In match Scouting ideation

[StewyJ]

As our robot begins to take shape, I was wondering if anyone's thoughts had turned to what quantitative measures you will be focusing on for your in-match scouting. It seems, with only 1 game piece per alliance, that in-match scouting should be easier this year?!?

Of course, from the scoring we will have alliance wins, alliance assist points, and alliance auto points.

We had a little ideation and came up with the following. Ideas on additional measures and which measures to be eliminated to make scouting easier for a small team would be appreciated. We had a little ideation can came up with the following (some of which are duplicates said another way)

1) Autonomous performance per robot - high goal attempt, low goal attempt, no ball, mobility attempt, positioning location, goalie position, individual robot points achieved, "smart" attempt at hot goal (only could be measured if hot goal is on other side for first 5 seconds), number of seconds to clear ball in auto, number of seconds for alliance to clear all auto balls and start first cycle.
2) Cycles - seconds per cycle, number of cycles attempted per match,
3) Cycle Play - total alliance points per cycle, assist points by robot, truss points by robot, catches by robot, high goal points by robot, low goal points by robot, successful receptions (does robot control ball and could divide this by number of cycles attempted to get a ratio of how often robot touches the ball)
4) Foul points (at 20 and 50, foul points could be very significant this year) - foul points assessed
5) Zone played in (do you think robots will stay in specific zones or will they move)
6) Is robot a teamwork robot or try to control match themselves (qualitative)
7) How well does team assist (qualitative)

Thanks for your help and any ideas

[mbshark]

Also think about defense - mostly qualitative
1) Does the robot go in the goalie zone
2) How many shots does the robot block out of how many did it try to block
3) Does it go straight to defense after the ball leaves their robot if starting end or white zone and is the robot in the end zone blocking the others when not trying to get the ball or in possession of the ball

Also Yellow/Red Cards if you want to be really detailed
Put in an area for a Dead Robot or No Show

[StewyJ]

Originally Posted by mbshark

Also think about defense - mostly qualitative
1) Does the robot go in the goalie zone
2) How many shots does the robot block out of how many did it try to block
3) Does it go straight to defense after the ball leaves their robot if starting end or white zone and is the robot in the end zone blocking the others when not trying to get the ball or in possession of the ball

Also Yellow/Red Cards if you want to be really detailed
Put in an area for a Dead Robot or No Show

Great thoughts on goalie play. I will definitely add a checkbox for dead robot.

[Dragonking]

In some ways match scouting is easier this year and in other ways scouting is harder. You really only need 2-4 scouts per match since there is only one game piece at a time (auto shouldn't be to hard to keep track of).
What makes it more difficult is that you have to focus on a lot of qualitative data and it can be difficult to tell what is happening on the field (ex. which robots have assists).

[WeJohnFriedIt]

Our team usually comes up with a sheet to fill out for one robot at a time during matches. This is what I have so far:

http://imgur.com/Ir4FToG

Let me know what you think!

[PVCpirate]

I honestly don't think quantitative data on Teleop scoring will be useful, as the numbers will be very dependent on alliance partners. If I was scouting this year, I would focus on the abilities each robot has(shooting, passing, driving, defense, etc.) and qualitatively how useful these abilities would be to an alliance.

[depth_Finder]

So far, most of our scouting this year is looking like trying to find efficiency. We are going to try to find the amount of time it takes for a bot to do its actions and then do a lot of math to it to arrive at a few metrics we can use.

However, I believe this is a year were OPR is going to be a much more accurate metric than in previous years. Because there is only one game piece for most of the game, many robot's contributions will be hard to accurately fit into concrete quantitative fields like in the last two years.

A good scouting system should be able to supplement the OPR calculations. However, we haven't figured out how to do that yet.

[Joseph1825]

Originally Posted by depth_Finder

So far, most of our scouting this year is looking like trying to find efficiency. We are going to try to find the amount of time it takes for a bot to do its actions and then do a lot of math to it to arrive at a few metrics we can use.

However, I believe this is a year were OPR is going to be a much more accurate metric than in previous years. Because there is only one game piece for most of the game, many robot's contributions will be hard to accurately fit into concrete quantitative fields like in the last two years.

A good scouting system should be able to supplement the OPR calculations. However, we haven't figured out how to do that yet.

I was actually thinking that OPR will not be good this year. When you think of last year, each team contributed roughly the same amount of points each round regardless of who their partners were. This year you won't contribute the same amount of points each round, I think that average match score will be just as accurate as OPR this year. I could be completely wrong though.

[Strategist257]

Originally Posted by WeJohnFriedIt

Our team usually comes up with a sheet to fill out for one robot at a time during matches. This is what I have so far:

http://imgur.com/Ir4FToG

Let me know what you think!

i think that this is quite good. most years my team has lots of trouble scouting, and actually getting the important information to the drivers.

[pandamonium]

Efficiency is key. I am contemplating timing each robots posetion.

[KrazyCarl92]

Originally Posted by pandamonium

Efficiency is key. I am contemplating timing each robots posetion.

This is what I am thinking. 6 people, each watching a robot, each with stop watches. Track how long a robot holds the ball, how they received it, number of zones occupied in possession, and the result of ejecting the ball (high goal fail/success, low goal fail/success, truss, to HP, roll on floor, mishandle). In general, less time is better. However, a robot with the ball for 10 seconds that receives a pass from a partner on the floor and rolls it on the floor to another partner while occupying 1 zone total is not apples to apples with a robot that spends 10 seconds with the ball by receiving it from the floor, occupies 2 zones in possession, then hits the high goal 95% of its attempts. The time is the same, but it's a different set of actions. This is where combining data in a digestible format and information analysis becomes incredibly important.

[Mr. Tatorscout]

Originally Posted by KrazyCarl92

This is what I am thinking. 6 people, each watching a robot, each with stop watches. Track how long a robot holds the ball, how they received it, number of zones occupied in possession,

Your time idea is interesting, although I can see issues with trying to keep track of too many things (physically as well as datawise). But, i'm a klutz, so it might be fine. Is it necessary, though? Can we get decent data about ball handling proficiency and efficiency by counting passes and receptions? If the numbers are high over a few matches, wouldn't that mean they were pretty quick?

Our team is struggling with scouting assists because it's so dependent on partner competency. If the partner bots haven't had much practice time and are struggling, the amazing scouted bot would have a low assist score and artificially deflate their desirability. But if all three bots are average, their individual assist scores would be artificially high. Or if the struggling scouted bot had great partners, it's assist score would be inflated. So how would counting assists be different than looking at the overall alliance score?

Finally, is anyone considering scouting human players? I can see their effectiveness at passing intelligently playing a roll in game play.

[Mr. Tatorscout]

Another quick question, how do you folks collect data on penalties? From what we've done in the past, if an inexperienced team member (or parent)is watching a single robot, he or she doesn't notice the flag waving judge. Is there a log kept somewhere of which team got which penalties that we can access?

[Iaquinto.Joe]

Scouting this year is very important because this game is based upon robot interdependency. Understanding how a team is going to fit into your strategy is key.

[Zmarken]

This year, I think that measuring timing is going to be important.

Getting a high number of cycles is going to be important to get the most points, so you need to find out things such as how long it takes a robot to pass/receive, how long they take to position, and stuff of that sort.

If I had to make a scouting sheet right now, I'd have to go with...

Autonomous:

1) Position: Goalie zone/White Zone, as well as Left/Right/Center

(If in the White Zone)

2) Scoring: pretty basic, how many shots taken, how many points gained

3) HOT Tracking: can the robot track the HOT Goal?

4) Time to Shoot: How long the robot takes to shoot. Right away? The full 10 seconds?

5) Intake: Can the robot pick up balls if another alliance member is in the Goalie Zone?

6)

7) Mobility: Can the robot get the mobility bonus?

(If in the Goalie Zone)

8) Blocks: How many shots did it block?

9) Pathing: Does the robot just go back and forth, or does the robot track the other robots/balls?

Teleop

1) Shots on Goal: How many shots the robot had on the goal, and how many got in

2) Passes: How many passes did the robot do successfully?

3) Receptions: How many balls did the robot receive from its teammates?

4) (Combine with #3?) Catches: How many catches, either over the truss, or in general, did the robot have?

5) Zone: Is the robot a roamer, or does it stay in one zone? Also, which zone?

6) Cycles: How many cycles were completed? Also, how long did each cycle take? Number of assists per cycle?

7) Assists: How many assists did the robot have?

8) Time to Throw: How long did it take from receiving the ball, to pass it to another teammate? Also, were they waiting on the receiving robot?

9) Time to Receive: How long did it take the robot to get ready to catch? Also, were they waiting on the passing robot?

10) Time to Shoot: Same as in Autonomous, how long does it take to aim and shoot the ball?

11) Truss Throw: How many passes over the truss did they do?

12) Truss Catch (Combine with 3?) How many catches over the truss did the robot have?

(Now onto defense)

13) Time to Defense: How long does it take for the robot to switch to defense after passing the ball?

14) Passes defensed: Total passes blocked, deflected, etc...

15) Goalie Zone: Does the robot use the Goalie Zone, and if so, how do they block in it?

16) Disruption: How much does the robot disrupt cycle time of the other alliance? How much time can they drain by defending?

I tried to make it less subjective, and easier to just fill in without paying too much attention (I know the scouts on my team...) The hardest thing about this game, is that it's more about teamwork, so if you just go for points scored, you'll really only get good results from the one robot in the alliances scoring zone.

Again, I think that the time it takes to do a cycle is one of the more important things to check, especially breaking it down to how fast can you pass/get ready to receive/shoot.

Assists are also important, because even if you've got a slow cycle, if you're doing 3 assists and your opponent's only doing 2, they need to double your cycles to win.

Also, defense is probably one of the most important things in this game, because if your defense slows down your opponent's cycle time, that puts you at an advantage.

That's it for me so far.