Implications of Autonomous Extension Election

While setting up some simulations to create statitics for a possible future game, I have to make some choices in the experimental setup. Assume:

Some future game consists of the standard 3 scoring periods:

  1. 15 second Autonomous
  2. 1:45 Standard Scoring period
  3. 30 second End-Game Change, overlapping the Standard Scoring period

Before Autonomous begins, Red Alliance may elect to extend their Autonomous period by 10 seconds. If they choose not to do this, Blue Alliance may then make that election. It is possible and allowable that neither alliance will extend.

During Autonomous, if one of the alliances elected to extend their Autonomous Period, their robots will continue to function for an additional 10 seconds while their opposing alliance robots are disabled by the field controller. At the conclusion of Autonomous (which may include the extra one-sided 10 second extension), the winner is awarded Autonomous bonus. Driver control then resumes.

This is obviously an early advantage to the alliance that chooses to extend. The price for this advantage is the “even-up” period. For 10 seconds at the end of the game, the alliance which elected to extend autonomous will be disabled, while the opposing alliance is free to score.

Assume that the standard scoring of the game is Zero Sum. That is, scoring for Red Alliance implies descoring for Blue Alliance. (Here, “standard scoring” refers to those points awarded for things other than Autonomous bonus and end-game bonus, like parking, elevating, or hanging.)

Now, the questions:

  1. Would you ever elect to extend?
  2. If so, under what conditions? Examples would be “Early in the season to see how it works.” and “If autonomous bonus is 5% of the available max score” or “If scoring objects are limited;” things like that.
  3. Would your answer change if the “even-up” period happened 40 seconds before the end of the match? (The point 10 seconds before the End-Game change. This year, the end-game change is adding a cube. Also, it’s implicit that hanging has to come late in the game, so that’s a kind of End-Game change as well.)
  4. Would your answer change if the even-up happened 10 seconds into the End-Game change. That is, it would start with 20 seconds left and end with 10 seconds left. This year, you’d then have 10 seconds to throw objects back, or to go hang.
  5. Would your answers to any of the questions change if the “even-up” period was not 1 to 1 but 1 to 1.5? That is, if 10 seconds early cost you 15 seconds late? How about 1 to 2, so that it cost you 20 seconds?

Let me know what you think.

Don’t have much time to answer this, but I would always choose to not extend the autonomous. Usually, the robots end up misaligned in some way at the end of autonomous because of interactions with partners and opponents or just mistakes in programming. Thus, it would be hard to do an extra ten seconds anyway. Also, those ten seconds of user control that you lose are very crucial. It would ne hard to overcome the points the other side scores in those ten seconds. Besides, 10 seconds user control will score more points than 10 seconds autonomous.

I am going to have to agree with @Bryon Tjanaka here, unless the autonomous bonus was significantly higher (I would think x3 or more), it would not be worth it. 10 seconds at the end of the match is a very long period of time, and the tables can really turn in that time. If you where able to get ahead far enough, and set yourself up in a way where you could prevent the other team from scoring then it might work, but I just don’t see it being used very much if it where implemented.

If you where able to pick whether or not to extend your autonomous after your 15 second autonomous ran, then it might be more lucrative. If you and your partner’s autonomous ran with no issues, and you where set up to score more points, it may be worth it to continue with autonomous even though you loose the last 10 seconds.

My biggest issue is that the proposal violate the K.I.S.S. principle. I am NOT in favor of these kind of complications.

I would really just rather see a full 30 second autonomous period across the board.

I think what you’re really saying is “In the past, it has been too hard to determine where your robot is on the field, and where it is in relation to game objects and other robots, and to react to the changing field condtions.”

Because if you could more easily accomplish those things, you could more easily code a reactive autonomous routine.

Consider a game where hoarding game objects confers a large benefit. What if you used those 10 seconds to gather the objects, so that there were many fewer objects to score with at the end?

I certainly am an adherent to the K.I.S.S principle. However, I think there’s evidence the GDC has been trying to spur more teams into writing better code. This would be another way to do that.

I would say it depends on the game. A game where common objects cannot be de-scored (like NBN) then an extra autonomous might not be recoverable and the field is clear by matches end anyway. A game like this year, 10 seconds at the end can decide the match.

I’ll answer question 1, then address every scenario I can think of that would make my answer change.

I’ll assume that this zero-sum game is similar to Starstruck.

  1. No.

Under pretty much any circumstance, 10 seconds of driver (on a clean field) is necessarily either as good as or better than autonomous. You can run auto code and have a manual override if something goes catastrophically wrong, making driver necessarily as good as or better than auto. If the game is zero-sum, no object is permanently scored (is this a logical deduction? I think so, but I’m not sure), and thus any lead I gain in auto can theoretically be surmounted.

Scenarios in which my answer changes:
+I become a decent programmer.
+The auto bonus is so large I would have an insurmountable lead.
+Once scored, objects are very difficult, although still possible, to descore and rescore.
+My robot strategy relies on winning auto and preventing the alliance from scoring (e.g. a wallbot). With a wallbot (for this game, anyway), if I won autonomous and raised my wall, it wouldn’t matter if we were disabled for 10 seconds.
+There are very few game objects, and they are up on a platform which it would take 20 seconds to reach. Thus, I grab the few objects, sit for 90 seconds through the even-up, and then score them afterwards.
+Entanglement and/or trapping become legal, and I trap both opponents in a net.
+The game requires immense precision which cannot be replicated by a driver (e.g. NbN flywheel speeds, not a perfect example, but it works)
+Drivers are blindfolded as part of the game.
+The game is not zero-sum, and thus I can permanently score 51% of the field.
+As a challenge, the RECF implements a 1 second latency into the controllers; any input from the joystick is received by the cortex 1 second later.
+To generalize a few of my scenarios, any scenario where the rules tremendously handicap the drivers.

That’s all I can think of.

I certainly concur with the GDC focusing more on code. That was behind both the move to Autonomous points being the first tie-breaker and the combining the Driver’s Skills score and the Programming Skills Score into a single score. However, I think one has to keep in mind (in my opinion anyway) 1) the game needs to remain accessible to many different levels from middle school to pretty advanced high schoolers and 2) the spectators. Make the game play too complicated and you make it far more difficult for young or inexperienced teams as well as very difficult for spectators to follow the game. My least favorite game over the years has been Gateway largely because it was very hard for a spectator to follow the action and get a sense of who was winning (or losing). And I would not be adverse to a longer autonomous (maybe), but to me, the above scenario is just WAY too complicated. One of the things I have appreciated about the games put out by the GDC for VEX is the fact that they could be challenging from a robot stand-point but easy to follow with no complicating penalties or rules that are overly confusing (as I believe the above is).

Maybe not quite that long, but something like that(at least 25 seconds) would be soooo fun and challenging. It does kind of depend on the game though; nbn couldn’t have had that. It would have been interesting to have that extra time in a game like toss up though.
Edit: it would leave a lot more room for autonomous Improvement during the season as well. At this point in the season and for quite a while now most people just have close to the same autonomous :frowning:

My biggest issue with this proposal is assuming 10 seconds of autonomous == 10 seconds of driver. Its a bad trade. Maybe a 10 extra seconds of autonomous is 5 seconds extra in driver.

but i agree with this statement the most:

15 is kind of pathetic when we are trying to teach kids the value of computer programming.

What I’d really like to see is a driver abort button for autonomous. it would save kids from thier riggs tearing themselves to pieces and when teams dont have an alternative autonmous they could just abort it instead of forcing the better teams to change from their (often times) better routine or any number of overly complicated solutions.

Thank you! That’s very helpful. I may have questions after I digest this.

I agree with that. I’m often in the position of having to explain game mechanics to other parents, grandparents, etc. I think it would be helpful to have some introductory material. Things like “How this game works” and “What you should watch for” and “Here’s a vocabulary cheat-sheet.” They have those video features in Olympics telecasts. It’s the only way to make, for instance, short track speed skating understandable. I wish the online challenge videos would ask for a “So, your grandchild is in a VEX tournament, and you’re here to watch…”

That said, the current situation is difficult to explain. Not sure how much worse this would be, but it would be worse to explain.

To keep tournaments managable, you need to stay close to the same overall time. So, longer autonomous probably means shorter driver. (This new scheme takes that into account as well, it looks like.)

I agree with you in principle. But I disagree that the recent games have been easy for a new viewer to undrestand. It’s easy for us because we’ve seen so many, and we know the general scheme. So we have an outline in mind. Context. With no context, it’s difficult to figure out. But you’re correct when you say this would make it worse.

Both teams have the option to extend their autonomous. These decisions are made independently of each other. If neither teams extends, match is played as normal. If both extend, for 10 seconds in driver during even-up no robot moves. If one extends, they get the extra auto at the cost of losing the even-up period.

Thank you for your comments. As part of our events here in Wisconsin, we make every effort at all of uor events throughout the year to produce a one page game description and team roster that spectators can pick up upon entrance to the event as well as our announcers giving a run-down of scoring, etc. particularly early in qualifying, but we try to remind them to do it at least one more time later in the day. The OP’s scheme would be much more difficult to explain plus what a headache it is going to give to DWAB trying to modify TM to fit this. Hopefully see you at Worlds, would be good to put a face to a name.

That is an interesting idea from a game theorist’s point of view, since it turns it in to a one-round Prisoner’s Dilema. The advantage (for a person who studies games) is that there is a very rich background of research into PD in its various forms.

In fact, upon reflection, I realize that bringing any adversarial game-affecting choice into VEX RC would make it a fantastic laboratory for game theorists.

Oh my. What a grand experiment it would be. They’ve set up a system that encourages, even requires, that all events report outcomes of every meeting of participants. The system assigns teams togehter in alliances in a way that leads to maximum data extraction from the available time. It does this by arranging that, to the degree possible, alliance pairings are not repeated in a tournament. The system also offers a constrained “free choice” to participants in alliance selection. A game season lasts for nearly a year, so there are multiple datasets to examine. And, the data is gathered from regions accross the world. This allows examination of cultural biases. Further, there is an arranged cross-cultural meeting, so that teams from the different areas/cultures play with and against one another. This means datasets can be normalized, those cultural biases and tendencies studied, and tools created to adjust for those factors in the data.

And, the game itself can be altered year-to-year, so that effects of particular factors can be isolated, amplified, or eliminated.

Is someone at REC F working on a PHD in game theory, philosopy, logic, or mathematics?

As was said in another post, it could be that:

@Karthik is playing games with us.

That is a very good set of ideas. The VRC event partner resources should include a writeup like that, and forms for creating an event-specific program. Maybe I’ll develop a tool to do that sometime this summer.

I’m definitely hoping to meet you and some others at Worlds. I know I’ll be there, since it’s less than an hour from my house. Currently planning to be there everyday.

I kind of want to make this a video now… Wouldn’t be high video quality but at least it would work as an explanation. But I have State tomorrow, so definitely not a high priority.

It would be a fantastic resource. Just remember and do it again next season, too.