The Math · 算账

Run the numbers on a UPoker bot

No "is it real" hand-waving — just the arithmetic. EV by stake, the hand count to repay the software, and the rakeback (返水) line that decides whether the sheet is black or red.

1 · Player-side EV by stake — no rakeback (solver edge minus rake)
Stake (blinds)	bb size	Edge bb/100	Rake bb/100	Net bb/100	Net / 1k hands
NL20 · 0.10/0.20	$0.20	3.0	6.5	−3.5	−$7.0
NL50 · 0.25/0.50	$0.50	2.8	5.8	−3.0	−$15.0
NL100 · 0.50/1.00	$1.00	2.5	5.0	−2.5	−$25.0
NL200 · 1.00/2.00	$2.00	2.2	4.4	−2.2	−$44.0

Estimates Edge = a steady solver overlay vs. soft club opponents; rake = typical heavy club rake. Without rakeback, the net line is negative at every stake — the bot loses money on the cards alone.

2 · Same stakes with rakeback (返水) added back — the agent's sheet
Stake	Rake bb/100	Rakeback %	Rakeback bb/100	Net w/ RB bb/100	Net / 1k hands
NL20	6.5	40%	2.6	−0.9	−$1.8
NL50	5.8	45%	2.6	−0.4	−$2.0
NL100	5.0	50%	2.5	0.0	$0
NL200	4.4	55%	2.4	+0.2	+$4.0

Estimates Add the rakeback the agent collects and the sheet creeps from deep red toward break-even — and only above NL100 does it tip positive. That tiny margin is the entire business. A renter who collects no 返水 stays on table 1.

Break-even: how many hands to repay the software?

Suppose a seller charges a monthly fee for a UPoker bot. Break-even is simply the fee divided by the net win-rate per hand. With a positive rate the answer is a hand count; with a negative rate there is no break-even at all — more hands just lose more.

Hands to break even = monthly fee ÷ (net win-rate in $ per hand)

If net win-rate ≤ 0, the denominator is zero or negative — break-even never arrives. This is the case for every player-side row in table 1.

3 · Break-even hand count vs. a $300/mo software fee
Scenario	Net bb/100	$ / hand	Hands to repay $300	Months @ 30k hands/mo
Renter, NL100, no rakeback	−2.5	−$0.025	never	never
Agent, NL200, +0.2 w/ RB	+0.2	+$0.004	75,000	~2.5
Agent, NL200, +0.6 (better RB)	+0.6	+$0.012	25,000	~0.8

Estimates $300/mo is an illustrative fee. The renter row never repays because the win-rate is negative before the fee is even counted. The agent rows repay only because rakeback drags the rate above zero.

What the three tables say together

Read top to bottom, the arithmetic is blunt. On the cards alone (table 1) a UPoker bot is a losing machine at every club stake — the solver edge is real but smaller than the rake. Add rakeback (table 2) and the line only reaches break-even around NL100 and barely clears zero above it. Run that through a software fee (table 3) and the renter never repays while the agent repays only on the thin rakeback margin.

That is the whole thesis in numbers: a UPoker bot is not a poker edge you can buy, it is a rakeback position the agent already holds. The same structure governs PPPoker and ClubGG — server-dealt cards, a 代理 at the top, settlement in diamonds the operator can freeze. GTO and solver math set the edge term; rakeback sets the sign. To see whose pocket those diamonds end up in, read Who Profits.

Have real rake, rakeback, and fee numbers from an offer you're weighing? Send them to the chat below and we'll run them through the same sheet.

Raul Moriarty

Poker Software Expert · Poker Bot AI

Tracks the software side of club-app poker — automation, detection, and the gap between what sellers promise and what code can actually do.

Bring the numbers

Send us the rake, rakeback, and fee

Have real figures from an offer? Send them over and we run them through the same sheet on these pages — whether the math lands black or red, and where the diamonds end up.

Talk to us on Telegram