Is there still a "wet bias" at the Weather Channel?

Nate Silver’s very interesting NYT piece on forecasting includes the following:

The use of the past tense implies that TWC has changed their policy, but I can’t find any confirmation of this. My family uses their website regularly (multiple times daily) because we do not have a car and so need to know what we are getting into when walking or bicycling to work, the pool, tennis courts, etc. As a veteran poker player, I am familiar enough with real, unfudged probability that I would like to have accurate percentages to work with when planning my day!

Always look to the eyeballs. The less decisive the chance of precip, the less likely you will stay tuned in.

Doesn’t expain why forecasting 40 would be smarter than 50 though. When might it increase viewership (or goodwill) to understate the chance of rain? Maybe when ad dollars are betting on sunshine?

It may be true. My local media is saying 70% chance Monday while the National Weather Service is saying 40%. But I don’t put any faith in any percentages. Even if chances are high, atmospheric conditions can keep it from raining a drop, and if you really take notice, it’s easy to find 80% days where the rain misses you entirely and 20% days where you get hammered by storms.

You might try going to and typing in your zip code, then when the page loads click “forecast discussion” on the menu to the left. This is text released several times a day from meteorologists at your local NWS office explaining the weather and what they expect to happen. It can be kind of science heavy but even if you don’t understand that it can still give you a good idea of what’s going on and the amount of certainty in the short term weather forecast, and even the timing. If there are changes in the forecast, you’ll probably see it there before TWC.

If you want accuracy, this site will show you which of several weather services has been the most accurate for your area when you enter a location; for me, it is the NWS followed by TWC (the further accuracy analysis link at the bottom will show you stats for precipitation alone).

Of course, for precipitation in particular, percentages apply to generalized areas; a 20% chance of rain is supposed to be read as a 20% chance of rain anywhere within whatever area they defined (there may also be a time element, as when a broad area of rain is occurring and they say 80% chance of rain, when it is obvious it will rain, or is already raining). Similarly, when the weather report says it is currently sunny or raining, they are usually reporting the conditions at the local airport (where official weather observations are made).

Also of note, look at how the Storm Prediction Center defines probabilities for severe weather, which is why it is common not to get anything when even major severe weather is expected (e.g. tornado probability is the probability of only 2 or more tornadoes, and wind and hail for 10 events each or 6 total).

But that’s no reason to put no faith in percentages. Only 14% of the days of the week have the letter M in their name, but it’s not a shock when Monday comes around, right? This is why I never understand when people say “The weather forecast was wrong” if it says a 60 or 70 percent chance of rain and it doesn’t rain (or vice versa). You need a large sample size before you can say those forecasts are wrong!

It strikes me that most people* psychologically distort percentages and odds. Anything above 75% chance is treated as though it were 95%+; conversely anything below 25% chance is treated like “not going to happen” (or if it does happen, the person or entity giving the odds is “wrong”). Or even people that are a little more realistic in their evaluation of percentages in the 80s lose that objectivity when it comes to something that’s “90 percent” or “95 percent”, treating it as a lock or as a 99.999 percent type deal, despite the orders of magnitude of difference.

There’s an 83% chance I’d survive one round of Russian roulette, but I’m not going to try it no matter how much you pay me, thanks all the same!

*Not I, as I said, at least not any more, because I have over the past few years become so accustomed to dealing with hard percentages in playing tens of thousands–maybe even hundreds of thousands–of hands of poker, and if you distort odds in that game you are going to lose money.

The sentence i have bolded doesn’t make any sense.

Leaving aside the “wet bias” issue, if percentages are calculated with as much rigor as possible, there is absolutely no contradiction between having faith in them, on the one hand, and recognizing that the less likely weather events could still occur, on the other.

If you put four blue marbles and one red marble in a bag, and ask me to pick one out, there’s an 80% probability that i’ll pick a blue one. But if i happen to pick the red one, it doesn’t lessen my faith in the percentages.

ETA: Also, what SlackerInc said. And i like his Russian Roulette analogy better. :slight_smile:

My point was that it’s better to know and understand what the weather is doing rather than rely on an X percentage of rain. If you know how all the conditions are supposed to interact and watch the situation every few hours, you’ll know if that 90% is going to pan out or not. That’s where something like the NWS discussion is going to be far superior than percentages from a national outlet that may only base them on models and may only update them once a day. I would plan entirely differently if i relied on “a 90% chance of storms tomorrow” instead of “if this cold front gets here before 2pm, it’ll kick off widespread storms with heavy precipitation, jail and heavy winds. If after 2pm, the cloud cover will probably hinder any development.”

I recognize that living in Southern California means that i don’t have very much actual weather to deal with. “Warm and sunny” with maybe a bit of coastal marine layer, is about as exciting as it gets around here most of the time.

But even when i lived on the east coast, and in other countries, i was not generally interested enough in the weather to “know how all the conditions are supposed to interact and watch the situation every few hours.” If they said there was a good chance of rain, i took an umbrella; if it rained, i stayed dry, and if it didn’t it was no big deal.

Unless there’s some torrential monsoon, or a 110-degree scorcher, or a big snow dump, or a tornado, on the way, the percentages and the general forecast are perfectly serviceable for most people most of the time.

What on earth does a 50% chance of rain mean anyway? Until we get that time machine working right these are unrepeatable events that either happen or don’t.

You see, this is why there is a “wet bias.” If they overstate the chance of rain and you take an umbrella to work needlessly, no big deal. But if they understate it and you get soaked, you’re pissed.

For what it is worth, they don’t use rain chance percentages here in Finland at all. They just show where the rain fronts are predicted to go on the map and if there’s localized thunderstorms they say so without any numbers. Sometimes they use “… with a chance of rain.” or something along those lines, but again without actual numbers.

It means that if you make a record of all the days with similar conditions, you’ll see that on half of them, it rained, and on half them, it didn’t.

A probability quantifies a trend over time. If you flip a coin, you know once of the sides is going to land face-up. You can’t predict which side it will be for that flip…but you can predict that after 100 coin flips, you’re probably not going to have gotten 100 tails.

This summers drought has added extra issues to forecasting rain.

We’ve been getting extremely narrow bursts of showers. One day I was returning from an errand and I had fairly heavy rain up until the 2nd red light before my neighborhood. My neighborhood was bone dry and literally two blocks away they had heavy rain. I think the weather guy called it funnel rain? Very, very narrow bursts of rain. Extremely frustrating in a drought.

The other new issue is the 7 minute rains. We’d get a forecast of 40% rain. All day it would be cloudy and even thunder clouds. If we were lucky we’d get a six or seven minute shower out of it. Sometimes we never got more than a 5 min sprinkle. Just a trace of rain that did nothing for the drought conditions. This happened repeatedly this past June and July. I’ve never experienced rain like that before. All the conditions were perfect for rain and it just fizzled out.

How can they forecast something like that?

Well, seeing as TWC has the highest accuracy in precipitations for the month in my area, and is only a tenth of a percentage behind in the yearly measure, I’d say no.

And, really, if you get mad when it rains when it’s not supposed to, the best thing you can do is just always assume it’ll rain, and bring at least an emergency poncho when it’s too hot for a jacket with a hood. And keep an umbrella in your car.

I understand how probability works, but I find the report percentages dubious. I think they’re made up, not based on actual probabilities produced from records of conditions.

I’m sure you’re right about the reasoning, but i don’t get pissed at the weather forecasters if they miss the mark. I’m rational enough to understand that, while they do their best to predict the weather, they don’t control it.

You don’t really have to understand it. The reason I mentioned the NWS area forecast discussions was because the NWS does understand it and relay their thoughts on the situation, and all you need to do is read a paragraph about it to get the details. Some of them can be heavy on the science but even if you don’t understand what exactly they’re saying, you can still get a grasp of what is going on. The OP checks TWC multiple times a day but, IMO, would be better served reading a paragraph from local meteorologists than a canned forecast from TWC.

I read the AFD every morning and get an idea of what’s going to happen throughout the day. If the weather is expected to change and I am planning something outdoors, I will read it again later to get an update. As a result, I haven’t paid attention to percentages in a long time and find them to be fairly worthless. As everyone who takes issue with my dislike of percentages pointed out, you need a large sample size to see a trend over time for them to be accurate. How on earth do you use that to plan activities?

Good point, but a little late. One a GQ or GD thread goes into hard-data territory, there’s usually no bringing it back.