Missouri 45, Florida 17: Our Worst Betting Day Ever

Saturday, October 3, 2026 was the worst betting day this site has ever published. We put $163,013 through the window and got $152,743.70 back, a loss of $10,269.30, or 102.7 units. The board broke everywhere: No. 25 Missouri beat No. 8 Florida 45-17, Minnesota beat Michigan at home, UNLV beat Cal 39-31, and Idaho ended Montana State's run at No. 1 in the FCS 35-34. This is the article about that day.

We have now lost three consecutive heavy college Saturdays. That is no longer variance, it is a pattern, and this article is about naming it rather than explaining it away.

What Happened

Florida collapsed. No. 25 Missouri beat No. 8 Florida 45-17. Not a narrow upset, a 28-point dismantling of a top-ten team by a side ranked seventeen places below them.

Minnesota beat Michigan at home. Another ranked team losing to an opponent the number did not respect.

UNLV 39, Cal 31. A non-conference result nobody had on their card.

Idaho 35, Montana State 34. The Vandals beat the No. 1 team in the FCS for the first time in program history.

Four results, four directions the board was not priced for, on one afternoon.

Three Saturdays, Three Losses

Here is the uncomfortable part, laid out honestly.

Three heavy college Saturdays in a row. Each one bigger than the last. Each one worse than the last.

Our NFL Sundays over the same stretch returned +110.6, +51.9 and +35.7. The weekday cards have been fine. The pattern is specific and it is specific to college Saturdays.

Why This Keeps Happening

Two explanations, and we think both are true.

The first is handle. $163,013 on a single Saturday is the largest card we have run. A college Saturday offers more games than any other day, which creates the sense of more opportunity, which creates more bets, which turns a bad read into a very large number. We wrote the same thing after September 19 and then did it again at a larger size twice. That is the part that is squarely our fault.

The second is that college football in 2026 is behaving differently from its numbers. Ranked teams keep losing to unranked ones. A No. 8 losing by 28 to a No. 25 is not a close call the market got slightly wrong. It is the market being badly wrong about the gap between two programs.

Live models inherit that error and compound it. When a ranked team falls behind, the in-game price keeps expecting the correction because the pregame rating says the talent gap is real. We bet the gap between that expectation and reality. When the rating itself is the thing that is wrong, there is no gap to bet, only a model and a bettor making the same mistake together.

That is the honest diagnosis. It is not bad luck three times. It is a read that has stopped working on one specific day of the week, applied at increasing size.

The Documented Day

This sits on the results page in the same format as every winning day. October currently stands at −45.1 units because of it.

What Changes After a Day Like This

Specific and boring, which is how it should be.

A hard daily cap on college Saturdays. Not a feel, a number. The handle grew from $113,724 to $163,013 across three losing Saturdays and the losses grew with it. Capping the day removes the worst outcome regardless of whether the read improves.

Treat a correlated slate as one position. When a dozen bets express the same idea, they are not a dozen bets. The bankroll math that makes this business survivable assumes independence, and a college Saturday is the day that assumption is least true.

Stop betting the rating when the rating is the error. This is the hardest one and the most valuable. The entire September and October pattern says preseason-derived ratings are mispricing the gap between ranked and unranked teams this year. Real football handicapping means noticing when the thing you are exploiting has become the thing exploiting you.

Why We Publish This

Because the alternative is worthless.

Every pick service on the internet will show you a 110-unit Sunday. Almost none of them will show you a 102-unit Saturday, and the ones that will not are asking you to trust a record you cannot check.

The test for any sports handicapping service is simple: ask for the worst day. Ask for the longest losing run. If those are not published at full size, in the same place as the wins, you are reading marketing. This article exists so that nobody has to ask us.

Evaluating football handicappers on win rate rewards whoever picks the most flattering window. Evaluating them on a complete daily ledger, drawdowns included, is harder to fake and tells you far more.

What a Drawdown Means If You Are Following Along

This is the part that matters if you are betting these alerts rather than reading about them.

A 102.7 unit day is survivable or ruinous depending entirely on what a unit is to you. At 1% of bankroll per unit, this day costs roughly a tenth of the roll and you continue. At 5% a unit, it is close to fatal. Nothing about the picks changes between those two bettors. The only difference is sizing, and sizing is the half of this that nobody sells because it is not exciting.

The standard answer is 1% to 2% of bankroll per unit. On a day like October 3 that range is the difference between a bad Saturday and a finished bankroll.

The second thing worth saying: a service posting a 102-unit loss is telling you something about the shape of its returns. Our card is concentrated, it swings hard, and it is capable of producing both a +110.6 Sunday and a −102.7 Saturday inside three weeks. If that volatility does not suit how you bet, the lower-volume packages exist for exactly that reason, and the honest advice is to take one of those rather than the top tier.

Anyone comparing college football picks services should be asking about drawdown shape before win rate. A smooth, modest record and a violent, high-variance one can post identical season totals and be completely different products. The best sports handicappers tell you which one they are.

What This Doesn't Prove

One Saturday does not mean the process is broken, and we are not going to perform that conclusion for drama.

September finished at +205.7 units with two losing days in it. The NFL card over this same stretch has been consistently positive. A single day inside a large sample is a single day.

But three in a row, trending worse, at growing size, is a different claim than one bad day, and we are not going to pretend otherwise either. That is a pattern and it is being treated as one.

Past results do not guarantee future performance. Every day, in both directions, is on the results page.

Frequently Asked Questions

What was the biggest college football upset on October 3, 2026?

No. 25 Missouri beat No. 8 Florida 45-17. The 28-point margin against a top-ten team made it the most consequential result of the weekend for the Top 25.

What other upsets happened that Saturday?

Minnesota beat Michigan at home, UNLV beat Cal 39-31, and in the FCS, Idaho beat No. 1 Montana State 35-34 for the first time in program history.

How much did The Best Bet on Sports lose on October 3?

$10,269.30 on $163,013 wagered, which is 102.7 units at $100 a unit. It is the worst single day published on this site, and the FanDuel statement is posted with it.

Why have college Saturdays been losing?

Two reasons. The card size has grown each week, which scales a bad read, and ranked teams have been losing to unranked ones often enough in 2026 that the ratings driving the numbers appear to be mispricing the gap. A live model built on those ratings repeats the same error.

What is correlated risk on a college football Saturday?

When a dozen bets all express the same underlying idea, they behave as one large position rather than twelve independent ones. A day where the whole board contradicts that idea produces a far bigger loss than the bet count suggests.

Should a pick service publish its losing days?

Yes, and the ones that do not should be assumed to have them. A record without drawdowns in it has been curated. The useful questions for any service are the worst day and the longest losing run, at full size, in the same place as the wins.

Where can I see every day including the bad ones?

The results page publishes every day with the sportsbook statement attached. Live alerts go out by Email and Discord, and packages start at $39.95 a month on the packages page.