FREDcast™ is an interactive forecasting game in which players make forecasts for four economic releases: GDP, inflation, employment, and unemployment. All forecasts are for the current month—or current quarter in the case of GDP. Forecasts must be submitted by the 20th of the current month. For real GDP growth, players submit a forecast for current-quarter GDP each month during the current quarter. Forecasts for each of the four variables are scored for accuracy, and a total monthly score is obtained from these scores. Scores for each monthly forecast are based on the magnitude of the forecast error. These monthly scores are weighted over time and accumulated to give an overall performance.
Higher scores reflect greater accuracy over time. Past months' performances are downweighted so that more-recent performance plays a larger part in the scoring.
All FREDcast™ series are data for the United States.
Release dates are the days that economic agencies release economic data to the public. Economic data is always released on a lag, for example, March payroll employment is release on the first Friday in April. For 2016, the release date for March payroll employment is April 1, 2016.
Output growth is the seasonally adjusted annualized quarter-to-quarter percentage change advance estimate of GDP from the previous final estimate of GDP, rounded to the nearest tenth of a percent. The advance estimate is the release that is available one month after the end of the quarter. The advance estimates for GDP are released at the end of January, April, July, and October.
GDP is a quarterly variable and the forecast period is a month. Thus, users forecast the same quarterly GDP value for each of the three months in the quarter. Each of these forecasts is evaluated against the same advance estimate of GDP.
Unemployment is the seasonally adjusted level of the headline unemployment rate in percentage points rounded to the nearest tenth of a percent. The unemployment data are usually released on the first Friday of the month.
Inflation is the percent change from one year ago in the headline CPI rounded to the nearest tenth of a percent. The CPI data are released in the middle of the month.
Employment growth is the seasonally adjusted monthly change in nonfarm payroll employment to the nearest thousand persons. The employment data are usually released on the first Friday of the month.
Yes. The forecasts for each month must be made by the 20th of that month. Let's use May as our example: You must provide a forecast for May's GDP, CPI, unemployment, and employment by May 20. In technical terms, this is a "zero-horizon" forecast.
Yes. Provided you do so before the deadline.
Yes. Use the name change link above to change your name. If your new username is reported and changed by the FREDcast™ administrators you cannot change it again.
You can only change your name once.
We have changed how points are assigned to each monthly forecast. Under the old scoring algorithm, forecasts that were more than 3 standard deviations or so away from the realization were given negative scores. At the start of the recession caused by COVID-19, many forecasters missed their realizations by more than 10 standard deviations, resulting in large negative scores. Thus, during this period of high uncertainty, not forecasting might be a good strategy. In light of this, we have redesigned the scoring algorithm so that any forecast produces positive points.To make the points across eras consistent, we are also resetting the all-time scores and increasing the rate at which older scores are discounted in the all-time score.
There are two types of scores in FREDcast™: monthly points and all-time scores. Monthly points are a measure of how well a player does in any one particular month. All-time scores are a weighted aggregation of monthly points. A player’s monthly points affect the all-time score, but the all-time score does not affect monthly points.
A player’s forecasts are scored separately for accuracy and then summed up to one monthly points score. A player can earn up to 250 points for each economic data series, for a potential total of 1,000 points each month. Points awarded are inversely related to the player’s absolute forecast error. A higher absolute forecast error (i.e., your forecast is farther away from the actual release) means fewer monthly points. Use the send feedback form to request exact formulas.
The all-time score is a weighted sum of monthly points. In other words, a player’s all-time score is adjusted up or down by the points earned from previous forecasts. A player’s current monthly points receive the greatest weight in the all-time score, while prior monthly points diminish in weight. A user's all-time score is a weighted average of the points earned in the most recent month and those earned in the prior 11 months. Use the send feedback form to request exact formulas.
For example, a user who earns the full 1,000 points in one month but who did poorly in previous months (or has not forecasted consistently in the past year) may receive a lower all-time score than someone whose forecasts consistently have smaller forecast errors.
The all-time score is weighted, so the most recent monthly points are multiplied by approximately 0.5. If you do not have a long scoring history to add to your current monthly points, it is likely that your all-time score is lower than your current monthly points score.
Yes. Negative scores are bad.
The scores for forecasting different economic releases also take into account how difficult it is to forecast those specific releases. The sample variance of GDP is higher than the variance of unemployment, so the score for forecasting GDP is weighted higher in a player's overall score for the month.
Yes. There are two scenarios where players may end up with different all-time scores following identical monthly forecasts.
Up to 250 points are awarded for each economic data series. The scoring formula starts with 250 and takes points away based on how far a player’s forecast is from the actual number. A larger distance means fewer points. If a forecast is sufficiently far away from the actual number, the total amount of points lost will exceed 250 and result in a negative score.
If the change in employment lies outside the range of values that a user can input (i.e., more than +10M or less than -10M), we will treat the actual value as if it were +10M or -10M. Thus, if a user forecasted -10M and the actual value is -12M, we will adjust the actual value to -10M and the forecast error will be zero. If the user forecasted -9M and the actual value is -12M, the forecast error will be -1M.
A league is a collection of players who have a private leaderboard for forecasts. This allows members of a class or a company to compete against each other. The league will have a separate page for a league leaderboard that displays the distributions of the league's forecasts and tracks how the players are doing.
The leagues are formed when a "commissioner" (e.g., teacher) solicits a league ID from FRED. The league ID includes a passcode that the commissioner shares with the other participants. The participants sign up for a FRED account, enter the passcode, and join the league.
A league invite code, or league ID, is a unique set of numbers and letters that is generated when a league owner creates a league. The league owner can distribute this code by email to his/her participants. League owners can view their league invite codes by navigating to the league manager page under the "Leagues" menu. If you're trying to join a league and don't have an invite code, ask your league manager to share the code with you.
Yes. Even if your league disbands, your FRED account keeps your account open and you can continue to make forecasts within the public league.
You receive FREDcast™ badges for achieving certain in-game goals such as making your first forecast. The full list of badges is here.
No. For example, if you receive a badge for being in the top ten percent of the leaderboard in a month after one series is released and scored, you will not receive another badge if you stay in the top ten percent when a subsequent series is released and scored.
For some achievements, you can receive a badge for each league if the badge is league- specific. For example, being in the top ten percent of a league is a league-specific achievement. You can receive one badge for being in the top ten percent of each of your private leagues, which will display on your scoreboard for those leagues, and a different badge for being in the top ten percent of the public league.
FREDcast Battle Leagues are head-to-head competitions where two FREDcast leagues compete against each other to determine the best individual forecasters, and best overall league forecasting average. Battle Leagues require no additional effort from league managers or forecasters, and will show up on the scoring page for your private league. The league that has, on average, the most accurate forecasters will achieve the highest aggregate score. The individual leaderboard for the Battle League will show which individuals from each league are scoring the highest in a given forecasting period.
To start a Battle League you must be a league manager of a current FREDcast league. You can request a Battle League from the League Manager page (League info> Your Leagues > Manage League). Battle League requests require that you provide the opposing league manager’s email, and the name of the opposing league. You must also specify an end date for the Battle League, which will be validated by the opposing league manager. Battle Leagues start when the request is confirmed by the opposing league manager.
Yes. One league can participate in as many Battle Leagues as the league manager wishes to challenge.
We cannot guarantee that we will be able to provide an opposing league manager, but if you would like to participate, email firstname.lastname@example.org and we may be able to suggest an opposing league.
There are two scores in a FREDcast Battle League—aggregate and individual. The individual score functions the same exact way as the monthly points score in regular FREDcast leagues, and shows the FREDcast user with the most accurate forecasts across all the series. The Battle League aggregate score is a trimmed mean calculated for each sub-league. The trimmed mean calculation removes approximately the bottom 20 percent of scores for each series in a given time period. This calculation gives greater weight to forecasters doing well, and limits the possibility that non-participants or erroneous forecasts can ruin an entire league’s average.
Each Battle League prefix is assigned by FREDcast staff so that individuals can identify who comes from which league on the individual Battle League leaderboard. League managers cannot choose or change their league’s pre-fix.
Yes. To Battle two of your own leagues, fill out the FREDcast Battle League Request Form on the league managers’ page and select the box that identifies you as the opposing league manager.