The Highest Levels of Play

[Andrew Schreiber]

Originally Posted by Kevin Leonard

But I think that's part of the problem for me: I'm trying to look too deeply into a game that we don't have that much time to play. The game isn't going to evolve the way other sports do over years, it's going to evolve over 8 weeks of competition (and evolve immensely), and will be limited by size restrictions, power restrictions, etc, not just by player skill and ingenuity.

Bingo. Teams get better, but overall they start out pretty bad. So, when I make my spreadsheets I also do run through a "what's likely" analysis too. My general rule? If I think an action will take 3 seconds I multiply it by pi because I'm an idiot and tend to overestimate skill. (Or teams just generally suck)

I also do what I call sensitivity analysis: at what point do strategies become worthless because you're taking too long to do them?

[logank013]

Originally Posted by Andrew Schreiber

You're mistaking important in an instance with important from a strategy perspective.

Ok. i thought you were just being general. But from a strategy, yeah, useless.

[The other Gabe]

The first reason for why alliances made are not always as good as they should be is bad scouting. whether it's because the team is biased, they don't have enough people to scout well, or their scouts just aren't accurate, this can cause them to create a poor alliance (or neglect to pick good robots... cough cough my team at DCMP the last two years cough cough)

as for gameplay at higher levels, not every team has the resources to create a robot competitive at the highest levels. another issue can be mis-analyzing parts of the game- members of my team, including myself, though that the totes would fall flat from the feeder of their own volition, and therefore didnt make any sort of ramp. this forced us at our first divisional to quickly change to landfill, which was quite an adjustment for the drive team. many other teams made similar mistakes, and later fixed their bots with ramps, full redesigns, etc.

also, to add onto andrew schriebner, 2 of the best PNW feeder bots this year, 955 and 4450, had first divisionals that were... less than stellar compared to their level of play at the end of the year. their strategies also changed completely; in playoffs, 955 was a capper, and 4450 was a landfill bot in their first competitions. both teams added ramps and became very good. nearly every bot from my division that made it to worlds this year improved an extent, but those two were the most radical changes.

[Lij2015]

In 2013 I was determined that full court shooters were the end all be all, and while teams like 148, 67, 303, and 469 were amazing bots, the championship was won by lightning quick cycle bots and amazing defense.

In 2014 I laughed off blocker poles(and this was pretty much spot on until Einstein)

And in 2015 I expected to see WAY, WAY more can specialists at regional events, not only because that function only required one mechanism but they were in seriously high demand among the lower seeded alliances at both Chesapeake and Virginia.
The only really top notch ones I saw at our two regional events were 540, 2537, and 2377.

[wesbass23]

The highest level of play in any modern FRC game minus Recycle Rush is being able to adapt to what your opponent throws at you and what sort of alliance you are playing with. A good team can carry most alliances on their own and beat most other alliances. A great team can utilize their alliance members no matter their skill level and together beat any combination of good teams.

[Abhishek R]

Originally Posted by wesbass23

The highest level of play in any modern FRC game minus Recycle Rush is being able to adapt to what your opponent throws at you and what sort of alliance you are playing with. A good team can carry most alliances on their own and beat most other alliances. A great team can utilize their alliance members no matter their skill level and together beat any combination of good teams.

Exactly this, word for word. That's what the highest level of play is.

[pabeekm]

Sorry if I’m reviving a settled topic… but I feel the need to defend death cycles (warning: super biased perspective incoming) . It wasn’t so much that death cycles were hard to build for (our robot was painfully simple: catapult + wheels, no pickup required) or hard to qualify for champs with. I would think obscurity killed the concept more than anything. And even so, it almost happened! A 900+1918 alliance totally could have been made in Archimedes (single tear… If only we had been the 8th alliance’s second pick. << come to think of it, that’s a very rare kind of statement in FRC).

I’d argue that the component bots of death cycles were plenty competitive on their own. At defense-heavy regionals, an undefendable trusser might have been a huge asset for the top scorers (pure speculation from a North Carolinian here). Scoring from in front of the low goal in and of itself was also of great value for some excellent high-gaolers.

Case in point: When 900 saw the trussing component of death cycles, suddenly no other strategy held a candle, even though we were aiming for regionals, not champs. We weren’t even thinking about the ultimate partners because, well …. North Carolina (no offense to NC teams, please don’t hurt me!); the benefits to cycle speed of truss to human player alone were enough to convince us.

I would think the digging required to figure out death cycles is what kept the components from popping up, and, by extension, the concept’s fulfillment; teams would need to have realized that defense would be killer, cycle speed would be super important, and that human players could catch truss shots. Most teams willing to be that observant were, by no coincidence, great teams anyway, and so they didn’t need to pursue niche roles.

What I’m curious about is if the possibility/value of death cycles had not been so obscure (i.e. the rulebook said plainly “you can totally throw over the truss straight to the human player” and/or the animation video warned that plowie could kick the crud out of guys on the field) if things might have been different. Do you all think more teams might have pursued the components of death cycles, or by extension, that death cycles might have been more or fully realized?

[Kevin Leonard]

Back on topic- I think a better title for this thread would be- "How do you get the game right early on?
How do you figure out what robots you'd see on the World Championship and IRI Championship alliances?

I'll use this year as an example:
I believed at the beginning of build season the highest levels of play would compose of: Landfill Stacker, HP Stacker, Capper.
No alliance would ever end up with more than 7 cans, which I believed a high-tier capper could manage to place all of, or at least most of. I figured the best human player stackers could make 4-5 uncapped stacks, and the best landfill stackers would be able to do the same.

What I failed to see is that while that could be effective in theory, so could three robots building stacks underneath cans.

When running human trials, we tried building stacks underneath cans and had the cans fall over and the humans couldn't carry the stacks well.

How do you account for the difference in what robots are able to do vs. what humans are able to do? Robots do some things faster than humans, while humans do some things faster than robots.

[Andrew Schreiber]

Originally Posted by Kevin Leonard

Back on topic- I think a better title for this thread would be- "How do you get the game right early on?
How do you figure out what robots you'd see on the World Championship and IRI Championship alliances?

I'll use this year as an example:
I believed at the beginning of build season the highest levels of play would compose of: Landfill Stacker, HP Stacker, Capper.
No alliance would ever end up with more than 7 cans, which I believed a high-tier capper could manage to place all of, or at least most of. I figured the best human player stackers could make 4-5 uncapped stacks, and the best landfill stackers would be able to do the same.

What I failed to see is that while that could be effective in theory, so could three robots building stacks underneath cans.

When running human trials, we tried building stacks underneath cans and had the cans fall over and the humans couldn't carry the stacks well.

How do you account for the difference in what robots are able to do vs. what humans are able to do? Robots do some things faster than humans, while humans do some things faster than robots.

Experience.

[EricDrost]

Originally Posted by Kevin Leonard

Back on topic- I think a better title for this thread would be- "How do you get the game right early on?

Kevin, I was asking similar questions not so long ago.

Like Andrew, I usually use a spreadsheet to match actions with their scoring values and predicted times and difficulties to find what I believe to be the optimal strategy for my team. The priority order I usually use in picking designs is:

1. Win matches.
2. Seed high.
3. Desirable to partners.

Depending on where a team sits on the "food chain", this priority is likely to shift.

I originally did not intend to post in the thread, but I felt like I learned two important lessons since I made my post in the 2010 thread.

The first lesson may seem like a cop out, but what has been most valuable to me personally in analyzing games is simply experience. The more games I see, the better handle I get on what is a "realistic" way every new game will play out. Each year I've been able to get a firm grasp of how match play will look earlier and earlier in the build season because I can relate it to previous games. This has been very helpful in guiding students to the right strategies, and asking them questions that deepen their understanding of the year's challenge.

The second lesson is to never accept your understanding of how the game will play as final. This applies equally during build season and competition season. There have been years where I've analyzed the game, had an excellent understanding of it for week one, and then failed to update my analysis. The best understanding of a game comes from constantly questioning, how does my strategy hold up under X, Y, or Z circumstance. A great way to do this is during build season is to find friends on other teams and compare their strategies to yours. During competition season, I find it easier to watch as many webcasts as possible and figure out how to react to the current metagame. For example, if you couldn't pass back to the human player by 2014 championship, you were at a real disadvantage. Strategy should constantly evolve as you learn new information -- getting your strategy perfect early on is not as important as adapting quickly and getting it perfect every match.

[Andrew Schreiber]

Originally Posted by EricDrost

1. Win matches.
2. Seed high.
3. Desirable to partners.

I'd have to disagree with this and would instead suggest:

1. Seed high
2. Win matches
3. Desirable to partners


While typically, winning matches is the key to seeding high it is not always the case. Reading the manual and understanding HOW to seed high is incredibly important to controlling your own destiny come alliance selection time.

[marshall]

Originally Posted by Andrew Schreiber

Reading the manual... is incredibly important

Fixed that for you.

[Deke]

I would add playoff versus qualifying strategies as well. Recycle rush had elements that were different in qualifying and playoffs with co-op and can races. At our second event, we realized that we made more of a qualifying robot than a playoff robot. We seeded high at every event in position to be a captain, but failed to bust into the finals losing in the semi finals all year (except MSC we survived octo and were eliminated in quarters).

After going through this season, I would say it is better to design a playoff bot that might seed lower and contend for the finals, than a qualifying bot that just can't quite cut it in the playoffs. This was a different element than previous years IMO.

[marshall]

Originally Posted by Infinity2718

I would add playoff versus qualifying strategies as well. Recycle rush had elements that were different in qualifying and playoffs with co-op and can races. At our second event, we realized that we made more of a qualifying robot than a playoff robot. We seeded high at every event in position to be a captain, but failed to bust into the finals losing in the semi finals all year (except MSC we survived octo and were eliminated in quarters).

After going through this season, I would say it is better to design a playoff bot that might seed lower and contend for the finals, than a qualifying bot that just can't quite cut it in the playoffs. This was a different element than previous years IMO.

This actually brings up something that is on my mind right now. When does it become necessary/advantageous for a more average team to build multiple robots to compete at both a regional/district level and at the championship level?

[Ben Martin]

Originally Posted by marshall

This actually brings up something that is on my mind right now. When does it become necessary/advantageous for a more average team to build multiple robots to compete at both a regional/district level and at the championship level?

When you are willing to commit to the almost-every-day-after-build-season schedule that you need to get your value out of the 2nd robot.

Since the second robot is a huge time and money sink, to get the full value out of it, you need to spend a lot of time practicing with it and upgrading it. There are many teams that build the second robot, but don't execute on it right (they don't behave the same, or they don't commit to the practice schedule required to truly get mileage out of it).

I believe that teams that don't build practice robots should focus on strategies with low movement that still score good points, with a focus on minimizing the impact of driver error and on game piece control in their robot design.

In regards to the priority discussion---

Our priorities in 2013/2014 were these:
1. Win local competitions
2. Do well at world competitions

For 2015, it was:
1. Get to Einstein
--though priorities from the past years factored a lot into the design.

We always try to build robots to be the #1 pick for elimination rounds and try to include all the features that might make someone want to pick us (we had a goal to have the fastest can grabber in MAR this year, in addition to wanting to have a 3-tote auto to seed high and to score points from the landfill, since we figured it was the harder task to do and would thus be more desirable at high levels where there would be a ton of feeder station bots).