Understaffed scouting

[viggy96]

Google spreadsheets are always the best.

[TogetherSword8]

What we have done in past years is a lot of gut feelings. When watching a match, what you see, whether in the stands or in queue van make a big difference in who you would choose. Trust your gut, and hope it the 10%of the time it is wrong isn't then.

[Sam_Mills]

Originally Posted by TogetherSword8

What we have done in past years is a lot of gut feelings. When watching a match, what you see, whether in the stands or in queue van make a big difference in who you would choose. Trust your gut, and hope it the 10%of the time it is wrong isn't then.

I would highly recommend any method of scouting that is not this. Championships are lost and won on scouting, and quantitative data is extremely important. (4334 for example would not have been a "gut" pick in 2012, but its a choice that got The Eh Team a division win)

That being said, I support the group scouting method where you work with other teams. Its a little harder to guarantee quality data, but its far superior to no data at all.

[artK]

Inter-team scouting is a good way of doing it, if both teams arrange it beforehand. (My one caution would be to have people from the other team scout your robot to reduce bias).

If you have <=6 people: Have two of them scout each alliance as a whole (swap them in and out as need), and someone compile the data between matches. Then have the remainder scout individual robots determined at the start of the match for specific qualities.

In the early matches, pick teams that are likely to be captains or first picks so you can sort out the top. As matches progress, start going with teams who could make second picks, but still check the top teams.

Create a sketch of a list at lunch, see if you can compare teams close to each other in close matches (if team A, B, and C are all next to each other in the list and all play in four matches, have one guy watch them during their respective matches, then do a sub sort).

A number of teams already do the above, in addition to the 6+data compiler scouting teams. The only thing if you do this is that the 2-5 people should be the same 2-5 people all day, not 2-5 who switch off with pit crew*. Like designing robots, it's better to have a few experienced scouts than a lot of okay scouts. If you have the same few scouts running the team the whole season, you will all get better at it by the time champs rolls around.

*If pit crew has input about how certain teams work with your team or suspect failure points in other machines, take it into consideration.

[XaulZan11]

Originally Posted by Sam_Mills

I would highly recommend any method of scouting that is not this. Championships are lost and won on scouting, and quantitative data is extremely important. (4334 for example would not have been a "gut" pick in 2012, but its a choice that got The Eh Team a division win)

What quantitative data would lead 1114/2056 to pick 4334? I suspect they were picked because they were very thin (to allow for easier triple balance), well driven and dependable. At least in my opinion, those are more qualitative measures opposed to quantitive. I'm not saying quantititative data isn't important, but how qualitative data is also necessary.

[Sam_Mills]

Originally Posted by XaulZan11

What quantitative data would lead 1114/2056 to pick 4334? I suspect they were picked because they were very thin (to allow for easier triple balance), well driven and dependable. At least in my opinion, those are more qualitative measures opposed to quantitive. I'm not saying quantititative data isn't important, but how qualitative data is also necessary.

Qualitative scouting is undoubtedly important, as is pit scouting, but especially at larger events where there are close decisions to be made, numbers are going to be the best answer.

Regarding the 1114/2056/4334 alliance (warning: *some* blind conjecture ahead) yes, they were well driven, dependable and thin, but there are elements to consider that can't be answered without numbers. 1114 and 2056 would be taking a gamble by having only 2 robots that could shoot when facing alliances of 3 if they didn't do their homework. With good scouting data, you can look at how much higher robots scored compared to their averages when they had 4334 feeding them. If shooters score an average of 2 more cycles when they have 4334, then it would be worth choosing 4334 over any robot that scores an average of 2 cycles or fewer. By self scouting, they can also put together hypothetical alliances and estimate not only their expected match score, but also the standard deviation of scores of different alliances (high standard deviation = higher risk/higher score ceiling). Lacking this information means you can't make decisions about things like your ceiling vs your average, or what your opponents are likely to be scoring. This is a very condensed version, but there is a lot more that can be gleaned from good scouting data than what I said here.

I have never won a world championship though, so if anything I put up is wrong and can be corrected by someone who has, I welcome it.

[Siri]

I think there's a lot of conflation of ideas going on here. What you do about scouting when you're resource-limited depends almost entirely on what you're trying to achieve. (Much like everything in life.) Are you legitimately aiming to be an alliance captain and make a pick list that can reasonably win you this competition? Then your best and only reliable solution is probably to not be understaffed (or under-trained, for that matter). Realize that this is a two step scouting issue: become an alliance captain and make your pick list.

But scouting that way when it's not your main/achievable goal that weekend won't necessarily help you. If your goal is to play well and get picked, 'gut' scouting isn't so bad. A few well-trained scouts/strategists can give you a lot of insight into your allies and opponents without actually tallying game pieces. In fact, when your team is understaffed but not under-trained in this way, qualitative can be better than quantitative. I've always opted for good qualitative over bad quantitative, and it hasn't let me down yet.

On the third hand, if you're aiming for either of these two and/or to get deeper into the FRC community, joint-team scouting can be great. Just understand it has its pitfalls.

On the fourth hand, if you goal is just to get better at the game (including scouting), you probably want a mix of qualitative and quantitative for your own team. I and most of the coaches I play with will keep at least one top scout on 'gut' duty--usually more than one--whenever we're fully staffed. That means at some point those guys need gut scout training. On the other hand, good quantitative scouting also requires practice.

Much like everything in this business, your scouting strategy needs match your competition strategy. Be honest with yourself - don't discount a strategy that (probably) won't make you Championship Alliance Captain when that's not your team's goal at the given moment.

[Kevin Leonard]

Originally Posted by Siri

I think there's a lot of conflation of ideas going on here. What you do about scouting when you're resource-limited depends almost entirely on what you're trying to achieve. (Much like everything in life.) Are you legitimately aiming to be an alliance captain and make a pick list that can reasonably win you this competition? Then your best and only reliable solution is probably to not be understaffed (or under-trained, for that matter). Realize that this is a two step scouting issue: become an alliance captain and make your pick list.

But scouting that way when it's not your main/achievable goal that weekend won't necessarily help you. If your goal is to play well and get picked, 'gut' scouting isn't so bad. A few well-trained scouts/strategists can give you a lot of insight into your allies and opponents without actually tallying game pieces. In fact, when your team is understaffed but not under-trained in this way, qualitative can be better than quantitative. I've always opted for good qualitative over bad quantitative, and it hasn't let me down yet.

On the third hand, if you're aiming for either of these two and/or to get deeper into the FRC community, joint-team scouting can be great. Just understand it has its pitfalls.

On the fourth hand, if you goal is just to get better at the game (including scouting), you probably want a mix of qualitative and quantitative for your own team. I and most of the coaches I play with will keep at least one top scout on 'gut' duty--usually more than one--whenever we're fully staffed. That means at some point those guys need gut scout training. On the other hand, good quantitative scouting also requires practice.

Much like everything in this business, your scouting strategy needs match your competition strategy. Be honest with yourself - don't discount a strategy that (probably) won't make you Championship Alliance Captain when that's not your team's goal at the given moment.

I have to disagree with a few points here.
The problem with scouting under the assumption you won't be an alliance captain is that sometimes you still end up an alliance captain.
Every team should have a picklist going into Saturday morning, because a) they could end up an alliance captain, and b) they could get picked by a team with no pick list.

In my opinion and experience, quantitative scouting is almost always better than qualitative. That being said, if you're unable to put together the people in your team or through multiple teams to have a quantitative scouting group, qualitative is better than nothing (and is often a good supplement to hard data anyway.)

[jaykris284]

1261 is at Palmetto (a 1st week regional) but is also sending 3 or 4 people to Perry (also a 1st week regional) who will be scouting in collaboration with team 2974 Walton Robotics. We will share data from Palmetto and they will share data from Perry, it is a win-win for both of us.

[Citrus Dad]

Originally Posted by josephus

We have a relatively small team but would like to have an effective scouting group. How might we scout matches effectively with fewer than six people? I'm not worried about pit scouting, as that can be done by one or two people.
--josephus

First, focus on WHY you are scouting? Are you expecting to be an alliance captain or a first pick? Do you want to use your scout data for alliance strategy or just for draft picks? Make an honest assessment of your team's likely success and position in the competitions. Being an alliance captain demands much more than being a draft pick, but you can still bring substantial value if you have a good means of identifying a 2nd alliance member that fits well with your robot. For alliance strategy focus on what your robot does and what are the other skillsets that other robots need to have to be successful.

Once you figured that out, then focus your scouting on the information you need. Try to keep it from being overwhelming for your scouts.

And as EricH said, pull in mentors and parents. We've done that in the past, even having junior mentors testing out specialized scouting skills, and I've sat in for students when they need a break. Really make it a team effort.

[New Lightning]

First to cover a few things on this thread. Using Google spreadsheets while not at an event could be a useful tool; however, since most events will not allow you to run your own WIFI, using Google spreadsheets is not a viable option while in the stands. Secondly for the person that asked, yes Scouting apps, many teams develop their own or use ones published by other teams. There is a whole thread on this topic Scouting the Scouting apps.

Secondly what has been mostly ignored is that in order to get the best data possible there needs to be a mix of both qualitative and quantitative data. Having one or the other is a very limiting factor and you may miss a key aspect of a certain robot that makes it either a great choice, or a horrible choice. That being said if your understaffed and you have to pick one over the other go with the quantitative. Quantitative data is unbiased and irrefutable. What did a team actually did, not what can the robot potentially do, not what the team said there robot can do, quantitative data tells you exactly how the robot performed. While qualitative data/ gut scouting is by no means bad there will always be some elements of human bias in this thinking and that can lead to problems when making a pit list and selecting people on the filed.

[BrendanB]

Great points in here mainly figure out what/why you want to scout. Aside from collecting numbers what I find more important is being able to look at a field of robots and determine what type of robots we want on our alliance and then rank the robots we see on the field that would best fit those roles.

Prioritize robots that have strengths you don't have. Pick robots who will work well with you. If all three of you need to go to the landfill to get totes you'll all be in each other's way. The same goes for the HP station.

The same can be said for autonomous: pick robots that have routines that work well with yours to achieve the highest score.

Keep an eye on the field for consistent performers who can perform tasks reliably from match to match. Keep an eye on teams progressing as the weekend goes on. Some teams increase their performance near the end of their rounds and can be huge dark horses in alliance selections. Find out why teams are under/over performing by talking in the pits.

Don't just look at the top robots on the field the more important partners are the ones who will be around for the last 8 picks and can often be the most important. Last year at early events it was hard to get a robot in the third round who could quickly and consistently gain possession for a three assist cycle or consistently score a ball in autonomous but a few were out there. Most of the time teams won't be in a position to pick the one of the top 2-4 robots at an event so don't dwell too much on them.

Having the numbers on who scores more totes/rcs in auto and teleop is very important but don't just focus on getting numbers because you need to examine each robot through a variety of lenses that numbers might not show. In previous years teams who had a low average score could have been slaughtered with defense in their matches while your higher average teams never had defenders on them. Defense isn't applicable this year since alliances are separate but whenever a scout tells me "This team is really good they scored xxx on their own" I want to know if they were defended.

Scouting with small groups isn't easy but its doable. You can throw 20 people at scouting but if they don't understand how to analyze a field of robots all the data in the world can't help you.

[Caleb Sykes]

Originally Posted by BrendanB

Having the numbers on who scores more totes/rcs in auto and teleop is very important but don't just focus on getting numbers because you need to examine each robot through a variety of lenses that numbers might not show.

I agree that the perfect combination for scouting is a mixture of quantitative and qualitative data. However, few teams actually have a good enough scouting system to collect meaningful quantitative data, much less qualitative data. I've reviewed data from teams (including my own) after events, and at least half of the data sets I have seen are absolute garbage. I believe that almost all teams should focus first on improving their quantitative scouting before even thinking about qualitative scouting.

Basically, if your scouts can't reliably count the number of totes a team can stack in a match, why would you trust that their "qualitative" assessments of teams mean anything at all?

[TDav540]

With a small team, it may be smart to avoid collecting a large number of repeats. Have a list of the teams you would like to collect during which matches to maximize the number of teams covered. Yes, you will miss some matches, but it's better to collect some data on every team than skip some teams that could have a vitally important mechanism.

Additionally, take pictures during your pit scouting operation. Pictures will help your scouters/drivers with recalling the specific bot, so having those available will help tremendously.

Finally, as many others have mentioned, work with other teams in the area to scout effectively. Almost every team has some sort of scouting operation, and many rookie/small teams will be understaffed. Combining resources will be like coopertition, beneficial for all simultaneously.

[BrendanB]

Originally Posted by Caleb Sykes

I agree that the perfect combination for scouting is a mixture of quantitative and qualitative data. However, few teams actually have a good enough scouting system to collect meaningful quantitative data, much less qualitative data. I've reviewed data from teams (including my own) after events, and at least half of the data sets I have seen are absolute garbage. I believe that almost all teams should focus first on improving their quantitative scouting before even thinking about qualitative scouting.

Basically, if your scouts can't reliably count the number of totes a team can stack in a match, why would you trust that their "qualitative" assessments of teams mean anything at all?

We've had our issues with bad data at during events as we have worked our way through various scouting methods: pen/paper input into Excel and our tablet scouting from last year.

That has put us in positions where we have no accurate data during an event which has been caused by technology failing or the input/collection failing. Either way I can still rely on my scouts to give me and our pick list makers a good evaluation of robots. Last year this played a critical role at two of our events and many before then where we had to say, "Data aside how do we feel about this robot from what we've seen".

Like many have said in this thread you need to prioritize what you are scouting and why which is why we work hard to make sure our scouts know what to look for and why. If our data collection fails we have over half a dozen students who can offer their honest opinions regarding what they saw and work through their thoughts as a group to get a good feel for the field of robots.

I think too many teams overlook the aspect of having a serious discussion regarding scouting and why teams need to do it which is why they get poor results. If you want to be successful at an event you need to treat scouting like you do the drive team in terms of importance and not a separate group in the stands.