Requesting Tips about General Scouting Strategy

[Danny Diaz]

Hey there, CD Scouters!

I've been doing scouting with FRC418 for about a decade, attempting to find a system of scouting that works for our team at the venues we go to. That's a lot to ask, since our team and its needs - as well as the venues we go to - are ever changing. I was hoping to hear how YOU GUYS do scouting.

Now, I don't want to know the mundane specifics about scouting calculations, what you're looking for (robot characteristics, average contribution, etc...) because, well, I don't care. I mean, it's not that I don't care - okay, I'll be honest, I really don't - I only care about the PROCESS that you use when scouting. Let me describe to you the process history and you'll see what I mean.

Back in the day, when we first started to formalize our scouting system, we were a large team (40+ members). We devoted one team member per team to watch the teams on the field each match, and kept detailed stats on what the robots did. On top of that, we would take pictures of robots in the pits so we knew what the robots looked like. That was very manpower intensive during the Friday scouting day, and even more intensive during the evening when we attempted to collate the information so we could have a semi-reasonable scouting "meeting" to determine our 23-team pick list. We were dumb, too, and allowed scouters to take non-empirical (e.g. subjective) notes and so nobody really could compare apples to apples and meetings lasted well into the night with a lot of frustration in trying to compare teams.

We then moved to keeping data on computers, and one year even set up a wired network in the stands - this was back in the day when Lone Star had power you could tap into in the stands if you knew how to get to it. We were able to take lots more information - and empirical information at that - but it was still too manually intensive to process (our lead scouter actually started crying because he was too stressed out trying to compile the information as fast as it was being generated and he couldn't come close to keeping up). Pictures would come in from the pit scouters and the pics had to be reviewed and often times retaken (because pit pictures are the WORSE) and then the pictures had to be collated and renamed to match the team number. In our meeting in the evening, we still had a fair amount of data collation to do (took almost 2 hours) but the data was better - we still argued a lot about "okay then, how did they drive?", "Did they do that quickly or slowly?", and so on. And the 23-team pick list would run well into the night trying to determine which teams we liked.

Recently we've moved away from taking almost any data at all - instead we've been recording the matches on video. But not just in any way, we wrote a LabVIEW app using the vision tools that FRC now has access to that can control just about any USB webcam or video camera with a 1394 interface. What we do is record each match to a file - but we don't record the entire match, we actually record still images at about 10-15 frames per second and let LabVIEW's IMAQ tools compile that into an AVI movie for us. This makes it so that the files are smaller, but when we're watching videos we don't have to spend 2:30 watching the entire match, we spend about 1 minute watching each match (we play it back at the usual 40 frames per second, so the matches are in somewhat-fast-forward). So in our 2 hour meeting in the evening, we can actually WATCH every team's match and see how they really did. And when we take pictures, we now ONLY take pictures of robots on the field - and not in the pits. On the field you can see what a robot REALLY Looks like WITH BUMPERS. The big element is that now we only really need a scouting crew of about 3 people total, and it takes us about 20 minutes in the evening to get set up before our 23-team pick scouting meeting.

However, this requires us to have POWER in the stands - we're running a laptop and a camera (or if using a USB webcam just a laptop). Last year Lone Star did not have power in the stands, so we resorted to using FRC batteries with inverters to help us power the laptop and camcorder, which one battery only lasted about 5 hours. So we reduced our video recording time to a window spanning from when matches started back up after lunch until the last match of the day. That worked for us, but we would have preferred to have more recording/editing time.

OK - so now you know my story. What is yours? How do you do scouting?

-Danny

[Conor Ryan]

Wow Danny,
I love that video idea. Should be very high quality.

Ok for the past few years now, my Dad and I have been writing a largely data entry based system. Our goal is to use our scouting data to predict what the score of matches will be before the matches are played. By the end of a regular event, ~100 Qualification matches, we can rank teams our own way, and figure out who really should be ranked where. Additionally we capture strategy related information for who starts where and how they play certain aspects of the game.

Its fairly labor intensive, we have 6 people watching 6 robots at all times. It gets more fun because you get to see how accurate we are in guessing the scores. Additionally, we do data pulls between each match we play in, so we generally know what the other alliance is capable of all the time.

My Eventual Goal: Use video scouting, with some video analysis to get the similar to what this small israeli company did before stats.com bought them (for a large sum) http://www.stats.com/sportvu/basketball.asp
http://www.stats.com/sportvu/basketball_technology.asp

[EricLeifermann]

Since my involvement in FIRST I've used paper scouting, laptop excel "app" scouting, Wiimote scouting and android tablet based scouting.

On Wave we started out with paper then moved to Wiimote scouting and for the last 2 years we have don't android tablet based scouting. Each year there is always 1 scouter per robot per match collecting data. Then on Friday night for pick list making we run through the data collected and look at pictures that were taken of each robot(we have a "strict" rule on what the pictures need to capture but we haven't always achieved it, maybe this year).

This year we are still in the works about what method we will use to collect quantitative match data, but we will have a minimum of 4 extra scouters each match collecting what we call qualitative data on each match, to aid in making a pick list and match strategy. We feel that this game is so different that each match a robot could be doing a different thing so the quantitative data (goals scored, assists, truss tosses, etc) will be misleading if you go with averages.

Ultimately it comes down to choosing the method you feel will give you the best data to aid in making the best pick list you can. Also having it be fun and enjoyable so your team actually wants to scout is always helpful as well.