Strategy Sub-Team

[Citrus Dad]

Originally Posted by Amit3339

Can you give more information about scouting strategy? I'm not sure I fully understand it's meaning

The scouting strategy is about in-competition strategy. We use it for both match strategy and draft picks. We need to put up our Powerpoint to explain it more fully. I'll see if we can get it up on the white papers here.

[Michael Hill]

Originally Posted by Citrus Dad

We tried heat maps in the first 2014 regional and it was to overwhelming for the scouts and we lost accuracy elsewhere. So we cut back to just counting type data.

Yeah, it is pretty difficult to do well. However, it provides so much valuable information. For example, in 2013, forcing a team to shoot from a spot where they make the fewest shots. You can also find the routes that teams travel through as well and are able to disrupt them enough to lose a cycle or two.

[evanperryg]

Originally Posted by Michael Hill

Yeah, it is pretty difficult to do well. However, it provides so much valuable information. For example, in 2013, forcing a team to shoot from a spot where they make the fewest shots. You can also find the routes that teams travel through as well and are able to disrupt them enough to lose a cycle or two.

That can be much more easily done with qualitative data. Good qualitative scouters can produce that information and a whole lot more, without polluting quantitative data.

[Michael Hill]

Originally Posted by evanperryg

That can be much more easily done with qualitative data. Good qualitative scouters can produce that information and a whole lot more, without polluting quantitative data.

Using qualitative data is fine...if you trust the person/people taking down the information consistently (which rarely ever happens). Having actual data is very useful for the drive team to have at hand when discussing strategy with other alliances. It really takes the "Yeah, we can do [x] every time. Let us do that." When we can say "Hmm...let me show you what you really do. Now lets win the match this way...".

[evanperryg]

Originally Posted by Michael Hill

Using qualitative data is fine...if you trust the person/people taking down the information consistently (which rarely ever happens). Having actual data is very useful for the drive team to have at hand when discussing strategy with other alliances. It really takes the "Yeah, we can do [x] every time. Let us do that." When we can say "Hmm...let me show you what you really do. Now lets win the match this way...".

I'm not saying rely entirely on qualitative data. I'm saying qualitative data streamlines the more detailed data into something much more usable. I only assign the most experienced, most knowledgeable scouters to qualitative scouting. If you can't trust your own scouters to be taking decent data, either they need more training or they just shouldn't be scouting.

[jajabinx124]

Originally Posted by evanperryg

I only assign the most experienced, most knowledgeable scouters to qualitative scouting. If you can't trust your own scouters to be taking decent data, either they need more training or they just shouldn't be scouting.

I agree with how you assign your scouts. It takes more skill, more game knowledge, etc. to scout qualitatively than quantitatively. Quantitative data is just data scouts are gathering that requires basic game understanding(Where scouts are collecting data based of dos and don'ts of a robot and numerical data). Qualitative data asks the scout to summarize the robot's performance and role during a match that the certain robot played during a match(Which IMO takes a deeper understanding of the game and understanding of important roles in a game), so it's a smart idea that a team should assign their more experienced scouts to this role.

You bring up another good point about trusting your scouts to take good data or not. How do you guys train your scouts if it comes to that point?

[evanperryg]

Originally Posted by jajabinx124

You bring up another good point about trusting your scouts to take good data or not. How do you guys train your scouts if it comes to that point?

Generally, we have finished the entire scouting system by week 4 of build, and only make minor changes after that point. So, during week one of competitions, prospective scouters come to a Saturday morning meeting where they learn to scout by watching the live stream of some week one event. This is a great time to assess how well the system works, and to see who on the team should be a scouter at events. Generally, qualitative scouters are students that did quantitative for an entire prior season, who did an outstanding job and have a very strong understanding of the game. Qualitative scouters require no special training, just occasional guidance from one of the scouting leads as to what they should look for. What we ask them to look for can vary event to event or even hour to hour, but there are a few basic things that they always know to look for- deadbots, repeated and egregious penalties, things like that that almost immediately put any team on the DNP come Friday evening.

[gblake]

To those of you familiar with how your own teams do scouting, do the scouts carry out specific evaluations defined/requested by strategists?

Given that scouts could measure/record an infinite number of things, is a strategist explicitly the person who narrows those possibilities down to what is actually collected?

[evanperryg]

Originally Posted by gblake

To those of you familiar with how your own teams do scouting, do the scouts carry out specific evaluations defined/requested by strategists?

Given that scouts could measure/record an infinite number of things, is a strategist explicitly the person who narrows those possibilities down to what is actually collected?

We have one "strategist," assuming your definition of strategist is "that guy that tells the drive team what the alliance probably should do." But yes, especially Friday evening or saturday morning I will ask a pit scouter to go ask a couple teams some very specific questions. Examples of this season's very specific questions include "how quickly can you upright a container?" and "do you /really/ need your ramp in order to load from the feeder?"

[Gregor]

Originally Posted by evanperryg

We have one "strategist," assuming your definition of strategist is "that guy that tells the drive team what the alliance probably should do." But yes, especially Friday evening or saturday morning I will ask a pit scouter to go ask a couple teams some very specific questions. Examples of this season's very specific questions include "how quickly can you upright a container?" and "do you /really/ need your ramp in order to load from the feeder?"

We have the exact same system. Lead scouting mentor and student come down and discuss with our drive team what our alliance should be doing. Time permitting, they come to the pre-match alliance meeting, but are normally quiet unless they drive coach misses something (which isn't uncommon, I miss things ).

[Citrus Dad]

Originally Posted by Michael Hill

Using qualitative data is fine...if you trust the person/people taking down the information consistently (which rarely ever happens). Having actual data is very useful for the drive team to have at hand when discussing strategy with other alliances. It really takes the "Yeah, we can do [x] every time. Let us do that." When we can say "Hmm...let me show you what you really do. Now lets win the match this way...".

When we first put up the smartphone app for our drive team, we called it the "b---s--- detector."

[Citrus Dad]

Originally Posted by gblake

To those of you familiar with how your own teams do scouting, do the scouts carry out specific evaluations defined/requested by strategists?

Given that scouts could measure/record an infinite number of things, is a strategist explicitly the person who narrows those possibilities down to what is actually collected?

I think a better term for what we call a "strategist" really is a match tactician. They're like a basketball assistant coach or football offensive coordinator who puts together the in-match strategy. They use data from the scouting app to identify the strongest teams on both sides and if a particular team can carry out a task. They give some feedback on the stats after the competition so that we can update the scouting app.

[Michael Hill]

Originally Posted by Citrus Dad

When we first put up the smartphone app for our drive team, we called it the "b---s--- detector."

I wanted to have a "BS Factor" put in that reflected the difference of what a team said they could do in pit scouting versus what they actually do. The programmers were about to do it, but since we were sharing with other teams, we decided we probably shouldn't.

[Citrus Dad]

Originally Posted by evanperryg

Generally, we have finished the entire scouting system by week 4 of build, and only make minor changes after that point. So, during week one of competitions, prospective scouters come to a Saturday morning meeting where they learn to scout by watching the live stream of some week one event. This is a great time to assess how well the system works, and to see who on the team should be a scouter at events. Generally, qualitative scouters are students that did quantitative for an entire prior season, who did an outstanding job and have a very strong understanding of the game. Qualitative scouters require no special training, just occasional guidance from one of the scouting leads as to what they should look for. What we ask them to look for can vary event to event or even hour to hour, but there are a few basic things that they always know to look for- deadbots, repeated and egregious penalties, things like that that almost immediately put any team on the DNP come Friday evening.

Our system is similar. We wish we had it built by Week 4...

[artK]

Originally Posted by Michael Hill

I wanted to have a "BS Factor" put in that reflected the difference of what a team said they could do in pit scouting versus what they actually do. The programmers were about to do it, but since we were sharing with other teams, we decided we probably shouldn't.

Why not just not get data points that you can't verify in the pit? It saves time and reduces the amount of noise in your data.