Concepts, in your hands.
The maths a model runs on, written for someone who wants to check it rather than take it on faith. Free, no account, and it includes the piece about when not to bet.
Three pieces, in this order.
Expected value before staking, because you cannot size a bet you cannot price. Then how to read a record without fooling yourself.
What a +EV bet actually is
Expected value, the math behind every Veritas Signals pick. Why a +5% EV bet can still lose — and why that is consistent with probability.
Bankroll · 8 min readHalved-Kelly, explained without the textbook
Why we cap stakes at 5 units per pick, why "full Kelly" is too aggressive for almost everyone, and what the halved version is actually optimising for.
Transparency · 6 min readReading a track record honestly
ROI, strike rate, units. Which one to trust, which one to ignore for the first six months, and why our charts include the bad months on purpose.
How the pipeline actually behaves.
What the model is doing, what happens when its inputs are bad, and why the book you use changes the bet.
How the V9 model produces a pick
A non-technical walkthrough of the feature pipeline, the ensemble, calibration, and the availability gate for the current run.
Trust · 4 min readWhy we'd rather show nothing
Stale, blocked, untrusted, missing. The four states that stop a pick from shipping — and why a silent fallback is the worst answer a model can give.
Operations · 5 min readBooks, regions, and best-price routing
Why a pick can be visible to a USA user and invisible to an AUS user. How the price you see is chosen, and what to do when your book of choice doesn't carry the market.
The part that isn’t about winning.
Betting is entertainment with a negative expected return for almost everyone who does it. This piece says so plainly, and says what to do about it.
Check the maths against the record.
Every settled pick is public, graded against the closing line, losses included.
