3UpGaming Bot Review: $1,800 for a “Pro” AI That Plays Like a Fish

TL;DR

A buddy of mine decided to try out a new player on the poker bot market — a "real AI bot running on poker AI + GTO that plays like top regs" — and paid $1,800 for it. Everything seemed serious and trustworthy, however, after a series of test runs (hand histories were scraped from all sessions + video recording was kept), non-stop patching of junky bugs from 3Up Gaming, we received a bot that plays worse than any fish and simply flushes your stack down the toilet — overall EV winrate: -87bb/100 hands and worse over a volume of ~10k+ hands.

I can say with confidence that this is just another well-hyped scam product on the poker bot market. 3UpGaming's goal is to sell you the bot immediately (2-3+ units for the max price) which looks like a working product, but in reality, is just an imitation. You will only realize this after a couple of days of testing, while the 3UP manager keeps gaslighting you along with promises of updates. Stay away.

I know 3UpGaming managers might try to claim this review is fake. If anyone—potential buyers or developers—wants to see the unedited video recordings or the full database of hands, email me at andrew@poker-ai.org. I have nothing to hide.

The Red Flags

I rarely do these breakdowns of other people's products, but I feel obligated to warn anyone wanting to buy a poker bot in 2026 about this scam. My friend fell for it (and he is far from a fool), so the chance of throwing a couple of thousand dollars into the void is high. Let's get to the point.

The 3UpGaming project appeared in 2021. In 2024, judging by the Web Archive, they offered iGaming services — you buy a ready-made poker app that is customized for your brand (so-called poker skins). The contact info listed was American — phone number, email, Skype. Okay, you can find a presentation of their services on the internet, I have no questions there.

In mid-2025, global changes occur — the site is significantly reworked, the CMS changes, the text changes, new sections and services are added. It’s visible that a good SEO specialist took over the site, and copywriters are writing mountains of AI-generated articles for them weekly (processed via humanizer services, but after reading them it’s clear it’s all fluff). The site is actively promoted, and backlinks are being bought. Services for poker AI robots (bots) are added, as well as a service for club owners to fill tables with their AI bots.

That’s when my friend saw them on the internet, studied the site, and messaged Alex from 3UP Gaming on Telegram. So, here is what they promise:

“fully automated bots that are a mix of GTO and AI”

“They play exactly like a professional player”

"2–3 bots sit at a table, see each other's cards and play team play"

“for each robot we give you a VM machine and a unique IP so that the percentage of robots being banned reaches zero.”

“Robots are trained and updated by AI every day.”

Setup of the bot for any poker room and poker format (Hold’em, Omaha).

And all this — for $1,200 setup + $600/month for each bot. Alex sent a few demo videos where their bots are playing on ClubGG.

Sounds not bad, right? But upon requests to show any HH (Hand History) graphs, game results, or successful cases, we get the answer:

“Most companies create proof for customers and you don't understand whether this proof is for the robot or the player. The best proof for you is to test the robot and use it.”

When we asked for a trial:

“Unfortunately we only have 1 hour test. Our robots cost a lot to configure. We can't prepare anything without money. We are not responsible for your bad experience. Please think carefully and then message me to rent robots.”

The maximum they can offer is a 1-hour test where we can watch their bots play against each other on ClubGG. And of course, they are "not responsible for your bad experience." Doesn't sound great, but given the specifics of the poker bot market, and considering they are newcomers — this is somewhat acceptable.

Anyway, the money was paid, and 5 days later pdf-instructions were issued for connecting to the VM via Web-VNC and launching the bot.

How we tested the AI "robot" from 3UP Gaming: Methodology

• Room: PPPoker
• Limits: NL4 cash games, 6-9-max tables

At the tables — a classic zoo of micro-stakes players:

• Newbie fish
• Amateurs
• Old rule-based bots like Warbot, Shanky

There are no regs, poker sharks, or advanced poker AI bots here that profile players and find exploits. But there are cheap bots running on various profiles that can show +EV at micro-stakes.

The VM already had the PPPoker client installed, the whole environment was set up, we just needed to sit at the table and run the poker bot's .exe file — from there it does everything itself. In the bot's terminal window, logs were visible — OCR was recognizing cards at the tables, balances, buttons, and interface elements, sending all data to their "AI", and then making an Action.

For game analysis and reviewing stats, we recorded the entire hand history using Hand2Note 4 and video recorded the entire gameplay — from start to finish.

Analysis of 3UpGaming Bot Sessions

At first, we received a raw product — their bot froze, didn't take actions, lost connection with their servers, and did weird things. After a detailed description of the bugs to Alex, an updated stable version was released 2-3 days later.

Session 1: The "Stable" Bot

• ~2,100 hands, NL4 PPPoker.
• Actual Result: about −$144.6.
• EV: about −$121.8.

Stats:

• Winrate: ~−309 bb/100
• EV Winrate: ~−251 bb/100
• VPIP / PFR / 3bet: 44 / 19 / 17
• AF (Aggression Factor): 0.6
• WWSF: 17%
• Won w/o SD: ~−245 bb/100
• WTSD / W$SD: 18% / 58%

Putting emotions aside, we get a bot that simply donates chips to opponents. And quite quickly. After a logical question to the 3UP Gaming manager: "WTF?!", we received assurances that their AI had updated, try tomorrow (it updates and adapts daily), and everything will be fine. The bot didn't have a good strategy, entered every second pot, called often, yet had a 3-bet of ~17%; the bot immediately surrendered after any aggression, even weak aggression, even with good cards. By then it became clear that things wouldn't be "fine," but there were some glimmers of hope (I apologize for any possible minor inaccuracies in the stats – I took the screenshots after writing the draft).

Session 2: Updated "Poker AI"

• ~2,300 hands, NL4 PPPoker.
• Actual Result: about −$69.13.
• EV: about −$45.99.

Stats:

• Winrate: ~−87 bb/100
• EV Winrate: ~−58 bb/100
• VPIP / PFR / 3bet: 22 / 8.7 / 8
• AF: 0.5
• WWSF: 15%
• Won w/o SD: ~−68 bb/100
• WTSD / W$SD: 17% / 57%

The second version turned the bot into a less maniacal fish. Aggression is noticeably toned down, but the gap between VPIP and PFR remains large. Too many calls, few raises and isolates.

The bot plays loose but has zero postflop initiative. WWSF is 15%, whereas for regulars it's 40-50%. The AI "robot" doesn't fight for pots, it gives up and folds. In both sessions, the EV graph goes steadily and predictably down. It’s not a question of whether the 3upgaming poker bot is running bad or not, a systematic approach is clearly visible where the bot drives itself into negative situations and surrenders.

Other flaws of the 3UPGaming bot:

1. The bot aggressively raises by 1/2, 2/3, 3/4 of the stack — regardless of the pot size or stack depth (whether 50bb or 500bb). Meaning, the bot does not account for or rely on the limit, on others' raises and calls — it is hard-coded to raise a certain percentage of its own stack.
2. The bot cannot auto-reload the balance, for example at =0 or when it drops below n-bb.
3. The bot is very easily interrupted by various pop-up messages from the rooms — these exceptions are not handled at all.
4. The bot often loses connection (Socket error, socket exception, invalid socket response) and mindlessly folds and sits out.
5. The bot FOLDS in spots where it can check. Imagine this — flop, no one raises, you can calmly check and see the next card. The bot clicks "Fold", gets a question from PPPoker: "Are you sure you want to Fold???", and the bot confidently clicks "YES". Insane! They fixed this elegantly — now the bot simply clicks Cancel on the question from PPPoker. This clearly speaks to the "professionalism" (amateurism) of the 3Up gaming developers.
6. On the virtual machine provided by 3up gaming, free space constantly runs out, the game is interrupted by Windows update messages, offers to backup the system, and many others.
7. Control was often lost on the VM — you couldn't move the mouse, although the server was active. The only solution — "press Reset and reboot the remote computer."
8. They claim that "each robot we give you a VM machine and a unique IP so that the percentage of robots being banned reaches zero". Yeah, except it's a Hetzner data center IP in Finland, which immediately tells the security team of any poker room that this is a bot playing, not a human.

How do you like that? Does it look like a "professional AI-GTO bot" for $1800? Moreover, the managers strongly insisted that it is necessary to seat 2-3+ bots at a table ($600/month for each) to have a confident profit. Because they understand that after one payment, there will be no continuation.

Why 3UP Gaming Robot is not "Poker AI", but an expensive fish-bot (My theories)

~4400 hands were played on a stable micro-limit field. For poker math, this isn't a couple of hundred thousand hands, but it's enough to see the "outlines" of the strategy and evaluate how the bot works — what it does in various situations, and to look at the trend of the EV line. If someone looks like a fish over 4k hands — that is already a signal, not noise.

I could assume that this is a quick student project slapped together: OCR + sending game state data to some publicly available LLM -> receiving an answer and executing the action. But even LLMs have a better strategy than the 3UP gaming poker bot. There is no GTO, no "Neural Network." It’s a simulation of a product designed to milk the marks.

Possibly, the 3UpGaming project changed hands that invested in promoting the site and creating new "products." Specifically described as a high-end AI solution with a high price tag to maximize earnings. But these are just my guesses.

Save your $1,800 and invest it in a coach or a solver instead. Stay away from 3UpGaming.

P.S. I wrote the drafts of this article a long time ago but didn't publish them. At the moment, the bot has gone through several stages of "updates" solving problems, the "poker AI" improved every day, and in fact, 10k+ hands have already been played, which is statistically more significant. The result... EV winrate -127bb/100 hands.

Source URL: How I bought a GTO+AI poker bot for $1,800 that plays like a Fish