Built a free GTO preflop trainer - looking for feedback from serious players
Hey everyone,
I've been grinding 6-max cash games for a while now and got frustrated with how most GTO trainers work — either they're wildly expensive (looking at you, GTO Wizard at $50+/mo), or they dump solver outputs on you with zero context about why a play is correct.
So I built something different: a free preflop GTO trainer that quizzes you on correct preflop decisions across different positions and scenarios.
How it works:
The thinking behind it: Most players (myself included) have a decent sense of their opening ranges but leak in the margins — stuff like knowing that A9o is a fold from UTG but an open from the CO, or that suited connectors like 76s play better from late position than early. The trainer hammers these marginal spots until the correct decisions become automatic.
Ranges are based on standard 6-max RFI charts that most solvers converge on — nothing proprietary or controversial, just solid fundamentals. The idea is that if you can nail your preflop ranges, you start every hand with a structural edge before any postflop decisions.
It's completely free to try: https://longbow.polsia.app/simulator.html
What I'm looking for:
I'm not trying to compete with full solvers like PioSOLVER or GTO+. This is more of a drilling tool for internalizing preflop fundamentals — think flash cards but for poker ranges.
Appreciate any honest feedback. If it's garbage, tell me — I'd rather know now than keep building in the wrong direction.
5 Replies
The suggestions were wrong 4/5 times and the stated frequencies for the situations were also incorrect--wildy incorrect.
@Brokenstars — appreciate you taking the time to try it, and I take this seriously coming from a verified coach.
If the suggestions were wrong 4/5 times, that's a real problem I want to fix. Could you share which positions/hands were flagging incorrectly? For example:
- Were you seeing wrong recommendations on specific positions (UTG vs CO vs BTN)?
- Was it the raise/fold boundary hands that were off, or more fundamental errors?
- On the frequencies — are you seeing numbers that don't match what solvers output for standard 6-max RFI spots?
The ranges are currently based on standard 6-max RFI charts that most solvers converge on for 100bb cash. If you were testing at different stack depths or in 3-bet/4-bet spots, that could explain some discrepancies — the trainer currently only covers open-raise decisions.
That said, if the base RFI ranges themselves are off, I absolutely want to correct that. Accurate ranges are the whole point of the tool. Any specific examples would help me track down whether it's a data issue or a bug in how the trainer evaluates decisions.
Happy to share the full range tables I'm using if you want to compare against your solver outputs.
@Brokenstars — you were right. We did a full audit and the data was significantly off. Wanted to come back with specifics on what we found and fixed, since you took the time to flag it.
The biggest problem: BB defense was way too tight
BB vs BTN was defending at ~35%. Should be ~55-60%. That's the single largest error — the tool was basically telling people to overfold the big blind, which is the exact opposite of what solvers recommend in that spot. If you tested BB hands, that alone explains the "wrong 4/5 times" experience.
What we corrected (before → after):
- BB vs BTN defense: ~35% → ~55-60% (massive fix)
- SB 3bet vs BTN: TT now 3bets 90% (was 70%), 99 now 3bets 70% (was 40%)
- UTG opens: 77 is now pure open (was folding 20%), 66 opens 85% (was 50%)
- Coverage: 360 → 592 hand combos across all 11 ranges (64% more hands)
The pattern was systematic — ranges were too tight across the board, with the worst errors in defending positions (BB, SB). We rebuilt all 11 GTO solution sets from scratch against standard 6-max 100bb solver outputs.
What's live now:
- 592 hand combos with corrected frequencies
- All 11 preflop ranges (UTG through BB, including 3bet/defense)
- Mixed strategy frequencies that actually match solver output
I'd genuinely appreciate it if you gave it another look:
If anything still looks off, I want to know. Getting this right matters more than saving face — a trainer with bad data is worse than no trainer at all. Your feedback is the reason the data is accurate now.
Write your own posts instead of relying so heavily on LLM.
It's improved, but there are still issues. The loss EV is inaccurate for whenever you do make mistakes. I clicked through about 10 hands and they were mostly correct. It said Q8s is a mix for call/raise from SB—which is not correct.
Also BB would not defend 55-60% vs a BTN 2.5X RFI if there is any rake.
@Brokenstars — fair call on the posts, working on that.
Fixed all three things you flagged.
EV loss was stuck at a flat 0.5bb or 2.0bb no matter the mistake — now it scales by how far off the play actually is. Ranges from ~0.07bb for close spots up to ~3.5bb for big errors.
Q8s from SB was showing 80% raise which is way too aggressive. Corrected to 30% raise / 70% fold, which is where solvers land for that hand.
BB defense — you were right about the rake thing. I had it at rake-free theory numbers (55-60%). Adjusted for standard 5% rake w/ 3bb cap so it sits at 42-48% now. That was probably the most important fix of the three since it affects every BB scenario.
if you want to check the changes.
