Wikipedia:Sandbox

Prediction is a noun in the English language, and it means "the act of predicting." To "predict" means to foretell, to state, foresee, or make known in advance, especially on the basis of special knowledge. Prediction is a fundamental part of the spirit of doing statistical research. Scientists and statisticians aim to build reliable models that enable them to make predictions; common examples include forecasting the weather, natural disasters, stock markets, and human population growth. Meteorologists watch the movements of clouds and storms, and with the help of scientific and historical data, they can forecast the paths of these objects. A quick Google search for the word "weather" brings up a weather forecast for your location. Typing "weather" followed by the name of a specific city or town allows you to see the current and near-future weather for that city or town. Try it. Try "weather Miami" or "weather NYC." Weather forecasting also makes it possible to determine, in advance, when and where a currently active blizzard or tropical cyclone will hit, allowing authorities to warn residents to prepare for the storm, or in severe cases, evacuate the area.

Ideally, we want there to be a high correlation between our input and output variables. However, the old saying is true: correlation does not imply causation. For example, we may find a high correlation between drownings and ice cream sales, but there is no causal connection between those two variables; in this case, there is a third variable that results in the increase of both: the time of year. In the summer, when it is hot outside, ice cream sales increase because ice cream is a popular food for cooling down in the summer. The high temperatures are also why going to the beach is a very popular activity in the summer. And unfortunately, there could be some deaths due to drowning. Although correlation alone does not imply causation, correlation is one of the three "ingredients" in establishing causation. To show that X is the cause of Y, you must establish:

• 1) X is correlated with Y.
• 2) X happened before Y.
• 3) No other factors account for the correlation between X and Y.

We can fit our data using a regression model. The letter "R" is typically used to denote the correlation coefficient. If R is very close to one, such as 0.98 or 0.99, this indicates a strong positive correlation between X and Y. If R is very close to −1, such as −0.98 or −0.99, this indicates a strong negative correlation between X and Y. In addition to a linear model, we can also use other functions such as a quadratic model, a cubic model, an exponential model, a logarithmic model, and so on. In the real world, your data won't fall perfectly on the line or curve, so we try to find the "curve of best fit." Here, "best" means we are minimizing the total squared error. The "best fit" solution is also called the least-squares solution.

In addition to curve correlations (such as the familiar "linear correlation"), we can also use the "sign correlation" or the "rank correlation." If X and Y have a positive correlation, then larger values of X usually correspond to larger values of Y. If X and Y have a negative correlation, then larger values of X, for the most part, go with lower values of Y.

Types of predictions

Predictions fall into three categories:

• Determinate predictions: Predictions that are certain due to a fixed, constant pattern or formula.
  • Example 1: "Two years from now, it will be November 2026."
  • Example 2: "In Boston, December 7, 2024, will have a sunset time of 4:11 p.m." (Within a location, sun/moon times for any particular point in the year is practically constant. Every December 7th in Boston, for the last 20 years, maybe even longer, has had a sunset time of 4:11 p.m. The sunset time for 8/25/2024 was only two minutes earlier than that for 8/25/2011 [13 years ago], 7:28 p.m. this year vs. 7:30 p.m. in 2011.)
  • Example 3: "June 1, 2025, will be a Sunday." This can be calculated mathematically by calculating the number of days between now and 6/01/2025, then by taking the remainder modulo seven. The date 6/01/2025 will be 183 days from now, and 183 divided by 7 leaves a remainder of one, and one day after today (Saturday) is Sunday. Even if you can't do the calculation yourself, you can easily search on Google for "2025 calendar" or "2030 calendar" or whatever year you want, followed by the word "calendar."
  • Example 4: "There will be a total solar eclipse in Spain in August 2026." Solar eclipses are very predictable, being known for decades, even centuries, in advance.
  • Example 5: "In December 2027, Taylor Swift will be 38 years old." This is just simple arithmetic: Taylor Swift was born on December 13, 1989, and subtracting these numbers gives: 2027 − 1989 = 38.
• Educated predictions: Predictions that are based on scientific or historical data. For instance, meteorologists make weather forecasts by watching the movements of clouds and storms and using historical data to predict their path.
  • Example 1: On November 29, 2024, at 05:00 UTC, "Baby Shark" had 15,343,065,154 views. Today, at 05:00 UTC, it had 15,346,704,150. Like usual, linear models can be used to predict the short-term growth rate in the near future. Subtracting these numbers, we see that it gained 3,638,996 in those 24 hours, so tomorrow, at 05:00 UTC, we can expect it to have 15,350,300,000 views.
  • Example 2: Album of the Year is a music review site that allows users to sign up and rate and review songs and albums. Album of the Year is similar to Metacritic, a site that allows users to rate and review video games and movies. This site is growing exponentially. And if you think this is yet another misuse of the word "exponential" as a loose synonym for "rapidly," it's not. This site really is growing exponentially, and I have the numbers to back that up. By January 1, 2020, there were 202,933 pages on the site. By January 1, 2021, this number increased to 320,840. The number of pages on the site was 453,237 on January 1, 2022, then 588,763 on January 1, 2023, and last but not least, 833,485 on January 1, 2024. Each year, it increases by 30% to 60% from its previous value. Between 1/01/2023 and 1/01/2024, it increased by 41.5% (833,485 / 588,763 = 1.4156545). If we look at all nine January 1st values from 2016 until 2024, we get an exponential function with a base of 1.4388 with a correlation coefficient of R = 0.9988. On June 28, 2024, at 4:50:11 p.m. Eastern Time (20:50:11 UTC), the site reached one million (1,000,000) pages, with "My Merry-Go-Round," a 1973 album by Johnny Nash, becoming the site's one millionth page. As I am writing this article on November 30, 2024, there are now over 1,147,000 pages on the site. Using the exponential model, it will be around 1,200,000 (833,485 × 1.4388 = 1,199,218) on January 1, 2025, and by July 1, 2025, it will be somewhere in the vicinity of 1,199,218 × sqrt(1.4388) = 1,199,218 × 1.1995 = 1,438,462.
  • Example 3: According to the current weather forecast for New York City, it is expected to be 34 F and sunny on Monday, December 2, 2024, at 10:00 in the morning.
• Psychic Predictions: Predictions that extend beyond the scope of modeling; instead, these predictions rely on gut feeling, intuition, as well as real-world blessings and curses that are known to be legitimate. An example of a real blessing is that the name "Clitus" has mystical, magical properties; with a name that means "glory", "splendid", "famous," it's no wonder there has never been a mass shooter with this name.
  • Example 1: There will not be a "superlative mass shooting" in the U.S. that is committed by a man named "Clifton" during 2025. A "superlative mass shooting" is defined as a mass shooting which breaks the record for the "deadliest mass shooting" in the country being studied. In the United States, there were three "superlative mass shootings" in the last 20 years. In April 2007, the Virginia Tech shooting resulted in 33 deaths, surpassing the 1991 Luby's shooting which killed 23 people. It has since been surpassed two more times, in 2016 with the Orlando nightclub shooting (50 deaths), followed in 2017 by the Las Vegas shooting (61 deaths). As we know, "Clifton", like other CL names, is a name with mystical, magical properties, and Cliftons are kind, helpful, and good-natured. Cliftons don't carry firearms and open fire. And we will never, in a million years, have a mass shooter named Clitus. Period. Let alone one who is coldhearted enough to break the "deadliest shooting" record. People named "Clitus" are very kind, caring, and loving, and they live up to the meaning of "glory."
  • Example 2: Justin Bieber's next album will be titled a spiritually positive word, such as "Clarity," "Comfort," "Connect," or "Healing." This is a reflection of Justin's sweet and kindhearted personality. Starting with "Believe" in 2012, his album titles have been one word, seven letters, and a positive word that, in the context of the album, reflects his values of growing spiritually, making the world a better place, and embracing life. He is a great role model. Sure, he did some bad things in 2013 and 2014 and got in trouble with the law multiple times, but in 2015 and later, he has turned a new leaf and has demonstrated time and time again that he is a kind and generous person with a good heart. In September 2017, he donated $25,000 to Houston, Texas to help with hurricane relief after Hurricane Harvey. This storm is tied with Katrina (2005) for the costliest Atlantic hurricane, at damages estimated at $125 billion. On February 7, 2020, Bieber donated $100,000 to Julie Coker, a 22-year-old fan who works in mental health awareness. Coker revealed that her own past struggles with mental health motivated her to work for mental health. She praised Bieber by saying, "[Bieber] has a big following, so if he has a good message about mental health, hopefully everybody else ... will want to start thinking about mental health in a different way." During that same month, Justin Bieber made a donation to Beijing Chunmiao Children Aid Foundation in China to support COVID-19 relief. It was also around that time, that he released his album "Changes," which in this context refers to making positive changes in the world. And his latest album, released in March 2021, is titled, "Justice." This one's straightforward: he is advocating for justice and equality, and he wants to put an end to discrimination and exclusion. Notice how "believe," "purpose," "changes," and "justice" are all seven-letter words? Other words with seven letters that fit this pattern include: "beloved," "clarity," "comfort," "connect," "healing," and "support." Justin Bieber isn't known for high ratings on his albums. That makes me quite sad because I love Justin Bieber, and I have been a big fan of him since 2010 when I was in the 8th grade. For the past two years, updates about Justin have become fewer, sparser, and further between ever since he canceled the Justice World Tour in September 2022. He got Ramsey-Hunt syndrome in June. On March 15, 2023, I breathed a sigh of relief when an article showed up in my newsfeed stating that, after 9 months, Justin had finally regained the ability to smile and regained mobility of the muscles on the affected half of his face. Justin Bieber has never scored above 80% on any of his songs on Album of the Year, and I am hoping that his next album and at least one of its singles will finally break the 80% barrier. Not counting "Home to Mama" which currently has only one user rating, his highest song is "Confident" (2014) with a 77% user score from 38 user ratings. It has been 60 months since the last time he went live on Instagram, in early May 2022. Let's hope 2025 will be the year Justin finally starts doing Instagram Live again. Speaking of Justin Bieber ...
  • Example 3: The next Justin Bieber single to be released and added to AOTY will have a rating of AT LEAST 5/100. Yes, I am 100% confident that Justin's next single to be released and added to the site will be rated higher (hopefully MUCH higher) than 5%. And to be clear, I am counting only brand-new singles, not old singles that were recently added to AOTY ("Home to Mama" was released ten years ago in 2014, but it was only added to AOTY recently, nine months ago in February 2024). This is a psychic prediction because of my clairvoyance about Justin Bieber. While his ratings haven't been "amazing" for the most part, peaking at only 77%, when other artists like Adele, The Weeknd, and Billie Eilish have multiple songs above 85% or even 90%, he usually doesn't get super low ratings, either. His first song in 2020, "Yummy", was a big exception. It has a user score of 12% based on 708 user ratings, making it by far his lowest single on Album of the Year, unless you count the Country remix (2%, 286 user ratings). No other Justin Bieber song comes even close to being as low as "Yummy" in terms of audience reception. Not counting remixes, his next lowest song, "I Feel Funny", got a 25% based on 229 user ratings. While 25% is still terrible, it is far superior to 12%. And "I Don't Care" by Ed Sheeran ft. Justin Bieber? That song got a 27% (260 user ratings). "Baby" has a 41%, which, while still lower than 50%, is surprisingly "high" considering its reputation as a very hated song. Plenty of Justin Bieber songs have had "good" or "decent" ratings. "Ghost," my favorite Justin Bieber song from the last five years, got a 66% based on 89 user ratings. I love that song. In the spring of 2022, whenever it came on, I would listen to the song and enjoy every second of it. All 150 seconds of this beautiful song. I am very disappointed that Todd in the Shadows put "Lonely" on the "worst of 2020" list, but I am very glad "Ghost" wasn't on the 2021 or 2022 lists for the "worst songs" of the year. Thank you, Todd, for not putting "Ghost" on your list of worst songs. And "Company" and "Heartbreaker" both got 73%. It is, thus, decently likely that Justin will get 70% or 80% on his next single. My philosophy about "negative records" is that they should not be broken. "Deadliest mass shooting," "costliest Atlantic hurricane," "deadliest day at Hospital X," are records we need to avoid like the plague. "Lowest-rated Justin Bieber song" is a record I hope will never be broken again. On November 18, 2022, Pokémon Scarlet and Violet became the lowest-rated mainline Pokémon games. With a score of 2.8/10 ("too much glitching"), these games made Sword and Shield's score of 4.6/10 look amazing in comparison. I hope this record never breaks again; in fact, I hope the Pokémon games will finally start going uphill and getting better again. There is strong consensus among today's Pokémon fans that the games have been on a steady decline since Pokémon X and Y (2013) with the exception of Pokémon Legends Arceus, which got 8.2/10. This "steady decline of quality" is a curse that has been going on for too long. It's been 11 years since Pokémon X and Y were released. Let 2025 be the year in which that curse is finally broken. Anyway, Justin Bieber will not set a new "personal low." My ironclad philosophy about negative records is that they should seldom, if ever, break, and if a record does break, it breaks by a little, not by a lot. Even in the unlikely event that Justin does score a new "personal low," it is more likely going to be a little lower than the record of 12%, such as 9%, 10%, or 11%, not 3% or 5%. The probability that Justin gets below a 5% with his next single is zero. I am hoping Justin Bieber will release a song with a very cool title, such as "Climatized Dance," and I hope it gets a 90% user rating and makes it onto Todd's "best hit songs of 2025" list.