THREE THINGS YOU’LL SEE

One published pick. Other views to check it against.

Every Week 7 forecast requires a saved research checkpoint, market benchmark and ten documented decisions. Judgment estimates are labeled unvalidated. A run receipt in its briefing shows the sequence, evidence and limitations.

HOW THE PICKS ARE MADE

Research first. Compare the market. Save the decision.

The market benchmark uses this process for the ten games selected by ESPN College Pick’em. The final card may differ when its documented review supports a different estimate. The same market method covers other games on the research page when the required odds are available.

1. Collect both sides of the same game

   We use paired moneyline prices from at least two named sportsbooks. A moneyline is the price for a team to win outright, rather than cover a point spread. Each quote keeps its source and update time.

2. Remove the bookmaker’s margin

   Odds usually imply a combined probability above 100%. We convert both prices into probabilities and divide each by their combined total. The two adjusted win chances then add up to 100% for that book.

3. Average the adjusted estimates

   Each qualifying book contributes an estimate. The benchmark takes the simple average and picks the team above 50%. If the required odds are missing, we show the gap instead of inventing a market forecast.

4. Save the pick and its evidence

   The published update retains its probabilities, confidence order, source times and notes. A later update gets its own record. Earlier picks remain available.

This describes the current method. Some earlier saved updates used different source blends or a stated confidence override. Their original methods and numbers stay with those records.

WHY ONE PICK GETS MORE POINTS

Ten is the strongest pick on this card.

The benchmark ranks selected winners by their unrounded win chances. The reviewed card may use a different confidence order when the football evidence supports it. Each change requires a saved explanation and an explicit comparison of expected points under the same market probabilities. Every number is used once: 55 points are available.

Assigning larger point values to stronger win chances maximizes expected confidence points under those estimates. A football review that changes this order gives up some of that benchmark expectation; we disclose the cost and do not claim an unproven payout advantage. It does not guarantee a weekly prize or maximize the chance of beating every particular opponent. Close probabilities can also make adjacent ranks nearly interchangeable.

The tiebreaker is a separate predicted combined score for ESPN’s designated game. It is saved with the update and does not change the confidence order. See the current Weekly Picks.

THE RESEARCH MODEL

What would the team data say?

This is a separate, experimental statistical model. It estimates the chance of winning from team information rather than using the current sportsbook prices to choose a side.

The formula was fitted on 2015–2024 games and frozen. New eligible team data can change a forecast; we do not refit the formula every time someone refreshes the website. The current game’s final result is not an input to its pregame forecast.

When it disagrees with Duck Force One’s pick, that is useful context to investigate—not an automatic reason to switch sides. Agreement is not independent proof that a pick is right. We compare both estimates with the actual results.

THE CASE BEHIND EACH GAME

The football case behind the numbers.

Game analysis brings together team statistics, official availability reports, weather, local coverage and attributed analyst views. The game recap and preview connect the pick to its evidence, gives the other team’s route to a win and identifies unresolved questions.

We do not add an arbitrary injury or weather adjustment to a win chance. A reporting note can flag a risk or a need to recheck the odds without silently changing the saved number.

ESPN probabilities are not analyst picks

An ESPN matchup win probability comes from its statistical forecast. A Football Power Index (FPI) rating is a measure of team strength, not itself a matchup percentage. A writer may predict an outright winner, a point-spread result or a total; those are different questions and are labeled accordingly.

AI tools help collect and organize evidence, build the site and draft explanations. They are not a news source or a substitute for the linked reports. The numerical estimates follow the saved methods above; generated prose is not another forecasting model.

100 POSSIBLE SATURDAYS

The simulations show how widely a week could go.

Each simulation draws a winner for every game using the saved win probabilities, then scores the card. The process repeats 100 times. Games are drawn independently and probabilities stay fixed within that saved run.

A team with a 60% win chance will not necessarily win exactly 60 of those 100 draws. The sample illustrates uncertainty; it does not improve the original forecast, simulate an exact football score, or prove that the picks work.

Play out your week lets you set particular results and see what they would mean for the card. Those scenarios do not change the published picks.

THE RESULTS THAT MATTER

Keep the original prediction. Then count what happened.

The season record uses each week’s registered original forecast. You can also grade another saved update, but it does not rewrite that original benchmark. A score refresh updates the outcomes; it does not replace the predictions being evaluated.

We show correct picks and probability error. The error measure checks how close the estimated chances were to what happened; lower is better, and confident misses are penalized more. Missing, unfinished or unverified results are excluded, never counted as losses.

DATES & LIVE UPDATES

Fresh scores and a new forecast are different updates.

Forecast saved

When that exact set of picks and win chances was recorded. It stays fixed for that update.

Report published / checked

When the source released a report, and when we checked it. A recent check can still refer to an older report.

Scores checked

When the score source was last successfully checked. It does not mean the forecast was rerun.

What updates automatically

A cloud collector runs on Supabase, independently of this website and your computer. It checks CollegeFootballData for scores and game status across the saved research slate every 30 minutes. Around active games, it checks about every two minutes. Oregon is included whenever its game is in the saved slate.

Verified scores are saved once and shared by Game day, game pages, team schedules and full-slate Results. Open pages check that shared record once a minute while visible. The collector keeps running when you close the page. This is periodic score checking, so changes can appear a few minutes after the play and depend on the provider’s timing.

When a check fails

The site keeps the last verified score and its original check time. An update panel shows the last successful check, the next planned check and any delay. The collector retries automatically; an unavailable score is never counted as a loss. Confirmed finals earn points, and score changes are retained for review.

What is not yet monitored continuously

There is currently no dedicated midweek news-discovery collector. The hourly availability task focuses on games within two hours of kickoff; it does not continuously search for Dante Moore or other unresolved players. New reporting appears when a source review is completed. A separate news-monitoring service is being planned; it is not active yet.

Statistics and reporting have their own clocks

A separate daily collector refreshes the season statistics used in matchup comparisons after 6 a.m. Pacific. It checks every game in the active published slate, using CollegeFootballData totals through the preceding week. A failed check keeps the last verified edition.

Team schedules, polls, rosters and player totals refresh daily at 6:30 a.m. Pacific. A full-slate research task is scheduled for 7 a.m.; postgame reviews are scheduled for 8 a.m. A scheduled task is not proof that its review finished. An hourly check from 9 a.m. to 11 p.m. looks for final availability and weather updates for games approaching kickoff. A separate 6 a.m. task prepares fresh odds, the upcoming slate and a new draft card once ESPN’s lineup and required evidence are available. These cloud tasks keep the last valid edition if a source fails. The update-status panel reports a failed or incomplete review.

Odds, availability reports, news, polls and detailed player statistics retain their own dates. The score collector does not review those sources, create a new forecast, change the confidence order or submit a card to ESPN. Submitted league totals remain the recorded ESPN results; Game day can separately calculate a card’s points from verified finals.

A pregame percentage remains a pregame percentage during the game. It is not a live win probability calculated from the score and clock.

MAKE IT YOUR OWN

Your card stays yours.

Build my card lets you choose winners and rank all ten picks. Your browser draft and shared link do not alter Duck Force One’s published picks or submit anything to ESPN. A draft can be edited, including after kickoff; an explicitly saved pregame record preserves the version and server timestamp.

Compare the choices, read the evidence, and make the call yourself.