top of page

Avocado vs. Cucumber: AbstainTrain


Game 3 of Avocado vs. Cucumber ended yesterday. I was really happy with a lot of how it went. And then, well, there's the abstain issue. Check out StuFLaW as well.


The Abstain Option

In the previous post about this game, I talked about the creation of the abstain option. It was a fix to a problem where some players had nothing they could choose that would allow them to win the round, and it feels bad to have no way to win. So I created a third option, mainly to satisfy people with this issue. It didn't actually ever need to be picked to work. The important part is that a player could abstain.


But in the back of my mind, I was thinking about a scenario I call AbstainTrain. AbstainTrain occurs when people started abstaining en masse; having very few non-abstentions would make abstain overpowered since abstain would virtually always win. People would have no choice but to get on the abstain train. This destroys the game by effectively eliminating players' choice in voting.


Obviously if the number of people playing Avocado vs. Cucumber ever got super high, then it would be stupid to ever abstain. After all, the chance of it coming down to a single vote is terribly low. So all that needed to happen is AbstainTrain needed to be avoided until then.


A Japanese bullet train. Image credit: BBC.

The AbstainTrain Scenario Occurs

In the most recent game, AbstainTrain started happening. A hypothetical player who abstained every round would have earned 4 points out of 5. (This figure for Games 1 and 2 is 1 and 2 points respectively.) By Round 5, more players abstained than voted for avocado and cucumber combined. AbstainTrain was happening.


In my previous Avocado vs. Cucumber piece linked at the top, I advised players that they "shouldn't abstain." However, if I were playing in Game 3, I would have abstained in several rounds for sure. That's not good.


Possible Solutions

One solution to this problem for future games would be to remove the abstain option entirely. After all, isn't abstain the source of this calamity? Well, mosquito bites are extremely annoying, but mosquito larvae are an important food source for some species of fish. Similar to the case of mosquitoes, abstain plays an important role: making sure there's something each player can vote for to win.


The art of the Magic card Prophet of Kruphix. Art credit: Winona Nelson for Wizards of the Coast.

Well, a player that abstains wins in each of three scenarios:

1. Avocado is ahead by 1 vote.

2. The vote is a tie.

3. Cucumber is ahead by 1 vote.


Introducing: Predict

Predict simply splits this up into three distinct options:

1. Predict: A+1. In this scenario, you win if avocado is ahead by 1 vote.

2. Predict: Tie. In this scenario, you win if the vote is a tie.

3. Predict: C+1. In this scenario, you win if cucumber is ahead by 1 vote.


So, now abstain is eliminated, and so voters have five choices: avocado, cucumber, predict: a+1, predict: tie, and predict: c+1.


Will This Solution Work?

One of the features of AbstainTrain is that if a couple of people are not on the abstain train, then abstain is still very likely to win. For instance, 3 people didn't abstain in Round 5. If 3 or fewer people choose randomly between avocado and cucumber, abstain wins about 70% of the time.* However, that number for predict: tie is only 14%, and predict: a+1 and predict: c+1 are each at 28% each. Therefore, PredictTrain is extremely unlikely to occur.


Isn't Predict Underpowered?

Yes, that's the point. Don't ever predict. It's like insurance in blackjack: it's a bad bet, but if you're frustrated that the dealer has a natural 21, then they can say "you could have taken insurance!"


A crystal ball. Image credit: Amazon.

Now, I'll move a cacophony of other Avocado vs. Cucumber-related things.


The Style Section

When I ran Avocado vs. Cucumber via Hangouts, I awarded players style points for doing something that is

(a) creative or "stylish," for instance, voting with emoji, and

(b) unique to that player.


They were ultimately a side thing that was just for fun.*** I stopped them when I switched to Forms since I was worried that it would devolve into "Who can come up with the coolest phrase to put in the "additional comments" box. However, while a survey I conducted suggested that most players don't care about style points, those that did were in favor of them, so I'm re-adding them to the game.


Google Forms to Google Sheets

Avocado vs. Cucumber moved to Google Forms this game, and it worked out really well. Therefore, if you want to play in future games, and you're not already a frequent participant, just give me your email and I'm happy to send you links to future games! I suggest the Wix chat function for this, but anything that gets to me (email, Hangouts, text, conversations, etc.) works fine.


A random screenshot from tinyurl.com/avcrawdata.

Another byproduct is that since Forms can feed into Google Sheets, it's possible to automate the vote-counting process.** This is pretty cool, and while it led to a mistake on my part because some votes were counted twice by accident, when used as a tool but not a crutch, this can be extremely useful. It can also help to bring the game to a large-scale in the long term. So that's exciting.


Greed Control

Someone mentioned that Avocado vs. Cucumber is similar to Art of Problem Solving's Greed Control.**** This actually makes perfect sense, since in both cases, if a strategy is too popular, it's no longer a good strategy. Also, it's important to look at past rounds to inform your idea of the future. Greed Control is super neat.


Strategy Idea: Use the Past

The strategy that I would play, by the way, is to play whatever would have won the previous round for me. (And never predict.) This simple-sounding strategy has a remarkable hit rate of 70%. Though the sample size is fairly small, and I just shared this strategy with everyone, I would actually play this strategy.


Anyway, what do you think about predict? The return of style points? The switch to Google Forms? Tell me. I hope you enjoyed this post.

–beautifulthorns


*The possible scenarios are:

(a+1) a

(c+1) c

(a+2) aa

(tie) ac

(tie) ca

(c+2) cc

(a+3) aaa

(a+1) aac

(a+1) aca

(c+1) acc

(a+1) caa

(c+1) cac

(c+1) cca

(c+3) ccc


**The COUNTIF function is very useful here.


***If style points aren't fun, ignore them. Heck, if points aren't fun, ignore them. There aren't actually prizes for this game.


****The rules are:


"The game of Greed Control is based on a simple forum game. Every day, pick an integer number between the numbers shown on the game page. For example, a number between 1 and 50.


At the end of the day, which is defined as midnight eastern time, 9 pm pacific, a calculation is performed to determine how many points you earn. The number of points is defined by (number picked) / (number of people choosing the number).


For example, if 4 people choose the number 20, each will receive 20 / 4 = 5 points. If only 1 person chooses the number 9, that person will receive 9 / 1 = 9 points. Notice how even though the person who chose 9 picked a number less than 20, that person still gains more points as the number 9 was picked less.


The current game is set to end when someone reaches 2000 points."

38 views

Related Posts

See All

Comments


bottom of page