How to Win a Robotics Competition

[JesseK]

Originally Posted by Carol

And another big factor in winning.


Luck.

Yep.

Luck in who's available to be picked during selections.
Luck in that the right positions on the field were chosen for any given moment during a match, knowing only slivers of information before a match starts.

[sanddrag]

It would be really great if someone did a "How to win a robotics competition" 30 second video spot, modeled after the Dodge Dart commercials.

[Racer26]

Originally Posted by sanddrag

It would be really great if someone did a "How to win a robotics competition" 30 second video spot, modeled after the Dodge Dart commercials.

Oh 1114 or 148? Anybody listening? I seem to recall JVN was the one that originally linked me to that Dart commercial, making a comment on how much it paralleled the 148 method...

[Alpha Beta]

I definitely agree with your first two points.

1. Build robust, easy to maintain machines.
Defense is taken to a whole new level during elimination matches, and their is very little time to put things back together. If we don't think your machine will hold up, we can't pick you to join our alliance.

2. Practice Driving.
Driving practice is a huge advantage, especially at earlier tournaments where even the best machines are in their first event. A well driven machine also makes an attractive defender for someone's pick-list (even if you've not taken the opportunity to play any defense during qualification rounds.) The chemistry on your drive team also leaves an impression.

On point 3...

3. Don't worry about ranking, just win.
If you're just trying to impress scouts maybe that works. As a potential captain you've got to play the raking system given. Intentionally ignoring it, especially when discussing pre-match strategy, makes the team seem uneducated. Negotiate a tolerance threshold with your alliance that balances the priority for winning with maximizing your seeding. Even better if you can incorporate scouting data from previous matches into how that threshold is calculated.

Additional Points...

4. As you eluded in your article, be prepared to be a captain. There are surprises in the top 8 every year. If a team is absolutely too small to scout then here are a couple of ways to get some data...
a. Set up a camera to record a full field view of the matches and then review at least the most recent 2 matches for each robot Friday night. It can make for a long evening, but some people really enjoy that kind of film study.
b. Tour every pit before the end of Friday and get to know what is available to you on the pick-list.
c. At least come to the tournament with a pre-list where you predict how good each team will do. At the World Championship that is fairly easy considering how much data is available to you.

5. Luck... "Chance favors only the prepared mind."
No one wins a regional by being lucky, but great teams can lose regionals by being unlucky. The qualification schedule can be a cruel thing. As long as you can guarantee that you have a great robot the odds will always favor your alliance having more great robots than the randomly assembled opposition. Sometimes the match-ups just don't work out. Sometimes a poorly driven robot on your qualification alliance goes on a penalty fest. Sometimes a weaker robot winds up as one of the top alliance captains and you can't say no to their invitation. And sometimes no matter how bad you think your situation is, the other side has worse luck and you still win. No point in giving up too early.

[edwardj]

Originally Posted by Alpha Beta

4. As you eluded in your article, be prepared to be a captain. There are surprises in the top 8 every year. If a team is absolutely too small to scout then here are a couple of ways to get some data...
a. Set up a camera to record a full field view of the matches and then review at least the most recent 2 matches for each robot Friday night. It can make for a long evening, but some people really enjoy that kind of film study.
b. Tour every pit before the end of Friday and get to know what is available to you on the pick-list.
c. At least come to the tournament with a pre-list where you predict how good each team will do. At the World Championship that is fairly easy considering how much data is available to you.

I'm not a scouting expert, but I think people tend to overcomplicate scouting with incredible amounts of data.

At a robotics tournament, the main difficulty with scouting is because there's way too much data on everyone's robot, and it's hard to process all of that data.

In my opinion, robots that I've seen fall into three categories (especially at FRC regional tournaments):

• Definitely want on our alliance
• Maybe...
• No way!

There's a LOT of "Maybe" and "No ways". Usually your first pick is pretty obvious, but the main problem is -- for your second pick, you run out of "Definitely want on our alliance", and how do you pick between the "Maybe..." or the "No way!"s that are left?

By then I just wouldn't really care, because it's very difficult to determine a better team from a heap full of "Maybes". Using a simpler vs. more complex scouting method is unlikely to influence the overall success of your team -- luck is more influential at that point.

Which makes me think: it's really bad game design to make a game rely on sheer luck. If good teams cannot consistently win over worse teams in the competition, it's a broken competition.

Even though poker's core game is luck-based, there's still a lot of strategy around a poker tournament -- where better players can consistently win over lesser players.

FIRST has a lot of goals with the competition -- and evaluating the most "talented team" at playing the game is the point. Inspiring students to learn more about science and technology is -- and they do a great job at it.

[$wimmer3138]

Originally Posted by edwardj

By then I just wouldn't really care, because it's very difficult to determine a better team from a heap full of "Maybes". Using a simpler vs. more complex scouting method is unlikely to influence the overall success of your team -- luck is more influential at that point.

It all comes down to how you collect your data and how it’s organized. Over the past three years we have tweaked with a scouting system and I my opinion it was a major reason that we won the Buckeye regional in 2011. Our second pick had a compatible mini-bot deployment which allowed us to win the regional by winning the mini-bot race.

[MichaelBick]

Originally Posted by edwardj

I'm not a scouting expert, but I think people tend to overcomplicate scouting with incredible amounts of data.

At a robotics tournament, the main difficulty with scouting is because there's way too much data on everyone's robot, and it's hard to process all of that data.

In my opinion, robots that I've seen fall into three categories (especially at FRC regional tournaments):

• Definitely want on our alliance
• Maybe...
• No way!

There's a LOT of "Maybe" and "No ways". Usually your first pick is pretty obvious, but the main problem is -- for your second pick, you run out of "Definitely want on our alliance", and how do you pick between the "Maybe..." or the "No way!"s that are left?

By then I just wouldn't really care, because it's very difficult to determine a better team from a heap full of "Maybes". Using a simpler vs. more complex scouting method is unlikely to influence the overall success of your team -- luck is more influential at that point.

Which makes me think: it's really bad game design to make a game rely on sheer luck. If good teams cannot consistently win over worse teams in the competition, it's a broken competition.

Even though poker's core game is luck-based, there's still a lot of strategy around a poker tournament -- where better players can consistently win over lesser players.

FIRST has a lot of goals with the competition -- and evaluating the most "talented team" at playing the game is the point. Inspiring students to learn more about science and technology is -- and they do a great job at it.

This is especially untrue at championships. While many of the powerhouse teams have relatively equal robots, the 3rd robot makes all the difference. Even at regionals a third robot can make much of the difference between winning and losing.

[Siri]

Originally Posted by MICHAELABICK

This is especially untrue at championships. While many of the powerhouse teams have relatively equal robots, the 3rd robot makes all the difference. Even at regionals a third robot can make much of the difference between winning and losing.

It's not even true pre-Championships, but the density of teams who realize it is lower.

The basic fallacy is that there's a single spectrum on which all teams should be ranked. Definitely/maybe/no way aren't bad classifications (higher resolution can be useful), but assigning them on a single-spectrum basis ignores the basic premise of elimination matches. I'll keep it basic, but what if you ended up with 3 shooters or 3 full-sized long bots? You can afford someone with a poorer autonomous if they can feed or you have the fastest guy to the bridge. You could skimp on the third 2011 minibot if necessary to nail an awesome feeder/defender or scorer. Do you need a low standard deviation feeder? A highly adaptable strategy-filler? Do you have triple balance compatibility (2012)? If there won't be reliable autonomous modes left on the second round, can you snag a double-tuber first (2011)? [This goes back through the years]

[PayneTrain]

The key to scouting is to not pick the two best robots left, but the ones left that complement your team's abilities the best. You can scout out the 5 best starting pitchers in baseball, but not even the Yankees can afford to buy all of them and have money left for position players that can best complement the rotation.

[JosephC]

Scouting isn't just important in Alliance selections; when your looking for the best robot to fit into your alliance, and not necessarily the best overall robot, but also in Match play.

Teams that do not scout will be going into matches blind, with no idea of what to do. Those who have collected data on robots can say, "Robot A is a really good shooter, we need to play defense on them; hey alliance partner B, since our data says your not the very best shooter would you mind playing defense on Robot A for us?"

Now for luck. Personally I believe that luck has very little to do with FRC. I've heard 5-6 teams say after a match, "Team A won because they got lucky." Unless there was a penalty that should of been called and it wasn't; then Team A used something called "Strategy". Strategy is this little thing that weaker teams use to beat stronger teams; and it isn't "Luck"

As for Quals; Teams can get lucky 2 or 3 matches and have some really good match ups; But I've never seen a bad robot be in the Top 8 after 12 matches. They may not be the best scorer, but they might be a great defensive robot.

Everything boils down to this:
FRC is 1% Luck; 40% skill and 90% Strategy.

[Ed Law]

Originally Posted by JosephC

Scouting isn't just important in Alliance selections; when your looking for the best robot to fit into your alliance, and not necessarily the best overall robot, but also in Match play.

Teams that do not scout will be going into matches blind, with no idea of what to do. Those who have collected data on robots can say, "Robot A is a really good shooter, we need to play defense on them; hey alliance partner B, since our data says your not the very best shooter would you mind playing defense on Robot A for us?"

Now for luck. Personally I believe that luck has very little to do with FRC. I've heard 5-6 teams say after a match, "Team A won because they got lucky." Unless there was a penalty that should of been called and it wasn't; then Team A used something called "Strategy". Strategy is this little thing that weaker teams use to beat stronger teams; and it isn't "Luck"

As for Quals; Teams can get lucky 2 or 3 matches and have some really good match ups; But I've never seen a bad robot be in the Top 8 after 12 matches. They may not be the best scorer, but they might be a great defensive robot.

Everything boils down to this:
FRC is 1% Luck; 40% skill and 90% Strategy.

I completely agree with you. Scouting and strategy are my passion. That's all I do at competitions. However I think you meant FRC is 0.76% Luck, 30.53% Skill and 68.71% Strategy.

[F22Rapture]

Originally Posted by edwardj

I'm not a scouting expert, but I think people tend to overcomplicate scouting with incredible amounts of data.

At a robotics tournament, the main difficulty with scouting is because there's way too much data on everyone's robot, and it's hard to process all of that data.

I agree 100%. While I'm sure this isn't true for everybody, I've always found that the majority of scouting data never really gets used to make a decision.

• Robot's technical ability to complete each of the subtasks of autonomous, scoring, defending, and endgame from 0 to 10, judged to a certain extent on Thursday to be revised later
• Overall performance from 0 to 10 from each of the 10 matches

Each of the subtasks will be weighted by importance to the game (30% for endgame, 40% scoring, etc.) and then averaged together, giving a single technical average to deal with.

The each of a team's performances from their10 matches are averaged together with the first 4 matches weighted 5%, second 2 matches weighted 10%, and the last four matches weighted 15%. This gives a single performance average to deal with.

It's very simple math, it's not much data to go through, and it can be easily done in Excel with no special scouting software. And once we've finished, it's very easy to sort, filter, and manipulate to get a good list of teams. It's perfect for smaller teams since it only takes 2-3 scouts to come up with a single performance score for each robot, and maybe a few comments to change their tech score and note their predominant strategies. And they can spend their time watching the match instead of writing.

It's not fancy, perhaps not Nate Silver levels of accuracy, but it's fast and can narrow a field of 40-80 teams down to only 10 or 15 almost instantly (with little information backlog), meaning that you can then focus your time and efforts on those teams when you start going around the pits to make connections.

[dcarr]

Originally Posted by F22Rapture

I agree 100%. While I'm sure this isn't true for everybody, I've always found that the majority of scouting data never really gets used to make a decision.

• Robot's technical ability to complete each of the subtasks of autonomous, scoring, defending, and endgame from 0 to 10, judged to a certain extent on Thursday to be revised later
• Overall performance from 0 to 10 from each of the 10 matches

Each of the subtasks will be weighted by importance to the game (30% for endgame, 40% scoring, etc.) and then averaged together, giving a single technical average to deal with.

The each of a team's performances from their10 matches are averaged together with the first 4 matches weighted 5%, second 2 matches weighted 10%, and the last four matches weighted 15%. This gives a single performance average to deal with.

It's very simple math, it's not much data to go through, and it can be easily done in Excel with no special scouting software. And once we've finished, it's very easy to sort, filter, and manipulate to get a good list of teams. It's perfect for smaller teams since it only takes 2-3 scouts to come up with a single performance score for each robot, and maybe a few comments to change their tech score and note their predominant strategies. And they can spend their time watching the match instead of writing.

It's not fancy, perhaps not Nate Silver levels of accuracy, but it's fast and can narrow a field of 40-80 teams down to only 10 or 15 almost instantly (with little information backlog), meaning that you can then focus your time and efforts on those teams when you start going around the pits to make connections.

If your final score for each team is based on its overall performance, however, this data isn't particularly useful in discerning the strengths of different teams and how they fit into an alliance.

[F22Rapture]

Originally Posted by dcarr

If your final score for each team is based on its overall performance, however, this data isn't particularly useful in discerning the strengths of different teams and how they fit into an alliance.

Well, my overall point was that you can figure out the overall strength directly after the seeding matches end, and then spend the next 2 hours working out which robots of those are most compatible with your own and making connections with those teams. It's primarily a way of focusing your effort, not doing everything for you.

[dcarr]

Originally Posted by F22Rapture

Well, my overall point was that you can figure out the overall strength directly after the seeding matches end, and then spend the next 2 hours working out which robots of those are most compatible with your own and making connections with those teams. It's primarily a way of focusing your effort, not doing everything for you.

Got it. That makes sense. You want to be choosing a robot that fits your alliance needs from a subset that has been determined to be good.