Alphago will win over Lee Sedol at least 4-1
Created by sole21000 on 2016-03-09; known on 2016-03-14; judged right by PseudonymousUser on 2016-03-15.
- sole21000 estimated 96% on 2016-03-09
- moridinamael estimated 95% on 2016-03-09
- PseudonymousUser said “Even if Lee Sedol’s per-game win probability were as low as 56%, the probability of AlphaGo winning >= 3 of the 4 remaining matches is only 0.44^4 + 4 * 0.44^3 * 0.56 = 0.228.” on 2016-03-10
- PseudonymousUser estimated 44% and said “But I’ll be cautious in my estimate, and average some of my other guesses in.” on 2016-03-10
- awesomemathlete estimated 85% on 2016-03-10
- PseudonymousUser estimated 76% and said “Updating based on the 2nd game, which is currently in-progress.” on 2016-03-10
- PseudonymousUser estimated 86% on 2016-03-11
- artir estimated 99% on 2016-03-12
- PseudonymousUser estimated 85% on 2016-03-12
- PseudonymousUser estimated 90% and said “Based on the third game’s result” on 2016-03-13
- artir estimated 100% on 2016-03-13
- Raahul_Kumar estimated 100% and said “Not sure if this is a flaw with reinforcement learning, but by now AlphaGo knows how to counter that mistake.” on 2016-03-13
- PseudonymousUser said “AlphaGo is frozen and is not being trained between the games in this match, so it has not specifically learned “how to counter that mistake”” on 2016-03-13
- PseudonymousUser said “After Lee Sedol wins one game of four (win rate = 25%), you put down 100% on Lee Sedol losing the last game?” on 2016-03-13
- nshepperd estimated 82% on 2016-03-13
- themusicgod1 estimated 51% on 2016-03-13
- JoshuaZ estimated 75% on 2016-03-14
- mortehu estimated 80% on 2016-03-14
- PseudonymousUser estimated 85% and said “Updating based on the 4th game’s result, and my own cognitive biases.” on 2016-03-15
- Houshalter estimated 55% and said “Just a basic estimate gives 66% probability to AlphaGo, and that’s not even factoring additional information, like Sedol having improved.” on 2016-03-15
- PseudonymousUser judged this prediction right on 2016-03-15.