Hi y'all,
I made an account on here just for this. I read this past week about a swarm AI project that used the combined brainpower of groups to make more accurate predictions than the individuals in the group could make. Specifically, this was the system that predicted the quad for the Kentucky Derby and the trifecta for the preakness. It also did stuff like calling 75% of oscar winners correctly.
Since I read about that a couple of weeks ago, I've been thinking about potential applications for sports betting. Are all of us smarter than one of us? It's hard to tell...but given that this is the first time this tech has ever been available to play around with, we could be on the forefront of something that is, at the very least, interesting, and at the very best, exceptionally profitable. The more contributors, the better the system seems to work.
This is where you come in. The creators of this swarm AI project have made it publicly available at unu.ai. Account signup for the beta is free and doesn't require any personal info. Once you can access the main page, you'll see a list of rooms in the top-left corner; look in the 'Featured Unus' box for UNU Sports. That's where I'll be hosting this experiment until the beta unlocks user-created rooms. Up to 100 people can participate at any one time.
You'll see, once entering the room, that there is a hexagonal display with a rotating circle in the middle. If you mouse over it, your cursor will turn into a magnet that will influence the movement of the circle towards it. The idea is that the closer you get to the circle, the bigger your magnet appears and the more influence you exert. In other words, the bigger your magnet, the stronger you feel about your decision. Each of the six points on the hexagon will represent a potential answer, so whichever of those (up to six) potential answers you pick is where you'll attempt to guide the circle. It sounds weird, but it's very intuitive when you're actually doing it. On the left is a text chat where discussion can occur simultaneously, opinions can be swayed, etc.
What we will be working to do is to pick the best underdogs of each day in a manner as similar as possible to the highly-successful applications of the technology so far. We do not want to over-complicate the process by introducing expected value into the equations, we only want to answer a simple question: Who Will Win?
We'll achieve this by process of elimination. Therefore, the first sets of questions will be: which of these 6 teams is LEAST likely to win tomorrow? Each chosen team will be eliminated from our pool. Once only 3-4 teams remain (depending on the number we started with,) I'll then switch the question to ask which team is MOST likely to win, 2nd most likely to win, etc, until we have ordered our 3-4 teams. Then, we simply track the results, knowing that anything over a 50% success rate will mean profitability.
The whole activity should take about 15-30 minutes, max. The data generated will of course be yours as much as it is mine.
If you're interested, I'll be hosting the first session at 6:30 eastern standard time. I would really appreciate your participation!! If you'd like to get in on it, here's the list of teams we'll be drawing from. Remember, we're not looking for 'best value.' We simple want the most likely to win.
LA Angels
Baltimore
Cincinnati
ChiSox
NYM
Miami
I'll also be here answering questions.