50 Pt. Climbs

[cmrnpizzo14]

This season, it seems as though there was a negative rep for the 50 pt. climb and dump. People said that it was a bad strategy for teams to solely devote themselves to and as a whole this was in fact a largely unsuccessful strategy for most teams.

My question, mostly to teams that attempted this but also to those who considered it, is why did you find this strategy successful or unsuccessful?

Keep in mind that according to EWCP's Twenty Four blog, a 50 pt. robot on an alliance should be more than pulling their own weight. The median alliance score was below 50 points (for qualification in week one, not the greatest sample but it should be indicative). There were few alliances in qualifying that could attain over 150 points and at most regionals few that could in the elimination matches as well. Statistically speaking, a robot that contributed 50 pts should be a huge alliance contribution.

Was it consistency issues? Was it too defensible? Why didn't this work?

[EricH]

I can think of a few reasons.
1) Lineup time. Depending on your climb start point, it could take a while just to line up.
2) Defense. As long as you didn't get called for interfering with a climb or contact with the pyramid, opponents that had to line up precisely had to go around you as a defender. If you played some physical defense, guides to help line up could be damaged (and in fact I saw that happen--the team who was trying the climb and dump couldn't line up the second time they tried it, despite the defender backing off).
3) Difficulty of climb. Take the number of 30-point climbers--reliable, consistent ones--out of the number of climbs attempted at any level. Not too terribly many, by comparison.
4) Now carry discs while seeing #3. And #2 due to defense being played as soon as color was spotted in your dumper. How many discs will make it?


When you factor in the rest of the game, say with a cycler (no floor pickup, automode shooting only preload with 100% accuracy, 10-point hang), that 50 points becomes only 40 (hang), then 22 (automode), or two cycles of 4 discs each. That's a moderate robot, one that would inhabit most regionals somewhere around the bottom of the alliance selections, maybe late first-round.

Essentially, a 50-point climb-and-dump with nothing else onboard to score is at about the same level as a robot that just does cycles with a good automode.

That said, when a 50-point climb-and-dump has something for, say automode, THAT's when it becomes difficult to beat. 3309 comes to mind--they shot during auto, and didn't show the C&D device until sometime around semis in L.A.--at which point they became the biggest target for their opponents' defender.

[KrazyCarl92]

For teams that could integrate a climb and dump with the ability to score discs in the high goal (67, 1114, and 1334 come to mind), the climb and dump is a way that they could raise their alliance's ability to score, but did not exclude any of their partners from scoring points. What I mean by this is that a 7-disc auto is great, but having two robots with 7 disc autos on the same alliance is superfluous. Having additional cycles beyond a consistent 13 total for an alliance becomes superfluous. Additional disc points past a certain point are exclusive from each other. Climbers that climb up the outside of the pyramid to 30 and dump in the top are not exclusive of other alliance partners' points (aside from colored discs being limited). Another interesting aspect of climbing that fits into alliance structure is that while climbing up the outside of the pyramid to 30 points would not exclude potential partners from points, climbing up the inside to 30 is an action that is largely exclusive to a single robot per alliance (perhaps not, but I'm not aware of 2 robots on the same alliance climbing up to 30 inside the same pyramid this year).

From an individual team design stand point, we made the decision on day 2 of the build season that based on our resources we would have to decide between primarily climbing/dumping and largely forgoing other disc points, or going for disc points with a 10 point hang. Based on our estimation of contributions we could achieve from different strategies, we figured we could choose either:
-about 50 points/match if we chose climb/dump
-about 80 points/match if we chose 10 point climb and scoring discs

We chose the second option because we felt that it would improve our chances of winning matches, and it also felt like a more repeatable strategy. Missing a single climb takes away all of our points. Missing a single cycle or a single shot is 12 or 3 points, respectively. All things considered, I would rather fail at missing a single shot than at missing a single climb.

With the decision on overall strategy determined, we moved forward with prototyping around functional objectives based on executing that strategy with our robot design.

It's not so much that the 50 point climb and dump was not valuable. It is just that perhaps many teams who tried it could have benefited from a competitive stand point by making a decision against the climb and dump early on. The teams that did execute it in addition to scoring lots of discs in teleop were extremely valuable as alliance partners, hence why 2 of them partnered with a 7 disc auto team and a great defensive robot to win IRI.

[Nuttyman54]

Originally Posted by cmrnpizzo14

Was it consistency issues? Was it too defensible? Why didn't this work?

Yes, and yes.

The simple answer is: it did work as EWCP predicted. Just look at 4451, a pure C&D bot who (as rookies) ranked in the top 8 at both of their regionals, reaching the semis at their first and finals at their second.

In addition to what Eric said, the first key to remember is that the Twenty-Four blog is about statistics. 50 points at early regionals will win MOST of your qualifying matches, IF you do it every match. There are no guarantees, especially when you factor in that opponents who recognize the threat will adapt to it.

The second key is that a pure 50pt climb/dump robot is capped at 50 points. They physically cannot score any more points, whereas a pure cycling robot will usually improve over the course of the event and season.

It was INCREDIBLY difficult to be a consistent climbing robot, and for most teams it took at least 30 seconds to climb, 5 seconds to dump, plus 10-15 seconds for alignment. If it takes you 30 seconds to get across the field, load up your frisbees and get back without any defense, you're already taking 1:15-1:20 seconds of the match to do your thing. If anything goes wrong, or if there's defense, or some combination, you really start pushing against the buzzer. Forget about trying to play some defense and then make a quick climb and dump at the end of the match. Not to mention that a simple human-load only disc shooter could often score more points in less time.

The kicker is that because it's a pretty all-or-nothing strategy, all it takes is a few seconds of delay and the 30 point climb/20pt dump becomes a 20pt climb and no dump (or worse). That's a pretty big points swing for anything going wrong in a match.

Lastly, climb/dump only robots had no auto mode contribution, which combined with my previous points means that they were generally a liability on an eliminations alliance. All other things being equal, if your opponent's robot scores all 3 discs in auto, and your climb/dumper scores none, that's a 36 point differential. You can theoretically make that up with the 50pt climb/dump, but some combination of defense on you and/or them scoring 5+ discs means that doesn't happen, and they win.

I saw more successful pure 30pt climbers that didn't dump in elims primarily because removing the need to go get colored disks and come back (the most defendable part of the strategy) meant that pure climbers could play some defense first.

[Chris is me]

Funny you mention EWCP's TwentyFour blog. A post Andrew and I are working on will be debating the "was / was not worth it" arguments for the 50 point climb.

I maintain that with a fast enough climb speed, a good enough driver, and the right circumstances, a 50 point climber could have been an effective alliance partner all the way up to the Championship elimination rounds. I'll either finish that blog post or post again in this thread if I don't have time for the former.

[cmrnpizzo14]

Originally Posted by Chris is me

Funny you mention EWCP's TwentyFour blog. A post Andrew and I are working on will be debating the "was / was not worth it" arguments for the 50 point climb.

I maintain that with a fast enough climb speed, a good enough driver, and the right circumstances, a 50 point climber could have been an effective alliance partner all the way up to the Championship elimination rounds. I'll either finish that blog post or post again in this thread if I don't have time for the former.

I'm glad I'm not crazy! Everything I was looking at was telling me that if it in fact worked a 50 pt. climb and dump would have been an incredibly solid alliance partner.

I don't have my team's scouting data from last year readily available, maybe someone else does. What was the average accuracy in teleop for the 3 point goal? Or really any goal for that matter? It seems that many assertions of why a RI3D robot was better than a pure climber was that it shoots at nearly 10% accuracy. I feel that there is almost no way that this could be the case and that most cyclers were shooting at around the 50% mark. I would assume that a quick cycler shooting at 75% would be the top scorer (or in that similar tier) at most regional events and possibly at a CMP division.

[EricDrost]

I think the main issue with this argument is that it's based on a robot's theoretical ceiling. There's a big difference between having a 50 point ceiling and a 50 point average. On 11, we had a ceiling of 18 points in autonomous. Do you think we hit that ceiling every single match? Not a chance.

To add to the problem, if you only got 2/3 or even 9/10 of the way to the top, you get only 40% of your theoretical point ceiling. There's also a risk of falling once you get to the top.

Teams like 4451 who got 50 points nearly every match were outliers.

(Side note, the folks at 4451 are awesome.)

[BrendanB]

Originally Posted by Chris is me

Funny you mention EWCP's TwentyFour blog. A post Andrew and I are working on will be debating the "was / was not worth it" arguments for the 50 point climb.

I maintain that with a fast enough climb speed, a good enough driver, and the right circumstances, a 50 point climber could have been an effective alliance partner all the way up to the Championship elimination rounds. I'll either finish that blog post or post again in this thread if I don't have time for the former.

I would say IF the climber was fast enough building a pure climber/dumper was worth it. Unfortunately most teams with a fast climber had a frisbee game with them (67, 1114, 1918, 1986, 254, 1334, 3467, 1640, 71, etc) or they were very slow/unreliable. Some dedicated climbers didn't have dump capabilities which really wasn't the greatest decision because that put their ceiling down at 30 points.

A very deadly combo would have been a 15-30 second climber/dumper that could play very effective defense and block full court shots before climbing. I could see this robot doing very well at a regional and making into eliminations at the championship but it really comes down to how fast your climb is and how good your driver at defense.

Looking back I can imagine most teams who were strictly climbers wouldn't do it again if they had to redo 2013.

[Ian Curtis]

Originally Posted by cmrnpizzo14

My question, mostly to teams that attempted this but also to those who considered it, is why did you find this strategy successful or unsuccessful?

In retrospect I think the biggest nail in the coffin is that there were easier and more reliable paths to 50 points. Scoring frisbees was easy relative to past FRC shooting games like Aim High and Rebound Rumble.

Lots of good insight in this thread

[Kevin Leonard]

Let's think from a picking perspective. At Connecticut, there was a very good, consistent Climb/Dump, Team 2170.
We (20) we the 1st seed, and we were trying to figure out who to pick. We were between teams 195, 177, and 228. 2170 came up in our conversation, but they simply didn't offer the same # of points or the same versatility as any of the other three options.

I think Climb/dumper-ers failed (at least at later regionals) because of that. At MOST regionals, there were at least a few robots that could outscore a 50-pt climb/dump. And if you're looking to beat the first seed alliance, what do you want? A robot that will consistently score 50 points, or a robot that has the potential to score 80.

We talk about 4451 as being one of the most successful climb/dumps, but they never won an event(you can srgue that they should have, but the fact remains that they didn't), and they weren't picked at Champs because of this.

[Jim Wilks]

What about IRI? 2 of the first 3 members of the winning alliance were 50 pt climbers. Other than having to share the 6 colored discs, this worked well and they came home with a blue banner.

Clearly, I have a bit of a bias on this subject as a member of one of those 3 teams.

[BrendanB]

Originally Posted by Jim Wilks

What about IRI? 2 of the first 3 members of the winning alliance were 50 pt climbers. Other than having to share the 6 colored discs, this worked well and they came home with a blue banner.

Clearly, I have a bit of a bias on this subject as a member of one of those 3 teams.

True, however none of those teams were a dedicated 30 point climber/dump. Both scored 3 of 3 in teleop with 3-5 cycles.

One of the arguments mid season was from teams who set out to solely climb the pyramid and/or dump and yet it wasn't a winning strategy getting beaten by simple cycling robots. The main problem in the dedicated climber strategy is it has to be something you can do reliably, quickly, and play defensively for the rest of the time to be a winning strategy.

I believe climbing can be a core part of a winning strategy. At IRI we had several matches where we were out of frisbees and all that was left was climbing. In semi final 1.1 our team finished 4 cycles with just under a minute left and I believe our partners were out shortly there after and 33 began sweeping the field.

[Francis-134]

You all bring up great points as to why climb and dump robots were not as successful as was predicted or hoped. As a member of a team that built such a robot this year, hopefully I can lend some insight.

The crux of the issue comes down to the fact that building a robot that can complete the wombo combo is really difficult! As shown through the season, a 30 point climb is very challenging, and brings a lot of risks if not looked at carefully. Take my team as an example. We never successfully completed a 30 point climb at our first regional and fell off the pyramid three times during the season (as an aside, it is quite a site to see your pit crew jumping on the robot frame to bring it back into square). Add in collecting and dumping frisbees and you have a very complex robot that is already not a realistic solution for most teams.

As far as raw points are concerned, a climb and dump must be part of other point differential methods to be competitive at a high level. At a regional, a 50 point play is already enough to put you above most third partners on an alliance. If you can throw in a 12 or 18 point auto mode, you can boost yourself to a very competitive robot. Championship is a different story. If your alliance's goal is to win the entire thing, you basically need to beat a four-cycle robot pretty consistently. Look at the following:

Four-Cycle
Auto: 18
Teleop: 48 (4 disks * 4 cycles * 3 points per disk)
Endgame: 10
Total: 72

Wombo Combo
Auto: 0
Teleop: 0
Endgame: 50
Total: 50

Somewhere in there, you have to create a 22 point differential between you and your opponent robots. As mentioned in the Twenty-Four blog, you can hope to do this by playing enough defense to remove a cycle, and then score at least 12 points in autonomous. Is it possible to remove an entire cycle’s worth of points in 60-70 seconds of defense? Yes, but it is very difficult (see GTR West QF 1.1 as we manage to remove at least one cycle from 1114/2056).

The big problem is that you now HAVE to score a complete wombo combo at the end of the match, which as I mentioned before, is very difficult to execute with a high degree of consistency. Failing to climb to 30, and/or not getting the disks out will put you way too far behind to win.

All of this also highlights the big issue with the strategy; you cannot carry an elimination alliance to victory off of a climb and dump alone. High powered disk robots that can complete 6+ cycles and a 10 point hang on their own can lift otherwise poor alliances above mediocre ones through sheer firepower. Taking GTR West QF 1.1 as an example, despite the fact that we did what we needed to do, there is no way that our robot could carry this alliance over 1114 or 2056. This comes back down to the ceiling issue, as pointed out before.

This begs the question "Why would anyone build a robot that cannot compete?" At the beginning of the seasons, if you looked at previous years robots and games that involved shooting, you could have make a fair argument that shooting is hard to do and would not be all that accurate unless you are a high level team. This played out to a certain extend this season, but the floor for shooting accuracy was much higher than it was in previous years (i.e. a 50 percentile shooter this year would score about 8 disks, and 12 points on average in autonomous, while a 50 percentile robot from 2011 is probably only good for points in autonomous or a handful of balls in the 2 pt. goal). The other answer to this question is that for many teams, there is something inherently fun and awesome about taking on a cool or unique challenge. It was a great feeling to finally see our robot pull off the wombo combo, and seeing it do it each and every time afterward seemed to be sweeter and sweeter.

From a game design standpoint, it would have been really cool for the upper parts of the pyramid to have been worth more points, bringing the climb and dump robots into the fold a little more. That being said, I have no regrets with our team’s decision.

[Brandon Holley]

I think it all comes down to 3 points:

1. Climbing to 30 was really hard.

2. Climbing to 30 consistently was even harder.

3. Because of points 1 and 2, you are now in a high risk 'all or nothing' type of strategy.

The last point is something many teams overlooked. Shooting points came in small chunks. While those chunks were smaller than a 50 point climb, they were generally easier to attain chunks. A team could build to 50 points and beyond over the course of an entire match, instead of executing one maneuver.

To me for a dedicated climb and dump robot to be picked they needed to have extremely high consistency.


At Orlando we chose 4451 as our first pick from the 2nd seed (declining the 1st seed, and passing up several other shooting robots) because based on our scouting data, they had the highest probability of scoring 50+ points a match. The reason for this was their near flawless execution of just getting to the top every match. They did not disappoint us as they climbed to the top in all 9 of our elimination matches.



-Brando

[Andrew Schreiber]

Originally Posted by Brandon Holley

2. Climbing to 30 consistently was even harder.

In qualification matches I believe the number of matches with >30 climbing points (ie all three bots climbed 10 or 1 or more climbed above 10) was less than 20% all season. (Ok, I know it was less than 20% but I can't recall the exact number my data showed)

I have some other theories but I need to think of how to word them...