scouting 2011

[Kims Robot]

Originally Posted by Siri

For instance, 1511 are scouting gods. Check out their full setup sometime, it'll blow your mind. But they've got a lot of experience in it and can get valuable information out of qualitative answers.

Wow this is humbling... lol. I would call 1114 scouting gods... and several other teams. We certainly put in the effort, but its all still being refined

One note to some of the comments, that is their very initial list. The database very often gets refined as they go. I can't speak to their exact plan this year, but from what I know that was the initial brainstorm. Once it is all put together, there will likely be refinements to make it more useful. It is just a starting point.

Originally Posted by Chris is me

Qualitative data is valuable - but translating opinions into numbers is misleading. Every scout has vastly different ranking systems and opinions for robot effectiveness so taking statistical data based on that is potentially a very bad idea.

Numbers should be objective ...

I've run 1511's Alliance Selection for the last 6 years... I think I posted about this before, but couldn't find it in a couple of quick searches so I will give you an honest rundown.

Yes, qualitative data is subjective, and yes, we change scouters every hour or so, so opinions can be different. But when you have a pile of match data from both practice matches and qualifiers, and 10 of the matches say the driver skill is "poor", and 2 say "good", you tend to lean towards the "poor" assumption.

Also, this data is compiled and given to the coach for each match, obviously the more matches that have been run, the better the data is.

We sit down on Friday night to do alliance selections, and I won't lie... its often a 4-5 hour meeting that doesn't end until 1am. We use the pit scouting data, robot photos, TBA, scores posted online and the output of our match scouting database to develop a ranked list. Its a setup of 5-6 laptops and a bunch of us spread out over a hotel room.

We start with this data and categorize every robot into Yes-high, Yes, Yes-Low, Maybe-high, Maybe, Maybe-Low and No. We then go through the list and rank each robot against all the others in the category, and move robots between categories as needed until we have ~30 robots fully ranked.

For the years that it is applicable, each robot is also given a role, if specific roles are needed... like in 2007, we were focused on picking a scoring robot first, then a ramp bot second. So once a scoring robot was picked, we had to pick from the ramp bots second.

It is a very manual process at the moment, and we have been working on the output of the database in past years to help speed up this process, but for now, it works, and the qualitative data fits.

Then on Saturday, I sit with a couple of other team members in the stands all morning and adjust the list for every single robot that is not a No. If teams were awesome all day Friday but then do terrible in their last couple matches Saturday, they get dropped in selection rank.

1511 is still aspiring to be closer to an 1114, 67, 365, 148, 217, etc etc... but their record does speak pretty well for them, they have been an alliance captain every single year, often multiple times, and sometimes even with some of their "not so good" robots. They spend a lot of time focusing on match and alliance selection strategy, and as their robots get better & better, I expect them to climb the tournament ranks more often.

I'm headed up to visit them this weekend and Im excited to see how they are doing With one of the students now heading up the database (which I think is now in SQL??) they may be able to release it ahead of week 1 for other teams to try.

I think the more teams share their scouting information & setups, the better we can make it and the better strategies you will see on the field Good luck all!