DEEPFOLD's answer:

DEEPFOLD is the only AI poker analyzer that reads X-Poker (formerly PokerBros) replay links directly. You paste play.xpoker.com/replay/... into the chat and get analysis — no export, no file conversion, no HUD. This matters because club-app players (a huge share of the Asian and LATAM online pools) have no hand history file to export at all, which makes PokerTracker, Hold'em Manager, and essentially every other analysis tool unusable for them. We parse GGPoker / Natural8 PokerCraft share links (gg.gl, my.pokercraft.com) the same way.

Beyond the import moat:

• Screenshots work too. Upload a table screenshot and a two-stage vision pipeline reconstructs the full hand history before any analysis runs — for the tables that don't even have replay links.
• AI on top of a deterministic engine, not instead of one. Pot sizes, equities, blockers, and draws are computed by a rule-based facts pipeline first. The LLM explains the spot from verified numbers rather than inventing them.
• A real GPU solver is included, not sold separately. DEEPFOLD-SOLVER is a CUDA-accelerated desktop GTO solver for Windows, bundled into the $25/mo PRO plan. It ships two things other solvers don't: a Runout Report (every canonical turn card fanned into a 13×4 grid colored by dominant action) and a 1326 Combo Drill (expand any 169-class label into its specific combos with per-combo blocker analysis).
• Natively multilingual. Full UI in English, 繁體中文, 简体中文, 日本語, and 한국어 — plus 65+ localized strategy articles, not machine-translated UI strings.

DEEPFOLD's answer:

Each alternative solves one piece. DEEPFOLD is the only one that covers the whole loop — and often the only one that can even read your hands.

• vs. GTO Wizard — it shows you frequency charts and expects you to interpret them. DEEPFOLD reads your actual hand and explains the leak in plain language. It also can't import from club apps; we can.
• vs. PioSolver / GTO+ / MonkerSolver — powerful engines, but they're a separate several-hundred-dollar purchase, they only solve spots you already know to ask about, and they don't touch your hand histories. Our GPU solver is included in the $25/mo plan alongside the analyzer and the coach.
• vs. PokerTracker / Hold'em Manager — they need an exportable hand history file. If you play X-Poker, PokerBros, or most club apps, that file doesn't exist and the tool is dead on arrival.
• vs. a human coach — $200–500/mo, scheduled, feedback in days. DEEPFOLD is 24/7, per-hand, feedback in seconds, from $0.

One subscription covers hand import from 8+ platforms, AI coaching chat, automated leak detection, GTO/MTT training drills, and a GPU solver. Free tier is 3 batch uploads and 10 AI questions a month — enough to test the import on your own hands before paying anything.

DEEPFOLD's answer:

Online Texas Hold'em players who are serious about improving but aren't served by the existing toolchain — micro to mid-stakes cash game regulars and MTT grinders.

Three groups specifically:

• Club-app and private-game players (X-Poker, PokerBros, QQPK, and similar) who have never been able to use a tracker, because these apps don't export hand histories. They're a large, under-tooled pool concentrated in Asia and LATAM.
• Players priced out of, or frustrated by, the existing options — a human coach at $200–500/mo is out of reach, and solver output reads like a spreadsheet, not an explanation.
• Non-English-speaking grinders. Taiwan, China, Japan, and Korea are core markets. Most poker tooling is English-only; ours is native in all five languages, coaching included.

Skill-wise they range from ambitious beginners who don't yet know what they're doing wrong, to established regs trying to break a win-rate ceiling.

DEEPFOLD's answer:

The problem was simple and infuriating: the tools everyone recommends assume you can export a hand history file. On the club apps where a large share of the world actually plays, you can't. There's a replay link and nothing else — so PokerTracker, Hold'em Manager, and every AI analysis tool are unusable, and you're left trying to recall hands from memory to type into a solver.

So the first thing we built wasn't the AI. It was a parser that reads an X-Poker replay link directly. Everything else grew from there: an AI coach that explains hands in the language you actually think in, drills built on real solver data, and eventually our own GPU-accelerated solver — because we wanted the whole loop, from "I played a weird hand" to "I understand why," to live in one place and cost less than a single hour with a coach.

Built by grinders, for grinders.

DEEPFOLD's answer:

DEEPFOLD is a consumer product — our customers are individual poker players on monthly subscriptions. Users span cash-game regulars and tournament players across Taiwan, Japan, Korea, China, and English-speaking markets.

DEEPFOLD's answer:

DEEPFOLD is not an LLM wrapper. The model sits on top of three layers we built ourselves, and it's the last piece, not the first.

• The parsers came first. Reverse-engineered readers for X-Poker (ex-PokerBros) replay links and GGPoker / Natural8 PokerCraft share links, plus importers for PokerStars, WPT Global, 888poker, PartyPoker, WPN, iPoker, and QQPK. No other AI poker tool can read a club-app replay link — this is the layer that makes the rest possible.
• Then the deterministic engine. A rule-based facts pipeline in Python computes pot math, hand strength, equities, blockers, and draws before any AI runs — covered by golden and property-based test suites. This is why our analysis doesn't hallucinate numbers: the LLM is handed verified facts and asked to explain them, never to compute them.
• Then our own solver. DEEPFOLD-SOLVER is a native C++/CUDA GTO engine for Windows — GPU-accelerated, with a memory-budget system that turns out-of-memory into a structured error instead of a crash. It ships post-solve tooling no other solver has: a Runout Report (every canonical turn fanned into a 13×4 grid by dominant action) and a 1326 Combo Drill (per-combo blocker analysis). Our training drills are built from this engine's own solve output, with path-probability conditioning so every spot is weighted by how often it actually occurs.