How we verify
Last updated 27 September 2026 · applies to solvedspot.com
Every strategy number in this trainer carries one badge saying how much you should trust it.
The badge is derived from where the number came from, not chosen by hand and not by the screen
you happen to be on — so the same number reads the same everywhere. The one rule underneath all
of it: nothing is called Solved unless its distance from optimal was actually measured.
When we cannot measure that, we say so rather than rounding up.
Solved
From a game-theory solve of that exact spot, with a measured exploitability — the number
of chips per hand a perfect opponent could win against this strategy. The solver runs until that
figure is small, and the figure is shown alongside the answer, so "solved" is a claim you can
check rather than take on faith. This is the standard the river and turn solves meet: with the
board complete, the outcome is enumerated rather than sampled, so there is no margin of error to
hide.
Cross-checked
A Solved answer that was also matched, node by node, against an independent solver on a
verified-identical game tree, and agreed within a few tenths of a percent of the pot. This is the
strongest badge: two engines that share no code arrived at the same strategy. Where you see it,
the number has been checked twice, two different ways.
Estimated
Computed, but not to the Solved bar. Two things land here:
-
The push/fold charts. These are a real heads-up shove-or-call equilibrium, reached by
iterating a solver over sampled equity — but their exploitability was not measured, only their
convergence (the strategy stops moving to within a fraction of a percent). They come with real
EV per hand, and they are the one preflop spot graded on EV loss rather than frequency alone.
Genuinely computed, genuinely game-theoretic at the edges — just not verified to the point where
we would call them Solved.
-
Flop and turn decisions in a hand you played out. With cards still to come, the result is
measured by playing the spot forward many times, so the answer carries a small margin of error
(a couple of chips per hand). That margin is shown; a difference smaller than it is inside the
noise, not a real edge.
Estimated sits one step below Solved and one above Heuristic: it was calculated, not hand-written,
but you should read it as "very likely right" rather than "proven".
Heuristic
Not from a solve at all. Two things land here too:
-
The 100bb preflop charts. Hand-authored approximations of well-known published ranges.
They are a good study default and they are honest about being one — they carry no EV, they have
not been compared against a solver, and grading against them is by frequency, which the app tells
you when it does it. This is the single biggest gap between this trainer and one that could claim
full preflop GTO accuracy, and it is labelled as such on purpose rather than dressed up.
-
Whole-hand play graded against the training opponent. When you play a full hand, postflop
decisions are scored by what they cost against the specific opponent at the table — the answer to
"what does this line cost against this player", not "what is GTO". Switch the source to solved
spots when you want the equilibrium answer.
Why label at all
Because a study tool that presents a hand-authored guess and a cross-checked solve in the same font
teaches you to trust both equally, and one of them does not deserve it. The badges exist so you know
which parts of the trainer you can lean on hard and which parts are a reasonable default you should
keep questioning. When a number is better verified later, its badge moves up — never the other way
around without the evidence to back it.
Found something that looks wrong?
Every graded spot has a "this looks wrong" control. If a number looks off, flag it there — it records
the exact spot so it can be reproduced and checked. That is part of how numbers earn a better badge.
Back to the trainer