Vector Core prices the whole field. Every runner in every Australian thoroughbred
race gets a win chance, a place chance and a fair-odds
line — so you can see at a glance which runners the market has too short and which it has
too long. It is a rules-based model: the same treatment for a Tuesday midweeker as for a Group 1,
and nothing hidden behind a number you cannot question.
This is the Vector Core prediction
On the boards and race cards the Core tip is tagged
Core — it's what your
Vector Core plan gives you, on every meeting we cover. Two other flows run
alongside it:
Vector Prime (tagged
Prime),
which adds a multi-factor AI read, and
Vector Boost (tagged
Boost), an experimental machine-learning model. All three
are shown and scored separately, on the same races.
1. What goes into the number
Each runner is scored from a small set of signals we trust, and the field is then turned into win
and place chances:
1
The live market — the
crowd's collective opinion, and a genuinely hard thing to beat. Core starts from it rather than
pretending it isn't there.
2
Our own rating — the
VectorOdds composite rating of each runner, computed here from its record. Our figures, not a third
party's ratings resold.
3
Recent form — how the
runner has actually been going, rather than how it reads on paper.
How much each signal counts is not a fixed recipe.
It is learned from settled results and moves as the evidence moves — which is the
point of the next two sections.
2. What that gets you
a
The whole card, not a shortlist.
Core runs on every runner in every race — no cherry-picked "tip of the day", no quiet skipping of the
races that are hard to call.
b
A price, not just a name.
A tip tells you who. A fair-odds line tells you whether the price is worth taking — which is
the part that decides whether a bet is any good.
c
A record you can check.
Core's strike rate and return are published alongside the other flows on the same settled races, so
you can see how it is actually going rather than take our word for it.
3. It re-tunes itself on real results
Core isn't frozen. As races settle it re-checks how much weight each signal has earned, and adjusts
— but only ever on evidence, and never without a way back:
a
A change has to prove itself first.
Nothing goes live because it looks better in theory. It is tested against real results before it
touches your predictions.
b
And it's watched afterwards.
A live guardrail follows what happens next. If the strike rate slips after a change, it
reverts automatically.
It can get better; it can't quietly get worse
That second step is the one that matters. Plenty of models improve on paper and drift in practice.
Core's blend is allowed to move only in the direction the results support, and anything that hurts
is undone without waiting for someone to notice.
Where Core sits
Core runs alongside Prime, which adds a multi-factor AI read of each runner, and
Boost, an experimental machine-learning model. All three are different approaches to
the same question, scored on the same races — not rungs of a ladder. Seeing where they agree, and
where they don't, is often the most useful thing on the card.
Read up on the other two:
How the Prime predictions work and
How the Boost predictions work.