Practice: Know what your mistakes actually cost

Jan Przysucha10 SEP 2026~7 min read

Every player I know can tell you their worst hand of the year.

Nobody can tell you what it cost them.

Not “I got coolered” — an actual number. How many big blinds that call gave away. That is the thing I wanted for years and never had, and it is why Practice exists in the shape it does.

You play the whole hand. Not a quiz question.

Most trainers stop exactly where poker gets hard. Preflop spot, pick an action, tick or cross, next.

That is not poker. That is flashcards.

In Practice you get dealt into a seat with a stack, blinds and antes. Players in front of you act. You make your preflop decision — and if the hand goes on, you see a flop, a turn, a river. The pot, the stacks and both ranges carry forward from what actually happened, and the opponent answers from solver strategy at every node.

A hand in Practice — nine-handed table, hero in the big blind facing a small blind raise, no solver hints on screen. — Nine-handed, 50 bb, big blind facing a raise. No hints — this is the mode you should be training in.

Every number on that felt checks out. Ante 1 bb, blinds 0.5 and 1 — that is 2.5 in the middle. Small blind raises to 4, so the pot is 6. You owe 3 more.

Pot odds: 3 / (6 + 3) = 33.3%. Exactly what the panel says. Do that arithmetic yourself — I want you to.

Postflop, every bet size is named as a percentage of the pot with the big blinds worked out for you, so you always know what a size actually costs before you press it.

Postflop action row — Check, Bet 15% (1.3 bb), Bet 30% (2.7 bb), Bet 60% (5.4 bb), Bet 120% (10.8 bb). — Every size as a percentage of the pot, converted to big blinds.

Then it tells you the price

This is the part I care about most.

Feedback card after a decision — Blunder verdict, 0.19 bb of EV loss, and the solver's full strategy at that node. — Blunder — 0.19 bb of expected value given away, and the strategy that says so.

You are not told you were wrong. You are told what it cost you: 0.19 bb.

Best, correct, inaccuracy, mistake, blunder — and the grade comes from how much EV you gave up against the pot, not from a list of approved buttons. A call that costs a hundredth of a big blind and one that torches half the pot are not the same error, and any trainer that paints both red is teaching you nothing.

Your range calls. Your hand folds. Both are true.

If you take one thing from this post, take this one. I have got it wrong myself more times than I want to write down.

The same hand with the Range Explorer open — the big blind's 169-hand matrix, continue 60.7% across 805 combos. — The range continues 60.7% of the time. That is the number people mistake for permission.

Facing that raise, the big blind range continues 60.7% of the time across 805 combos. Calling looks fine. But the hand was J♥6♣, and for that combo, at that node, the solver folds 100% of the time.

Both are correct — they answer different questions, and swapping one for the other is how you talk yourself into calls a chart seems to bless. So we put them side by side in plain words: Solver, with this hand and Solver, across the whole range.

You decide what gets dealt

Random hands are a waste of your evening. So you tell it what to deal before you sit down.

The session composition dialog — format, stack range, simulation type, ten preflop categories, hero and opponent seat, starting hands and nine board textures. — One dialog: format, depth, ten preflop categories, both seats, 169 starting hands, nine board textures — and Save as drill.

Cash, MTT or Spin & Go. Any stack range. Ten preflop categories — opens, facing a raise, three-bets, squeezes, blind versus blind, limped pots. Pin your seat, your opponent’s, or neither. Narrow to board textures, or to specific starting hands if a class of holding keeps butchering you. Mine is offsuit broadways out of position.

Play this spot instead of reading about it

The trainer deals the situations this article describes and grades every decision. Three-day trial.

Start Free Trial

The other half of the cockpit is how it behaves.

How it plays	Your options
Hands	5 · 10 · 20 · 50 · no limit
Feedback	Study · After · Errors only · Cheat
Clock	Off · 8s · 12s · 25s
Opponent	Samples solver frequencies · always most frequent
Villain cards	At start · at end · showdown only
Tables at once	1 · 2 · 4

That last row is the one people miss. Practice can deal four tables at once, each running its own hand.

Four tables at once in Practice — four independent hands, each with its own Fold / Call / Raise / All-In row. — Four hands at once, four independent action rows. Every decision still graded.

Not a gimmick — the closest thing to the pressure of a real session. Multi-tabling costs you the same EV here as it does online, except here somebody counts it.

Four ways to be told you are wrong

Only After and Errors only count towards your ranked stats. A decision made while staring at the answer measures nothing, and I am not going to let the app flatter you.

Turn your leak into a machine

Anything you can configure, you can save as a drill and start with one click. Blind versus blind at 20 bb. Facing a three-bet in position, 100 bb deep, paired boards.

Every drill keeps its own record — runs, dates, accuracy — so the list stops being a menu and becomes a scoreboard. Mine says blind-versus-blind went 33% → 44% → 62%, and that the cold four-bet drill I built a month ago is still on 0% because I have been avoiding it. Uncomfortable in a useful way.

There is a catalogue of ready-made drills next to your own, so you do not have to build them all yourself. Plans go further: multi-week structures, six sessions a week, each day with its own setup waiting. If you have ever opened the app thinking “no idea what to work on today” — that is what this is for.

Mark the hands that bug you

One key during a session flags a hand for later. They pile up in Marked — every hand you pressed M on, across every session you have played.

The Marked screen — flagged hands grouped by day, with filters, a pattern panel and a grade breakdown. — Your cards, the action that led there, your decision beside the solver’s, and what the difference cost.

Two things there are smarter than I expected when we built them.

It looks for a pattern in what you flag. If five of your marks are big blind facing an open, it says so — and offers to deal those five back to back. Your instinct about which hands were interesting turns into a drill without you noticing the theme yourself.

And it splits them by grade. Plenty of marked hands are not errors: nine of my twenty-eight were correct plays I still wanted to see again. The screen says it plainly — marks are yours, grades are the solver’s. It also counts how many you actually went back to. Mine says 15 of 28.

Opening one gives you the replayer: the hand street by street, your decision at every node, the verdict, and the solver strategy beside it — for your exact combo and for the whole range.

The hand review screen — replayer on the left, the chosen decision with the solver matrix in the middle, and the full 15-hand history on the right. — The replayer, the decision you made, and every hand of the session — sortable by what it cost.

Playing 400 hands teaches you almost nothing by itself.

Playing 400 hands, flagging the 11 that confused you, and then actually sitting with those 11 — that is a study session. The first is just clicking.

And then the part that hurts

Every graded decision goes into your record. Seven days, thirty, ninety.

The Performance screen — accuracy and EV loss headline figures, weakest measured track with a drill button, and trends over thirty days. — Accuracy, EV loss per decision, and the track you are worst at — with a button that drills it.

Two headline numbers: accuracy, and average EV loss per decision. Then the breakdowns that actually find the leak — by position and by street. That is how you learn your flop play is fine and your preflop is quietly bleeding.

It also names your weakest measured track, tells you how far below your own average it sits, and puts a button right there that drills exactly that. When it stops being your weakest, the app tells you that too.

One promise about the numbers

Every frequency, every EV figure, every verdict comes from solved solutions, read at the node you are standing on. No lookup table of “typical” frequencies. No heuristic wearing a solver costume. And when there is no solved data for a spot, the app says so instead of inventing a number.

A trainer that always has an answer is a trainer that is sometimes making it up — and you would never know which times.

Start here

Open Practice. Leave the defaults. Set feedback to After. Play 20 hands. Ten minutes.

You will finish with a verdict on every decision and a number for what the session cost you. Then read the review, mark what surprised you, and turn the recurring one into a drill.

Play, review, narrow, repeat. That loop is the entire product.

Go find out what your calls actually cost. Cheers! 🙂