Your watch is wrong for 1438 out of 1440 minutes each day, or 99.86% of the time, assuming minute precision. I haven't made a very high percentage of bad calls yet, AFAIK. Clever one-liners that are barely relevant to the topic really don't add anything to the conversation.
If you use milliseconds you can even make that 99.99% or higher!
The point is, that watch is verifiable accurate twice daily, you can look at it and will tell you exactly the right time. It will do so 365 days per year, which is 730 times. This is a very large number of accurate predictions without being useful at all.
The only kind of useful predictions are the ones that you are prepared to back with either money or deeds. Words alone really don't cut it, in fact, with words alone your accuracy decreases.
A good way to find out where people stand on a prediction is to ask them to back that prediction with money. If you're willing to stake $500 on each of your predictions then you may still be wrong but you'll be a lot more careful with what it is exactly that you predict than if there is no downside to you for your prediction being wrong.