Wanted to post a few thoughts before we get started with the season tomorrow. Unlike most years, we get to have seen a full competition play out before week 1 has even begun. Let’s use that to our advantage.
Palmetto
What happened at Palmetto? Breaches happened about half the time, and captures were very rare (didn’t occur at all until playoffs). The average team scored about 13 defense points per match. Let’s look a bit more in-depth. Here’s a chart of quals breaching % versus defense points per match:
You can draw a lot of conclusions from that chart. First off, I find it impressive that even the lowest scoring teams were averaging most of a breach (kudos to the GDC on a very playable game!). The next thing I’d draw your attention to is the teams at or above 45 defense points per match. These teams were getting breaches in about 3/4 of their matches. Even really good teams can’t guarantee breaches - 179 was excellent, but still breached in 6 of 8 matches.
Michigan is a game of district points, and the easiest and most consistent way to get points is through seeding. Many teams will ask themselves “how can I maximise my seed?” Well, here’s how teams at Palmetto did:
I’ve checked the correlation of points scored per match by method with ranking points accrued per match. They tell us something that seems obvious, but isn’t always true: the most common scoring methods had the strongest correlations with ranking. Defenses were about half the points scored, and that had the strongest correlation with ranking. Auto scoring was more correlated with ranking than you’d expect from its point output, but having autonomous scoring says a lot about a team’s preparedness. Neither scaling/challenging nor goal scoring proved an effective predictor of ranking.
However, these are past conclusions. Maybe goal-scoring is a poor method for ranking well, maybe there weren’t effective goal-scorers at Palmetto. I would expect goal-scoring in particular to improve significantly (in volume and effectiveness) as the season goes on. Defense points will never correlate as well with ranking as it does this week. Scaling won’t make a huge difference in rankings at any point this season, but it’ll make the difference in many elimination matches once captures become more common (for now, it’s something to do for practice).
Low Bar Bots
This has been a hot topic on Chief Delphi all season. How many teams are going low? Ugh, why so many? Let me say right now, I’m just as terrified as Karthik (maybe more). The proportion of robots going low is way off.
Let’s make a comparison to 2012. At the highest level in 2012, an elimination alliance very nearly required two of its three members to be widebots to go far (there was only one 4334). Many teams didn’t realize this, while others overlooked the importance of triple balancing. The result was lots of longbots and lots of good longbots not playing saturday afternoon.
Now let’s take 2016. You mostly need one of your three robots to be low. And the low/tall dichotomy is pretty much spelled out in the rules. True, it’s for all of qualifications, but it’ll be hard for multiple robots to consistently use the low bar (not to mention that’s a great way to not breach). So mayyybe 50% of teams should go low. But because it was obviously incentivized, the actual proportion is a lot more.
I highly expect every top team in Michigan to go low. 33 doesn’t tend to go too far out of the box. 67 is struggling with the loss of Adam Freeman and I therefore would’ve guessed they’d go with a pretty conventional design. This has been all but confirmed in another thread (they’re making one of those all-in-one arms, which isn’t much of a surprise for anyone who saw them in 2012). 469 certainly goes out of the box at times, but usually as an afterthought (those cangrabbers last year, their entire upper structure from 2010).
The most prominent Michigan teams that have revealed their bots (2054 & 68) are both low. 2054 tends to get creative, and their 2016 robot certainly is, but they didn’t buck the low trend. 573 is low as well. 4003 is tall, so that’s cool. If I had to bet on one great Michigan team that went tall, I’d guess 1918 or 1025. But realistically, the chances any particular great Michigan team went tall are pretty low.
Defense Selection
Alright, so the low bar is definitely damaged. How do I wreak havoc on my opponents’ strategies? TBA comes in clutch yet again.
There’s two good pieces of information here: which defense was picked more often, and which team’s struggled with more. Let’s go category by category.
A: The real portcullis is much nicer than the one teams expected (a guillotine). I suspect teams would’ve picked the CdF more often if it was available all the time/they weren’t afraid it would break. It’s harder and less of a burden to look past (although, again, the real portcullis is easier to see through than the team version). Both success numbers will increase as teams realize that having an A mechanism is pretty valuable and a good way to stand out from the crowd.
B: Not really unexpected. Both the moat and the ramparts are harder than they look at first glance. I’m sure many teams said “2.5 or 3 inches? That’ll be super easy.” But it’s not! The moat might be harder to do, but I think teams take more time on the ramparts. Could rationalize choosing either.
C: With so many low bots, alliances can pick one defense to not do. This was that defense. The most effective way to do Category C is working together, which doesn’t happen much in early weeks. Alliances that failed another defense and had to go for C had to work together under time pressure late in the match - that’s rarely a good proposition. Instead, I’d suggest teams do category C right at the start of a match. It can be done in seconds, less coordination is required because you’re choosing to do it right away, and you can take your time on the tougher tasks.
Also, teams picked the sallyport more. Because it hurts their viewing angles less. Great choice.
D: Turns out the rough terrain is really easy! Don’t pick it!
What should my strategy be?
We know breaches happen about once per match and captures almost never. That will affect your strategy.
Well, let’s ask who you are first. Look again at that Palmetto chart. You’ve got those top 6 teams in the top right, then you have a large group of above-average teams scoring 39+ defense points per match, then there’s everybody below that.
If you’re one of the top 6, a breach is mostly a given for you. If you don’t breach, it was probably a fluke accident. Once you’re certain you’re in that group, you should stop worrying about it and go for bigger goals. Since captures probably aren’t happening, go for the win. Maybe that means scoring fewer balls in the high goal (vs. more in the low goal) for more points, or maybe it means convincing your partners to play some legal defense.
Let’s say you’re in the second, “above average” group. You probably have mechanisms to do things other than go over defenses, and you want to practice them. Don’t do it until you have a breach. Don’t do it until you have a breach. If you play your cards right, you’ll be an alliance captain. You must guarantee the breach to do that. Getting over the hard defenses will help you to stand out while also maximizing your RPs.
Finally, the bottom group. Your goal is to play on Saturday afternoon, and you really don’t care where you seed. Find something you do relatively well, and do it a lot. Pick one or two categories you can do, and do them every time. Play defense if asked, because that’s a role that will be selected for.