Scouting Aerial Assist After Week 1

[jblay]

For Aerial Assist I have really struggled to get my head around what stats will be useful on a scouting sheet, I have always prided myself on being able to recognize which stats are useful and which are useless, but in this game I am trully lost. Luckily my team doesn't play until week 4 so we still have some time to sort things out.

For all the scout masters out there in the CD world who played in week 1, which stats were most useful? Which ones gave you no benefit? Any non intuitive stat that gave you way more information than you ever could have imagined?

[Joel Glidden]

More than anything else, I am looking for robots that can catch from a human player, robots that don't fumble the ball, and whether they interface with the top of the ball or the sides of the ball when giving/receiving (looking for compatible mechanisms for controlled hand-offs).

[PandaHatMan]

Our team attended CIR this past week and we have determined an important factor is being able to get a robot where it needs to be. For us, that meant we were looking at robots with beefy drivetrains.

Another factor to consider is ball control. We ran into a few robots at CIR who lost the ball after a bump or shove. It's important to keep the ball inside the robots as much as possible.

Another nice thing to look for is easy passing and/or truss shots. Hope this helps.

[XaulZan11]

We are dedicating our best scouts to qualitative information, but still will be recording some statistics. We tested our data collection system out at Central Illinois. You can find the results here (http://www.chiefdelphi.com/forums/showpost.php?p=1351143&postcount=61). With just two scouts and a quick 20 minute discussion, we were able to make a pretty good pick list which predicted most of the selections.

I think the best predictors of overal robot ability were 'average auto points', 'possesions per match', '10 pointers made per match' and '10 pt shooting %'. There really isn't one great statistic that can quickly and easily rank teams (in the past you could do total points scored, but that doesn't really work for this game).

For Wisconsin, we will likely add 'possessions lost per match' (for teams that drop the ball alot), 'human player loads per match', and fouls.

[Anupam Goli]

At one point in this season, we may open our scouting system to the public, but for now, we're making our own modifications to it before we do so. Our group is tracking assists made, assists received, truss shots, catches, low goal scores, and high goal scores. We use tableau to sort through this data when generating a pick list to find alliance partners that will match the strategies in our playbook and should theoretically enable us to score the most points we possibly can.

[jblay]

We are dedicating our best scouts to qualitative information, but still will be recording some statistics. We tested our data collection system out at Central Illinois. You can find the results here (http://www.chiefdelphi.com/forums/showpost.php?p=1351143&postcount=61). With just two scouts and a quick 20 minute discussion, we were able to make a pretty good pick list which predicted most of the selections.

I think the best predictors of overal robot ability were 'average auto points', 'possesions per match', '10 pointers made per match' and '10 pt shooting %'. There really isn't one great statistic that can quickly and easily rank teams (in the past you could do total points scored, but that doesn't really work for this game).

For Wisconsin, we will likely add 'possessions lost per match' (for teams that drop the ball alot), 'human player loads per match', and fouls.

This was incredibly useful. I've been thinking over the benefits of including shooting percentage and your numbers seem to justify it.

[D.Allred]

Joe,
The link below shows my initial thoughts going into Palmetto. I'm definitely adjusting before Orlando.

http://www.chiefdelphi.com/forums/showthread.php?t=126617

This game requires more qualitative statistics than most games. SPAM's "super scout" observation method will be critical since more teams can play a role of ball possession. The question is how well do they accomplish the task.

Key stats to keep:
Autonomous mobility plus high goal / low goal accuracy
What roles did they play in a cycle - inbounder, truss, catch, pass, floor possess, high goal, low goal.
Number of defensive plays.
Shooting accuracy high / low.

I'm dropping pre-cycle shots and cycle counts from my original sheet. Autonomous and tele-op goal accuracy is plenty to track. Cycle counts were too cumbersome.

I'm thinking about adding truss shot accuracy. Seemed odd at first, but there were many deflected or completely missed truss shots - huge time waster.

Ball security is also key. It could be difficult to quantify "unintentionally" dropped passes. I believe this should be part of the super scout job.

David

[CalTran]

Until I see a Team Update addressing this, and I hate to point it out, but one thing I saw that unfortunately needs to be scouted is how well a team's HP handles the ball, and whether they incur G40's or not.

[z4t143]

We are dedicating our best scouts to qualitative information, but still will be recording some statistics. We tested our data collection system out at Central Illinois. You can find the results here (http://www.chiefdelphi.com/forums/showpost.php?p=1351143&postcount=61). With just two scouts and a quick 20 minute discussion, we were able to make a pretty good pick list which predicted most of the selections.

I think the best predictors of overal robot ability were 'average auto points', 'possesions per match', '10 pointers made per match' and '10 pt shooting %'. There really isn't one great statistic that can quickly and easily rank teams (in the past you could do total points scored, but that doesn't really work for this game).

For Wisconsin, we will likely add 'possessions lost per match' (for teams that drop the ball alot), 'human player loads per match', and fouls.

Would you be willing to share your program? We were at Central Illinois and used your data after Day 1. We'd like to replicate your process for the STL regional.

[XaulZan11]

Would you be willing to share your program? We were at Central Illinois and used your data after Day 1. We'd like to replicate your process for the STL regional.

Yes. I had to confirm with our strategy team, but we'd be willing to post our database and scouting sheets. We're making some adjustments to it (any suggestions welcome) and will create some instructions, but should post it by Tuesday of next week.

[XaulZan11]

Would you be willing to share your program? We were at Central Illinois and used your data after Day 1. We'd like to replicate your process for the STL regional.

Everything you need can be found here (http://www.chiefdelphi.com/media/papers/2970?).