When we get it wrong
Corrections policy
Every calculator and simulator on actfap produces a number someone might weigh a real decision against, and every guide states odds, edges or terms as fact. Getting one of those wrong is a real failure, not a typo.
This page is the honest account of how we catch that and what we do about it.
Written and reviewed by Alexander Korsager · last reviewed September 21, 2026 · corrections
Seen a wrong number?
Email the page and the figure you think is wrong.
- Confirmed corrections on the record
- 9
- Pages corrected
- 8
- Most recent error found
- Sep 28
The record
What we said, and what is true.
Every confirmed correction, newest first. Found by our own fact-checking and review passes in every case below, not by a reader report.
Hedge & Cash-Out Calculator
Found September 28, 2026Stated
At a full hedge, the result panel headlined "Guaranteed profit" regardless of the figure underneath it, including when that figure was negative (for example, -$1.52 at the page's own worked-example odds of American -333).
Correct
The headline now keys on the sign of the computed figure. A full hedge that locks in a loss reads "Locked-in loss" in loss ink; "Guaranteed profit" is shown only when the figure is genuinely non-negative. The same fix was applied to free-bet-value's equivalent label.
Fix written September 28, 2026. It ships in the same release as this record, so it is live wherever this page is.
Middle Calculator
Found September 28, 2026Stated
The FAQ answer on how often an NFL middle lands stated the low end of the range as 4.3% (the two-result window landing least often among the totals a market realistically hangs).
Correct
The true floor over that range is 3.3% (the 55-and-56 window, 225 of 6,720 games). The FAQ sentence and the page's own reference table are now both derived from the same declared band, so the two can no longer disagree about which window is lowest.
Fix written September 28, 2026. It ships in the same release as this record, so it is live wherever this page is.
Slot Volatility Simulator
Found September 27, 2026Stated
The methodology summary stated the shipped defaults' median session balance as about $57.
Correct
The tool's own 1,000-session sweep at those exact defaults gives a median of $55.36, already documented in the component's own code comment. The copy now reads "about $55", matching the precision of the $87 figure quoted beside it.
Fix written September 27, 2026. It ships in the same release as this record, so it is live wherever this page is.
Roulette Spin Tracker
Found September 27, 2026Stated
A misuse note said the amount of data it takes to find a bettable edge is "roughly two orders of magnitude larger" than the sample-size validity floor.
Correct
On the European wheel the tool defaults to, the real gap is about 3.2 orders of magnitude (roughly 1,573 times, not roughly 100 times), understating the true gap by close to a full order of magnitude. The note now states a range, two to three orders of magnitude, roughly three on the default European wheel, since the true multiple varies by wheel (about 2.25 orders on triple-zero).
Fix written September 27, 2026. It ships in the same release as this record, so it is live wherever this page is.
Free Bet Value Calculator
Found September 27, 2026Stated
The methodology section stated the retention asymptote at the shipped 2% commission as 95.2%.
Correct
The page's own formula, (1 - commission) / 1.03, evaluates to 95.1456% at 2% commission, which rounds to 95.1%, not 95.2%.
Fix written September 27, 2026. It ships in the same release as this record, so it is live wherever this page is.
Hedge & Cash-Out Calculator
Found September 27, 2026Stated
The cash-out panel unconditionally read "The book is taking X% on this buy-back", which could render a negative percentage (for example, "taking -6.1%") whenever an offer priced above fair value.
Correct
The sentence now branches on the sign of the computed margin. A negative margin, an offer that beats fair value, now reads "This buy-back beats fair value by X%" instead of the self-contradicting negative "taking" figure. The misuse note beneath it was updated to match.
Fix written September 27, 2026. It ships in the same release as this record, so it is live wherever this page is.
The number the market hung
Found September 24, 2026Stated
The piece compared the totals-market effect to the spreads-market effect as "about a fiftieth" of it.
Correct
Computed on a like-for-like basis, the ratio is about a fourteenth, not a fiftieth. Two further findings on the same page were also corrected in the same pass: an overclaimed "never landed in 26 seasons" hook that does not survive a significance check (demoted to a caveated note) and a stated confidence-interval width that understated the true Wilson interval by half (three points stated, roughly six actual).
Fix written September 24, 2026. It ships in the same release as this record, so it is live wherever this page is.
Risk of Ruin Calculator
Found September 20, 2026Stated
Across six shared probability panels (bankroll-calculator, ev-calculator, parlay-calculator, poker-hand-odds, risk-of-ruin, variance-simulator), a probability between 99.5% and 100% could render a literal "100%" headline over a sentence reading "100 of every 100", short of true certainty, plus, at exact half-points, two counts on the same panel that summed to 101 instead of 100.
Correct
The headline now shows ">99%" for any probability short of true certainty (99.5% up to, not including, 100%), matching the existing "<1%" treatment at the other end, and the two counts now derive from one shared rounding so they always sum to exactly 100. A literal "100%" and a fully filled result grid now render only at true certainty.
Fix written September 20, 2026. It ships in the same release as this record, so it is live wherever this page is.
The mean is not the outcome
Found September 17, 2026Stated
A bet365 bonus's terms were sourced to Casino.org, owned by Legend (the site founder's employer), with no on-page disclosure of that connection. The methodology section also hand-typed three spin-frequency figures, one of them ("once every 285,000 spins" at the steadiest calibrated setting) about 30% off the model's own computed rate, on a page whose stated pitch is that nothing on it is typed by hand.
Correct
The Casino.org citation now carries an on-page disclosure naming Legend and the founder's role there, at the point of citation. All three spin-frequency figures are now interpolated directly from the calibrated model rather than typed, and re-derived: the steadiest setting reads about 1 in 374,000 spins, not 1 in 285,000.
Fix written September 18, 2026. It ships in the same release as this record, so it is live wherever this page is.
Scope
What counts as a correction
A calculator off its formula
A calculator producing a number that doesn’t match its own stated formula.A stale or wrong figure
A house edge, RTP, payout table or bonus term stated as fact that’s out of date or wrong.A misstated explanation
A guide or FAQ answer that misstates how odds, staking or wagering requirements actually work.A misdescribed method
A methodology statement that doesn’t honestly describe what a tool assumes or where it breaks down.
Wording or design you think is misleading, not just factually wrong, is worth flagging too. A page that’s technically accurate but reads as advice to gamble is exactly the kind of thing this process exists to catch.
Reporting
How to report one
Email contact.alexanderkorsager@gmail.com with the page and the figure you think is wrong.
After a report
What happens next
Step 1: Fixed on the page
A confirmed correction gets fixed on the page directly, we don’t quietly edit a number and leave the surrounding claim unchanged.Step 2: The date stays honest
The page’s “last reviewed” date only moves when someone has actually reviewed the content for accuracy, so it stays a real signal rather than one that drifts with unrelated site changes.Step 3: The rest of the site is checked
If a correction points to a mistake that could plausibly affect other pages (a shared formula, a stale reference figure) we check the rest of the site for the same error rather than treating it as isolated.
Before the fact
How we work
Getting a correction right after the fact matters, and so does the reasoning that went into a number before anyone had a reason to question it. The specific analytical choices behind our studies and tools, what was included, what was excluded, and why, get written up on their own, as each one earns an entry, so a figure here can be checked rather than just cited.