If anyone hadn't seen, the Reddit April Fool's thing this year is a game where you vote on one of three images, and the goal is to vote on the item with the middle number of votes. If you do, you gain 3/6/9 points (based on how soon you vote). If you pick the most or least possible, you lose 1/2/3 points (based on how soon you vote). Every 15 seconds it gives a quick preview of the current votes to give you an idea of where people are voting. You can't change your vote, so if you go early you risk more points but can't gauge other votes as well.
If you guess about 1/3 correct, and you get +3 points for corrects and -1 for incorrects, I think you average a net +0.33 points per guess over time?
I think I figured out a pretty solid technique, and I set up a quick Python script to automate it overnight to see how it goes. I've noticed that if there's a clear loser after the first preview of votes, that item never catches up at all. The other two are always a pretty clear neck-and-neck tie due to the feedback influencing late votes, and it's common for the vote split to be like 1000 vs 1002. I've been picking whatever has the most amount of votes after the first reveal, since we know the thing in last is out of the running, so it's pretty much a 50/50 where you get +3 on correct and -1 on incorrect, which I think is +1 point on average per turn.
I just have Pillow taking screenshots every couple seconds, Pytesseract parsing the votes as numbers, and PyAutoGui clicking specific points. Hopefully I'll get some nice results in the morning. I'm tired and don't want to set up something to track the wins/losses so hopefully points will be enough
If you guess about 1/3 correct, and you get +3 points for corrects and -1 for incorrects, I think you average a net +0.33 points per guess over time?
I think I figured out a pretty solid technique, and I set up a quick Python script to automate it overnight to see how it goes. I've noticed that if there's a clear loser after the first preview of votes, that item never catches up at all. The other two are always a pretty clear neck-and-neck tie due to the feedback influencing late votes, and it's common for the vote split to be like 1000 vs 1002. I've been picking whatever has the most amount of votes after the first reveal, since we know the thing in last is out of the running, so it's pretty much a 50/50 where you get +3 on correct and -1 on incorrect, which I think is +1 point on average per turn.
I just have Pillow taking screenshots every couple seconds, Pytesseract parsing the votes as numbers, and PyAutoGui clicking specific points. Hopefully I'll get some nice results in the morning. I'm tired and don't want to set up something to track the wins/losses so hopefully points will be enough
Please Login To Post
It looks like the top players have an average of +2.6/round assuming they played every single round since it came out, or 3/round if they started 150 rounds (~2.5 hours) after the game launch. I need to figure out their strategy.
I'm considering changing it so that it only picks between the higher of the left or right, but I don't think that supplements my strategy very well. If the descending order of votes goes left, middle, right and people are intentionally not picking middle, then there's no way that the left one can win second place, and it still doesn't seem that the lowest vote gains more momentum. I'm afraid that if middle is in second place during reveals, then it might stay second place for the rest of the round, so only going left or right might be harmful