Scouting 2016

[The Doctor]

I'm wondering what types of data you will be collecting on other teams during scouting. I'm not super experienced in it and would like to get an idea of what everyone wants to know about each other, in the hope that it will benefit anyone who reads this.

I'm more interested in data points that can be expressed as variables in a database. Stuff like "strategy" is far too broad, in order for a computer to parse it, it would be better to list specific yes/no questions like "can you shoot in the high goal?".

[Boltman]

Originally Posted by The Doctor

I'm wondering what types of data you will be collecting on other teams during scouting. I'm not super experienced in it and would like to get an idea of what everyone wants to know about each other, in the hope that it will benefit anyone who reads this.

I'm more interested in data points that can be expressed as variables in a database. Stuff like "strategy" is far too broad, in order for a computer to parse it, it would be better to list specific yes/no questions like "can you shoot in the high goal?".

We stated scouting last year and it made a huge difference.

What we like to track is past performance (current year ) and (previous years) both indicate "Class" and expectations.

Then we look at Min and Max score range for current year matches

Then we look at how they achieve max score. What they do best.

Then on ground visuals and interviews exchanging info ahead of matches.

Its good to have a convincing captain to help orchestrate strategy each match based on collected tendencies. That can elevate game results.

I personally think to many data points is unnecessary to do effective scouting. So we try to keep it simple and relevant.

[chandrew]

Are you referring specifically to pit scouting or to match scouting as well? In general data should be more quantitative such as number of high goal hits vs. number of high goal misses, number of defenses breached, etc... Instead of simple yes/no questions.

[Boltman]

Originally Posted by chandrew

Are you referring specifically to pit scouting or to match scouting as well? In general data should be more quantitative such as number of high goal hits vs. number of high goal misses, number of defenses breached, etc... Instead of simple yes/no questions.

Both. There is tons of offensive data posted so no need to ask. This year it'll be more important to find out what defenses teams can't cross as that won't be posted most likely.

[Jarren Harkema]

Are you referring to pit scouting, where you ask teams questions about their robots? Or are you referring to field scouting, where you watch what their robot is actually capable of?

This year, I feel that pit scouting has very little value. Most teams will have practiced with team versions of a few Defenses, and some will be lucky enough to practice with all 8 of them. I think that a team's response is going to vary greatly from actual performance on the field. In fact, if our team is unable to find an official field to practice on before our first tournament, I'm going to push our drive team to be upfront with other scouters. If we haven't tested our robot on an official rough terrain defense, our people in the pit should say that. I'd rather our robot prove it's merit by showing its ability on the field.

That being said, there are a lot of data points to collect by watching a robot perform on the field. Focus on the ways a robot can score, and try to break those down into easy to define variables. As your team's robot is developed, think about the ways other robots can compliment you, or be a detriment to you, and record that information as well.

[Boltman]

Originally Posted by Jarren Harkema

Are you referring to pit scouting, where you ask teams questions about their robots? Or are you referring to field scouting, where you watch what their robot is actually capable of?

This year, I feel that pit scouting has very little value. Most teams will have practiced with team versions of a few Defenses, and some will be lucky enough to practice with all 8 of them. I think that a team's response is going to vary greatly from actual performance on the field. In fact, if our team is unable to find an official field to practice on before our first tournament, I'm going to push our drive team to be upfront with other scouters. If we haven't tested our robot on an official rough terrain defense, our people in the pit should say that. I'd rather our robot prove it's merit by showing its ability on the field.

That being said, there are a lot of data points to collect by watching a robot perform on the field. Focus on the ways a robot can score, and try to break those down into easy to define variables. As your team's robot is developed, think about the ways other robots can compliment you, or be a detriment to you, and record that information as well.

The way we do it is get eyes on every bot.... that is assuming there are videos, in person or watch prior matches...nothing replaces eyes on bots.

Then we look at collected statistics to either confirm or question what our eyes saw.

The pit scouting is a second layer of confirmation, perhaps they are working on something new. Or to understand more about what we already witnessed and to find ways to enhance play together. Same with the opponents we share info back and forth because you never know if they will be in eliminations later.

So instead of solely numbers, we put in the eyewitness accounts that show all the little nuances in play.

In almost every case we could predict the matches we would win or lose ahead of time and get the score pretty close too. drive team said how valuable the intel was each match.

[The Doctor]

Originally Posted by Jarren Harkema

Are you referring to pit scouting, where you ask teams questions about their robots? Or are you referring to field scouting, where you watch what their robot is actually capable of?

I wasn't really thinking of a specific type. I assume that we could go around to pit areas and ask questions, and then corroborate those answers based on their performance in matches. It's the same data in most cases.

[New Lightning]

Originally Posted by The Doctor

I wasn't really thinking of a specific type. I assume that we could go around to pit areas and ask questions, and then corroborate those answers based on their performance in matches. It's the same data in most cases.

Especially for this years game I would be very dubious with what teams say they can cross and what they can actually cross. I am my teams scouting/strategy mentor and what I try and teach my kids is that a lot of the time the people you meet and talk to in the pit, don't really know the full capability of the robot. Relying on the numbers that you take on the field, and data that you can collect beforehand from videos and previous competitions is where you should put the most stock in terms of accurate information.

[chandrew]

Originally Posted by Boltman

This year it'll be more important to find out what defenses teams can't cross as that won't be posted most likely.

Chances are if a team can cross one of two defenses in a group it can cross the other, with the exception being the cheval de frise/portcullis that requires a separate mechanism from the drive train that differs between the two obstacles. You can see if they have that very quickly by looking at their robot during pit scouting. Pit scouting in general however is often times very useless as it often fails to obtain reliable information, something that will especially be an issue this year (last year in Utah I believe I had 20 teams or so that told me they could do a full autonomous mode/they were capable of it when in reality only one team had that capability). In terms of match scouting what you look at is more of speed/point efficiency during a match and other ways that the robot either succeeds or fails. This includes a large amount of "number of _____ per match" statistics. The most important part of scouting is that you focus on qualities that you would want in an alliance partner/what you want to avoid. A robot's performance in a match is the only way to tell how good it actually is. Even things like videos can be misleading, high shooters this year may appear very accurate in videos, but when a team hits a match they could be unable to align their robot in a way that they could get that shot off.

[ns3517]

A good place to start would what be defenses they can do, as well as how fast it takes them. Also do they have climbing capability? Can they shoot boulders, if so low or high goal? On average how many boulders can they shoot in 1 game? Can they cross all defenses within 1 match? Then of course the usual how well would their bot work with yours.

[IronicDeadBird]

The questions I would ask in scouting are super strategy and super structure abilities, as well as auto plans. Really want to keep it light this year for scouting I find I get a lot more information then I actually need for scouting.
Pit walk/observations would be heavy emphasis on drive base for defense counter picks.
On field it looking for synergies for alliances.

[Kevin Leonard]

Some obvious stuff is like # of high goals scored, # of times they crossed each defense, whether they climbed, etc.

More data is usually good as long as you can manage it all.

[cdogcdl]

Preliminary planning on our app pretty much has this looking like some past years besides scouting crossing the defenses which will be huge. Shots in each goal, missed shots, scoring actions in auto, fouls, endgame, and effectiveness of defense are a good place to start. It's nice to break data down beyond scoring to actual capabilities since we like to play to our alliance's strengths.

And some friendly advice for pit scouting: don't believe what people say, just what you see. If they really want you to believe they can cross 'X' defense, have them show you.

[RAGTOPSTANG]

Our scouting strategy is most likely going to consist of statistics not gained through interviews. As this year's game is so performance driven, it is unlikely that a team would not show its full potential on the field.
-Our scouting criteria will most likely consist of the following-
- Defenses which the robot can cross
- Speed at which the robot can cross said defenses
- Number of balls the robot can score in a match
- Effectiveness at defense
- Type of shooter (linear / arced / catapult / other)
- Effectiveness at climbing
- Which goals the robot an score in

[tindleroot]

The primary purpose of scouting is to learn about the strengths and weaknesses of every other team at the competition. This is easily quantifiable this year (for example, being able to cross a defense is a strength, while not being able to cross it is a weakness). Also take into consideration speed and consistency, as scoring faster or more accurately is a strength. Every data point you take should be an indicator of a strength or weakness compared to other teams. Some obvious ones I see are:

Can they go through defense a, b, c, ...?
Boulder scores vs. misses (and which goal they score in)
Cycle times
How many damaging crossings they have
How many shots they block/defend on the other team
Do they challenge the tower in every match?
Can they scale? How consistently?
etc.

Using the information you collect, you can figure out who you would want to have as an alliance partner. Generally, the optimal alliance would be one where the 3 robots minimize each other's weaknesses and boost each other's strengths. If you are in a captain's position, form an alliance with that team. If you are not in a captain's position, make sure to communicate with teams you're interested in allying with who are in a captain's position. If you can provide reasoning as to why they should pick you, then they might pick you.