scouting

[BX MARK]

Originally Posted by musicman

Hi every one I am wondering if any one has a scouting sheet already made and would like to share it with team 2531 so we can get some ideas from it. So if you have one made and ready for some one to look it over I can send you a response and tell you if it was good or not.
Thanks Collin Leck

Try this...our team is most likely using it. http://www.chiefdelphi.com/forums/sh...uminum+falcons

[Chris is me]

http://team2791.wordpress.com/2011/02/15/scouting-2011/

[boomergeek]

Who's collecting stats for defensive centric robots?

If so, what objective metrics do you use?

# seconds successfully impeding best bot on other alliance?
# seconds successfully impeding second best bot on other alliance?
# seconds impeding more than one robot at a time?

[Chris is me]

Originally Posted by boomergeek

Who's collecting stats for defensive centric robots?

If so, what objective metrics do you use?

Defense is not something that can be quantified well at all. On the off chance you scout for defense, I strongly recommend against using statistics for defense.

[LoganAdams]

Originally Posted by Chris is me

Defense is not something that can be quantified well at all. On the off chance you scout for defense, I strongly recommend against using statistics for defense.

I could not agree more. That said, I think that really good defense will be hard to find, and therefore stand out.

[boomergeek]

"Defense is not something that can be quantified well at all. "

That is not what the pro scouts in the NBA think...

Defenders that can shut down a good offensive player like Kobe are scouted and play a role all the time.

Pro scouts in the NBA, and the NHL are able to come up with useable metrics for defense: why wouldn't the brain power of FRC want to do the equivalent?

How could a good scouting/picking team decide if defensive pick is better than another offensive pick if the scouting team has not come up with tangible metrics for defense centric bots?

If the only tangible collection is of offensive metrics, then the data would naturally bias those using it to pick three offensive teams instead of 2 offensive and one defensive.

I've only been around FRC for three seasons, but I've looked at videos from previous years as well: consideration of defense seems to be generally an afterthought by most teams until it's shown to be effective.

When defense is effective, there are tangible metrics to describe the effect. The question is, will scouts look for them?

[Kpchem]

Originally Posted by boomergeek

How could a good scouting/picking team decide if defensive pick is better than another offensive pick if the scouting team has not come up with tangible metrics for defense centric bots?

By my experience, the question isn't whether to pick another offensive robot or a defensive robot based on the metrics you find. That is answered through your own unique strategy to the game. Some teams will decide three offensive robots is the way to play this game, others will decide that it's two offensive and defense, and I'm sure there will be other combinations based on how many feeder bots there are at that regional

However, I am also curious as to how people are quantifying defense. In general we like to use simple ratings of 1-5 for defensive ability, but these can often be prone to bias as it is a purely subjective rating, even when it is limited to one or two people making that decision.

[boomergeek]

"By my experience, the question isn't whether to pick another offensive robot or a defensive robot based on the metrics you find. That is answered through your own unique strategy to the game."

I'm really curious on this one.
Say for example, you are the second team picking and the first team already has the two best offensive robots. For your second and last pick, do you choose an offensive robot that is significantly less effective at offense than you and your first pick or do you pick a defensive bot that is shown as effective at shutting out one of the two best bots on the first team (let's assume everyone's minibot is equivalent).
If you haven't collected tangible defensive based metrics, how would you know whether your unique strategy should be changed based on the available picks?

[Chris is me]

Good defensive bots stand out enough to be noticeable subjectively.

Most other defensive bots average out to be about the same. Pick teams based on lack of penalties at that point, IMO.

[Kpchem]

When I talk about my own "unique strategy" to the game, I mean what kinds of robots work well with my team and what I have observed works well at the competition. If it looks like 3 teams all trying to run and score tubes gets clogged up and messy and is not an advantage, I obviously don't want to do that. If defense looks like it will help more than a third scorer, I'll probably go with that. It's a very fluid strategy now, but by the time the end of the weekend roles around, between my observations and conversations with our drive team, it becomes apparent what combinations of robots will work best for us.

I'm not saying I don't collect any defensive data at all, but that the data cannot necessarily be compared to the offensive data because I seem them as two entirely separate aspects of the game. So we scout both sides of things, offensive and defensive, and then use both of those data sets together with our strategy, to pick the alliance we believe will work best.

[boomergeek]

"Good defensive bots stand out enough to be noticeable subjectively. "

While attempting to play defense they shutout, or almost shutout, awesome offensive robots.
If they don't do that, they aren't good defensive robots.


Ability to play defensive against average bots is not typically important.

In 2009, at BAE, we decided to play defense based on the realistic comparison of our offensive capability against the top 20% of robots we saw in the pits.
We focused on playing defense against the best robot in the opposing alliance. That strategy and a lot of luck got us a third place finish in the prelims.

My point is that effectiveness of defensive robots can usually be tangibly measured: if scouts know what to look for.

[MagiChau]

Scouting itself is a subjective thing. You can choose how much you trust the numbers and or opinions. In the end the human representative out there on the field makes the final call.

[Carter12s]

If you had enough people it seems that you could count the number of hits the defense robot scores on a offensive robot. You could count forced fumbles, you could count the amount of time that a defensive robot delays another robot. All of these metrics require close observation thou, and even if tabulated often will not be able to provide a complete view of a defensive robots capabilities, not to mention the difficulties in objectively collecting the information.

My advise would be to calculate the EOA (Estimated Offensive Abilities) of a robot based off of the number of tubes it scored each match, and then compare this to the number it scored when playing against different dedicated defensive robots. This would let you see if there can be a EDA (Estimated Defensive Abilities) calculated for defensive robots based off of how other robots predicted offensive scores compare to actual scores. This would of course require accurate records of every match for every robot including breakdowns and signal loses which would ruin the data.

This all seems like a lot of work considering the advantages it brings. It would seem unlikely that a team would pick a defensive robot for their first pick (unless it had extraordinarily defense) and in general by the time the second pick comes around the remaining robots are fairly easy to rank. Specifically at my regional (Utah) where only 29 teams are registered, this would be a waste of time since for the second pick there are so few teams remaining that the choices are usually obvious.

[Wizzard of Izz]

Have you thought of cowscout.com? It's new this year, but it's nice. 2993's using it in correlation with Myscout (This is a just in case), developed by team 1094. http://www.chiefdelphi.com/forums/sh...hlight=myscout
All you need is a laptop for Myscout, and if you use Cowscout, you just need a LAN cable provided you have access to an outlet. Oh, and if you choose to use a scouting program, bring duct tape for your cords.