the matches in a range we want for both data types, the system will prompt the scout leader to compare and check the data. A lot of the crucial data that comes from TBA we will require scouts to collect.
Our team does have a protocol for training tablet users. Even last year with our very basic app, we made sure to get all students (who wouldn’t be paper scouting, driving, or doing something else that’s crucial during comps) and teach them the basics of the app and what to look out for. Granted, this app did not include things like starting position as we did not rely so heavily on tablet data. Later on, we found out that the data collected could be more useful in match prediction and strategy, but that was a little bit too late as the pandemic situation started to develop after our first competition.

A lot of what I described for the TBA delay plan is currently in a demo or idea form. Preseason for strategy is going to be packed as it’s hard to discuss scouting app stuff other than what I’ve worked on in the past week during a weekly virtual Zoom meeting. Your post will make me think more about how much of a difference TBA delay might make.

Thanks!

Keep in mind the key factors when trying to predict strategy and outcomes. Sample size is a huge factor. I personally don’t see much value in a constant check of BA data versus app data . What we implement is an end of the Day 1 check. This gives us around SIX matches played. Two-Thirds of Qualifications.

Not a huge sample size by any means, yet better than doing validation at that point rather than every single game played per team.

The other thing is you must take BA data with a huge “grain of salt” as its effectively an alliance based stat keeper. Individual teams are one-third of that and "strength of schedule/opponent makeup " plays a huge part.

Therefore, I personally value unlimited individual team observations over the event of a fidelity level that is unlimited in scope. During the event, a strategist might find a metric they want to be tracked. Hard to code that into an app without throwing off the set data collection it offers.

Things to think about. Therefore, we will test/use an app for certain things (items an app does better) but never replacing paper observations that don’t have alliance makeup bias. In effect BA is used only to validate the eyes once.

1 Like

I most definitely agree with both of your points.
With TBA data checking, it’s impossible to get the absolute correct power cell data. There will always be a discrepancy that you cannot find the answer to no matter how many ways you look at it with your app data and the official data (except for rewatching the video, but even then things get a little sketchy). However, with confirming the robot positions on the field, it is possible to check the TBA data to the app data to at least confirm the Y/N specific robot scoring data points. As someone who likes cool shiny new features, it saddens me to say that I am forcing myself not to assume data from alliance based data. However, when it comes to constant TBA collection, I disagree only because of the data we are checking. If my team can completely confirm that in TBA, Red1 robot on the field is red_alliance[0] in TBA?(or something similar to that), there is no downfall in doing a quick check to make sure that the Y/N specific team data is correct.

There is a lot that is not possible to achieve on a scouting app during competition. Paper scouting does much better at seeing what the team and robot are, while tablet scouting is good at showing what they do (at least in my team’s case).
A big reason for this app was the split between the value of paper scouting and tablet scouting. The best plausible scouting app will always be worse than the best plausible scouter and data set for paper scouting. The issue was more of that tablet scouting was being used (again, mainly on my team) to only confirm what was being written down. Being able to see and have a say in the paper scouting sheet, I would say that the data being collected on paper did not intersect with tablet scouting enough to make the data a “confirmation”. The only time that our tablet data actually made a difference was when the paper scouter was not able to fill in some of the fields of our given sheet. A big portion of what I am trying to present to some of my scouting acquaintances is that the tablet data can be used more in the data analysis portion and match strategy (along with having more of an influence on picklist if the data is modeled correctly). If a really good and fast data analysis method is found, then pre-match predictions and strategies with drive team can be possible without having to manually pull all the data from paper sheets.

Thanks!!

For things like power cells, correct.

However, recent years have had some per-robot information delivered through FMS (generally, whether a certain robot accomplished a per-robot autonomous or endgame task). We don’t even put that on our scouting forms anymore, because we know we can scrape it and get the truth.

(You may well know that, but I imagine someone newer to this doesn’t.)

1 Like

This is easy list Top 28, everyone knows the top 12 (captain level) you see it by watching the event even the order most would rank is consistent.

So, what about the rest… What do they add to your game? Yes, scout and rank your own team the same as the rest. What does your team need to beat teams led by the top 12? Track those metrics on the other 16. So, let’s take Deep Space, winning combo was Two rockets ship teams with L3 climbs… does not matter what the app stats say. The app could tell you who climbed how did they climb? How did they play the rocket? How long was the climb at the end etc? Could they still score versus defenses? do they foul ? Do they draw fouls?

Top 28 for your team is all you really need to figure out. The rest likely can’t help you overcome another elimination alliance.

This is how you select the next two teams, which is observational and not stat driven.

Do you want to create the best app,? Take some of this into account rather than raw data collection and number crunching. A simple input from scouts at end of the match : Would you want this team on your alliance, if given a choice? Not a bad data point for an app to track

Assuming you can do the rocket. L3 climb
1st selection best available rocket bot that can L3/2 climb
2nd selection best defender /cargo bot left with L2+ climb if available

Teams scouts liked is best

1 Like

I most definitely agree about what you are saying. Observational scouting is much better for picklists, and the direction I have the app going in right now, paper scouting should stay the norm for that on my team. Tablet scouting for picklist should stay at a confirmation level, at least on my app.

Once the picklist is done however, it becomes a different situation as there is no likable team or better team. It’s about getting the drivers the correct information that the strategists feel is the best strategy for the match based on the alliances. Paper scouting is capable of doing that, but this is where the number crunching comes really in handy. If there are enough fields in the app (already proposed adding more qualitative fields like shooting position and loading position), then a good interpretation can be made of the upcoming match.

Adding an opinion driven question at the end is a great idea to add to an app. I just feel that in the interest of my team, we do not add something like that at the end as we have not taken it into account beforehand. Of course, if I decide to release this out further to the public, I am going to have to do some major changes to the data collection style and model to account for what other teams do.

In a general sense (and even more specific), yes, you are absolutely correct on this. Top 28 is all we need to create an alliance that may help take down the other ones. Based on a lot of these comments in this thread, I have started making a change to my model to incorporate more qualitative things. A lot of the app is not focused on pick list as much, but more on the actual live match strategy and match predictions (predictions aren’t really so important though).

Thanks for your feedback! I’m trying to understand a lot more from my cut off rookie season!

2 Likes

I recommend putting a “comment” box in any app to cover this edge case.

3 Likes