I want to include a defense section in my personal match scouting database. What I have is the following:
Actual Blocked Shots
Points Prevented
Teleoperated Score Differential
I don’t know how I should make a Defense Efficiency Formula with what I have (by adding each together after it is multiplied by the weight/coefficient). Also I know I need more categories for my defense section but I don’t know what else to add. Any and all help would be greatly appreciated.
Just like professional sports, point prevention is a very difficult thing to quantitize. Just ask any stats whiz who enjoys baseball or hockey. There’s some “glamor” stats that are relativley easy to track, such as blocked shots, but those don’t always correlate to actual defensive performance.
In matches not featuring full court shooters, blocked shots are very rare this year. There are a handful of machines that will still attempt to block shots around the pyramid, but usually no more than one or two teams per regional will have any luck with it (and even then, it will only work against certain opponents shooting from certain locations). A majority of defense this year is played on robots in transition between the loading areas and their scoring locations. Thus it is very difficult to create a truly objective and standardized measure of their performance, as countless variables impact it.
I would also add a weight if a tean has a device to block full court shots. It can be very important during elims.
Record penalties committed by that team, if possible. Keep defense clean
I realized that defense is a lot more subjective than objective. I was also thinking of just making the defense section just a rating like 1-10 just by the eye test. I was also thinking that the scouter can just make a note about the robot if they believe the stats don’t truly represent the ability. Like if they have very little offensive stats then it could be due to they are a defensive type robot or a team was playing defense on them. I might just remove the defense stat section all together and like i said just use the eye test.
If you want to quantify something, here is what I recommend: Count the number of shots attempted by the opposing robot. That is, the number of shots that actually make it to the alliance station wall. Whether they are blocked, or just never fired in the first place doesn’t really matter, because if it doesn’t get near the goal, it doesn’t matter. The best kind of defense is preventing any discs from getting anywhere near the goal. This is a tricky stat to track when the defense bounces from one robot to the next. But, if the defender is actively getting in the way, count the total shots that get through.
Another important stat is the robots ability to prevent a hang. A good defender can make up for not hanging itself by keeping an opposing robot away from their pyramid.
we just use pit scouting to scout defense. Meaning, scout for type of drive train, gearing, and wheels. Any robot with a good drive train could play killer defense.
The best defense I’ve seen this year didn’t block any shots. All these teams did was slow down the robot(s) as much as possible in the transition between the station and the pyramid.
Oh man, I can’t wait for Corsi and relative Corsi stats for FRC. 
SPAM has the best formula for scouting defense I have found. They show the formula in “Poor Man’s Scouting System” published here on CD Media. Essentially they count “defensive maneuvers” during a match and assign a coefficient value to per move.
I’m sure they can explain it better than my attempt.
This is definitely one of the bigger fallacies in FIRST regarding defense… On the surface it makes sense - one of the only mechanical things you need for defense is a drive train that doesn’t fall apart, and has some decent traction, torque, and speed.
Really though, what makes defense effective has very little to do with the drive train and much more to do with the drivers. Find a team whose drivers are smooth (the robot goes quickly and smoothly where they want), who understand the rules, and have good “heads-up” drivers… and you’ll have a very good defensive partner. I’ve seen admirable defense played by good drivers behind slowish drivetrains or a drivetrain with little pushing power (mecanums). I’ve also seen mediocre drivers behind the sticks of a great defensive drivetrain (high traction, low and high gear) that simply weren’t effective.
I’d say look for the teams that have the best drivers out there (they are quite potentially also a scoring robot). Look for them and watch them. If you have a couple people watching each match for good drivers and talking to each other, you likely won’t need “objective” statistics (even though I’m a huge stats fan). See if they seem to make “heads-up” driving decisions. Talk to your drive team to see what they’re impressions are of the teams on your list. This should be as effective (or more effective) than almost any defensive stat.
Fact of the matter is that defense isn’t something you can see immediately. Field awareness is key to having a successful defense. A team that has been playing “bump and run” all qualifying matches will be comfortable doing that in elims. Bump and run can be effective if it is just one robot that you have to contain. Finding a team comfortable playing defense is key because sometimes it is more efficient having 2 scoring robots than 3 robots getting in each others way trying to put in Frisbees.
The best defense I’ve seen has been teams staying in a general section and picking at teams as they come by (this is effective against a team with 2 scoring bots), defenses that chase shooters and seem to have no general plan don’t do much. People may say drive train isn’t important but it is. Higher traction = more push. So the only excel formula that might give some coloration may be point differentials in matches, and also compare the OPR of teams they have played. Bet we normally find defense robots based on field performance and if we can work well with their drive team.
Point differentials and number of blocked shots only tell part of the story because a couple variables affect point differentials and some of the best defensive tactics don’t involve directly blocking shots.
It’s also hard to identify all the teams that are prepared to play excellent defense because some teams who play great defense in the elimination rounds actually play offense throughout the qualification matches.
In addition to a good drive train, good blocking devices, and a heads-up disciplined drive team, teams that are ready to play good D are backed up by very good scouting systems. When my team allied with FRC 118 in the Lone Star elimination rounds, we had the opportunity to look at their scouting data and we were amazed by the amount of valuable qualitative data they collect on teams. FRC 118 has a very high scoring machine, but they can switch to very effective defensive play and can coach allies to up their defensive game because of their scouting data.
Seeing their system inspired us to revamp our scouting system prior to our next Regional and it has helped TREMENDOUSLY. In addition to all the quantitative data that teams collect, our scouts also gather favored robot drive patterns, favored feeds, places where robots shoot and accuracy at each spot, etc. This info helped our drive team strategize with our alliance (Exploding Bunny Appreciation) to upset some really high scoring alliances at the Alamo Regional.
Moral of the story: If you want to quickly ID some teams that are preparing to play excellent defense in the elimination rounds, send a super scout into the stands to peek at the data that other scout teams are collecting on robots. In addition to quantitative data, look for qualitative data that’s designed to give drive teams a clear picture of the driving habits of their competitors and systems that get that data to the drive teams in time for them to prepare for upcoming matches.
I think that defensive formulas are tough to come up with in regards to this years FRC game. Some of factors that influence this are size limitations related to specific locations on the field, obstacles of the course(towers), etc.
I am an ex basketbal coach with 22 years of coaching experience (high school and college) as well as coaching several other sports for many years. As a basketball coach, I am of the opinion that making a player get to their sweet spot (so to speak) makes a player work harder to be comfortable. If they are uncomfortable, then you can begin to effect them.
I do not think that blocked shots are the key to great defense unless you are talking specifically about full court shooters. Most full court shooters want to align to the goal and line up for the shot. If you get in the way of the shot, they move to the pyramid to shoot thus making them a cycle runner. Very few are willing to stay true to the long range shot by just moving to another location on the field.
For pyramid shooter/cycle runners, make them work hard to get up and down the court. Middle cuts, side cuts and under the tower cuts are all moves that a defensive bot should be able to do. FRC 2789 did an excellent job of all of these at the Lone Star and Alamo Regionals. In Semi Final 2 match of Alamo, they occupied two other robots for 100+ seconds of the total 240 seconds for those two robots. That time equates to stealing cycles from their strategy. Martin did an excellent job of driving as well as taking coaching to change strategy throughout the matches. Watch the middle of the field in this video for reference: http://www.youtube.com/watch?v=xM1WIp2qe-I Notice that on several occasions, the opposing robots have to take a different path to get to the feeding station thus stealing time off the clock.
Defense of a cycle runner is not just trying to push them around but rather making some contact, not afraid to be physical, and cutting off paths.
Patterns are easy to predict if you watch enough matches. Coaches of sports teams watch endless hours of video and use all kinds of ways to try and predict the patterns of not only plays but even more importantly specific players.
Excellent observations, Norm. Basketball is the most similar sport to Robotics with its multiple players and contrasting flows.
Defense gets little glamour and very, very few teams have designed for it this year.
Nonetheless, in the NCAA Final Four for basketball, Louisville pulled it off last night with its relentless full court press, that eventually got into Wichita States “head” late in the game, when they committed three turnovers in one minute after none in the first 26 minutes. It changed the outcome of the game.
Syracuse, on the other hand, with it’s noted 2-3 zone defense, fell short against the run and gun (cyclers) of Michigan.
So, how to scout for defense? I like this approach:
Using a subjective coefficient requires a well trained scout team. At the CMP with over 100 matches, this can be tiresome. Share with another team and provide much caffeine. :eek: :ahh:
Defense will most likely be the deciding factor in the CMP divisions and on Einstein.
I think we will see some <84" pyramid shot blockers to shut down cyclers, as well as very effective mid court “presses”.
The strategies will align and the winners will be on the alliance with other ones whose shooters do not jam.
That, and autonomous is more important this year than any before.
Defense is an attitude for sure.
You cannot turn a robot/drive team into one if they do not have time to practice it or get comfortable with it. Penalties are easy to accumulate trying to play defense.
BTW, we started the year using SPAM’s system at Hub City in Lubbock but defense evaluation was all over the place. What one student saw was not anything like what others might see. The data was not usable for us due to this factor. The scouting system was very effective though not in match evaluation but also in helping with alliance selections.
At Alamo, we had two students who were scout leads. One or both of them watched all matches. We also shared scouting with a few other teams (801 and 1592) using our system based on SPAM’s system. We had students who watched videos before the tournament to get an idea of teams that had competed already. We had a few other students who watched defense for special defensive focus by teams. Two of the teams using the scout system were in the finals against each other.
2789 was chosen for defensive abilities and proved their worth in the matches for us. 4063 was a pyramid shooter/cycle runner but also played opportunistic defense when allowed. They interfered with Torque on a few occasions as they were moving from end to end.
The difference in matches comes down to a few frisbees scored or not. Defense is not about shutting a team down completely but limiting their production.
A lot of the teams looking for somebody to play defense also expect some offense, too. Our second round pick list consists of teams that can score in autonomous (#1) and can hang for 10 (#). After that, we try scout for “field presence”. Field presence is a subjective quality that is difficult to assign a number to (and even harder to get consistent data on from student scouts). On each team’s match scouting sheet, we have a “general notes” section where the scouts check off applicable items from a list of “strengths” (fast chassis, always on task, good driver, etc.) and “weaknesses” (penalty magnet, slow, no obvious game plan, etc. During pit scouting, we ask teams how they feel about playing defense, note their drive train & quality of construction, etc. In summary, we don’t specifically scout for defense. We look for teams that will be an asset on offense, but have what it takes to step into a defensive role. If any of the robots on the alliance have a partial breakdown during a match and can’t score, then they go on defense and the “defense” bot goes on offense.
Well said Wayne. Kudos to you and your team.
Martin is an amazing driver, and we did pour over hours of web streams as a team to analyze strategies for our zone defense. What made our job easy was that we were on an alliance with three very different robots that approached the game in different ways. We were able to tangle up several opposing alliances because of the simple fact that they shot from and fed from the same spots as their partners. The other essential ingredient in the recipe of ur success was how well we coordinated to your field marshaling, Coach Norm. When you needed us to set picks, you told us, and that kept the offense up and running.
Since everyone’s using sports analogies, I’m going with my favorite sport: football. I think that the key to being a good 2nd pick has been knowing when to play linebacker to stuff the running game of the opposing alliance, and knowing when to play fullback to push defenders out of the way. We don’t have a fancy robot, but our smart driveteam and scouting goes a long way 