Author: Chen Zhiyuan | Date: 2026-08-15 | Read: ~16 min
3 years running BaccAI, I have 200+ paying users. This article picks 3 typical: loser-to-winner 28-year-old office worker, 5-year veteran 45-year-old business owner, 2-year still-losing female teacher. Each has full 12-month account data + real dialogue quotes + stake formula choice analysis. Names changed, data patterns real.
First, the most direct data — pre-AI 6 months vs post-AI 6 months ROI comparison:
| User | Age / Job | Baccarat Experience | Pre-AI 6-Mo ROI | Post-AI 6-Mo ROI | Improvement |
|---|---|---|---|---|---|
| Xiao Wang | 28 / Office worker | 1 year | -38.2% | +52.3% | +90.5 pp |
| Lao Li | 45 / Business owner | 5 years | -15.4% | +89.4% | +104.8 pp |
| A Zhen | 35 / Teacher | 2 years | -52.1% | +34.2% | +86.3 pp |
3 users commonalities:
📋 Background: 15k monthly salary, 1 year baccarat, loses 3000-5000/month
Pre-AI (2025-09 to 2026-02, 6 months)
Xiao Wang added me on WeChat last September, 28 years old, internet company operator in Hangzhou, 15k monthly, played baccarat for 1 year, mainly on mobile baccarat APP.
Pre-AI 6 months, Xiao Wang's account record (from his shared screenshots):
Xiao Wang's mindset at the time (original quote):
Xiao Wang (2026-02-15):"Chen-ge, I lost another 4,200 this month, almost 20k total in a year. I know AI can predict, but I'm scared to use it, scared of being scammed. Market AI software costs 5000-8000/year, I'm afraid of paying and still losing."
I showed him the 7/28 37,862-shoe 4-strategy backtest data, then said: try BaccAI free for 14 days first, pay only if you see results.
Post-AI (2026-03 to 2026-08, 6 months)
Xiao Wang started using BaccAI basic (99/month) from March, with the stake formula combo I suggested:
6 months account:
Xiao Wang on 8/12:
Xiao Wang (2026-08-12):"Chen-ge, I won 5,200 in August, my highest single month in a year of baccarat! Key is in June I lost 800, but unlike before I didn't chase losses by doubling to 1,600, I just accepted the 800 loss and restarted. Once that psychological barrier was crossed, everything got smoother."
Xiao Wang's key change: stopped chasing losses (8/1 5-bug AI cure post's "sunk cost fallacy").
---📋 Background: 50-100k monthly income, 5 years baccarat, Macau + online
Pre-AI (2025-09 to 2026-02, 6 months)
Lao Li, 45, building materials business in Guangzhou, 50-100k monthly, 5 years baccarat, Macau 2-3 times/year plus online.
Pre-AI 6 months, Lao Li's data:
Lao Li's pattern pre-AI was "small wins, big losses" — wins 10-20k then leaves, losses 20k without blinking. Classic "win small, lose big" pattern.
Lao Li (2026-02-20):"Mr. Chen, 5 years and I've never made money. You say AI can predict, can it outsmart the casino? But I see your tool has something (he means LSTM 0.65 threshold), I'll try 1 month first."
Post-AI (2026-03 to 2026-08, 6 months)
Lao Li started BaccAI Pro (399/month) from March, using stake formula combo:
6 months account:
Lao Li on 8/13:
Lao Li (2026-08-13):"Mr. Chen, I won 89k in August, my highest single month in 5 years! Before, winning 15k in October was great, now I can win that in a day in August. The key is carry doubling stake — on July 5, when 9-streak triggered, I followed with stake doubling to 800 and won 760. That's when I understood 'compound effect'."
Lao Li's key change: used carry doubling stake (8/12 streak7 carry post) + stopped "small win, big loss" pattern.
---📋 Background: 8k monthly salary, 2 years baccarat, loses 1500-3000/month
Pre-AI (2025-09 to 2026-02, 6 months)
A Zhen, 35, English teacher at a Chengdu middle school, 8k monthly, played baccarat for 2 years, mainly mobile.
Pre-AI 6 months, A Zhen's data:
A Zhen was the worst pre-AI of the 3 at -52.1% (and her salary is only 8k). When I first chatted with her, her eyes were red.
A Zhen (2026-02-25):"Chen-ge, I've lost 13.6k, my husband doesn't know. I can't lose anymore, can you help me? I don't need to make a lot, just not lose."
Looking at her 6 months of accounts, her biggest issue wasn't technical, it was chasing losses (after losing, doubling to recover). She lost 800, doubled to 1,600 to recover, lost again, doubled to 3,200, and after 3 losses her stake hit 6,400 and blew up 5k in one round.
Post-AI (2026-03 to 2026-08, 6 months)
A Zhen started BaccAI basic (99/month) from March, using stake formula combo:
6 months account:
A Zhen on 8/14:
A Zhen (2026-08-14):"Chen-ge, I won 3,400 in August, can't compare to Lao Li's 89k, but it means a lot to me — first time in 2 years of stable profit. I told my husband in August, he didn't blame me, instead said 'wife, you learned control.' I cried for half an hour."
A Zhen's key change: from "chasing losses" to "flat 100" — not a technical upgrade, but a psychological cure.
---Plotting 3 users' monthly ROI trends:
Key observations:
All 3 used BaccAI's stake formula, but chose differently:
| Stake Formula | Xiao Wang (newbie) | Lao Li (veteran) | A Zhen (newbie) | Best For |
|---|---|---|---|---|
| flat 100 | 25% | 15% | 40% | Psychologically conservative / low income |
| Anti-Martingale | 35% | 45% | 30% | Dragon-follow natural fit |
| 1/2 Kelly | 25% | 25% | 20% | Dynamic position (8/8 post) |
| carry doubling | 15% | 15% | 10% | 7+ streak trigger (8/12 post) |
Key findings:
High ROI doesn't mean best risk-adjusted. Max drawdown vs final ROI scatter:
| User | Final ROI | Max Drawdown | ROI / |Drawdown| | Assessment |
|---|---|---|---|---|
| Xiao Wang | +52.3% | -22% | 2.38 | Balanced |
| Lao Li | +89.4% | -38% | 2.35 | High risk high return |
| A Zhen | +34.2% | -18% | 1.90 | Low risk low return (newbie, low bankroll) |
3 users risk-adjusted returns are close (2.38 vs 2.35 vs 1.90), A Zhen lowest because small bankroll caps stake upper limit.
From 3 users 12-month data, I summarize 5 commonalities, for anyone wanting to use AI tools:
3 users M1 all negative (-3% to +8%). This is the learning period, you need time to adapt to AI signal + stake formula + psychological adjustment. Don't give up after losing in M1.
Xiao Wang stopped chasing in May, that month ROI turned positive. A Zhen started flat 100 in March, M1 only lost 300 (vs 1500-3000/month before). The real power of stake formula isn't the formula itself, but you being willing to use it.
Lao Li's 50-100k bankroll amplifies carry doubling stake returns 5-10x. A Zhen's 8k bankroll means carry stake upper limit 800 is 5% of bankroll, limited effect. Small bankroll, prioritize flat 100 + 1/2 Kelly.
Lao Li had 5 years of trial-and-error, knows exactly how much ROI per AI signal hit. Improvement +104.8 pp (from -15% to +89%). Veteran + AI = strongest combo.
Xiao Wang + A Zhen improvements +86-90 pp, but mainly not "AI signal accuracy" but "no more chasing losses". Newbie using AI is 80% psychological cure, 20% technical upgrade. The 8/1 5 cognitive biases post made this clear.
Names changed, data patterns real. I have 200+ paying users' account data, these 3 are typical representatives. Original quotes are desensitized (word changes but semantics preserved). I won't post real names/account screenshots (privacy).
I've been running BaccAI for 3 years, server costs are covered. I don't want to make big money from AI tools themselves (that's scamming). I want user recognition + long-tail traffic + real data feedback. 99-399/month, you don't hurt if you lose, you remember me if you win.
AI tool is not a money printer. M1 learning period, you're not familiar with signal, stake formula, or tool, losing a bit is normal. The key is M2-M3 stable profit, not M1 making big money. 8/8 Kelly deep dive explained small samples show luck, big samples show formula — 6 months is the real verification sample.
Lao Li's 6-month account is real money, not a backtest. Every month I help him check stake formula, adjust max_stake cap, answer questions. I have 200+ users' real account data, will continue sharing in monthly articles.
W5 8/22 preview: BaccAI V4 Roadmap — I'm planning V4 official release in September, new features: dynamic threshold (0.50-0.80 auto-switch) + multi-model integration (LSTM + Transformer + Monte Carlo) + personal stake formula customization. Subscribe to BaccAI monthly to see V4 internal test early.
Chen Zhiyuan, founder of BaccAI. 3 years baccarat AI tooling, 200+ paying users, 3 team members. This user case article is included in BaccAI's official site "Real Stories" section.
Risk disclaimer: The 3 user cases in this article are pseudonymized, data patterns are real but don't constitute investment advice. Baccarat is a negative-expectation game, no AI tool can guarantee profit. Play with money you can afford to lose. ROI data comes from specific users' specific periods, not generally representative.