Every betting system you can run on Stake, with the bankroll it demands and the drawdown it produces. No system beats the house edge — what they change is the shape of your results, and that is worth understanding before you automate one.
Stake's original games run at a 1% house edge — 99% RTP. Every dollar you wager returns 99 cents on average, and that figure is set by the payout table, not by how you size your bets. A betting system decides the order and the size of your wagers. It never touches the 99.
This matters more with a bot than without one. Automation multiplies your bet count, and the more bets you place, the closer your realised result lands to the expectation. Variance is what lets a human walk away up after 200 bets; at 5,000 bets it has mostly cancelled out. Flat betting a dollar at 600 bets an hour costs about six dollars an hour, and no progression changes that arithmetic — it only decides whether you lose it smoothly or in one sitting.
So the useful question is not which system wins, it is which distribution of outcomes do I want. Do you want many small wins and a rare wipeout, or a predictable slow bleed? That choice is real, and it is the one the rest of this page is about.
Bots don't tilt. SSPilot follows your strategy exactly, never deviating on a bad streak or chasing losses beyond your limits.
Every bet is calculated and placed instantly. No delay, no miscalculation, no accidentally placing the wrong amount.
Track exactly how each strategy performs over thousands of bets. Make informed adjustments based on real data, not gut feeling.
Don't have a Stake account yet? Sign up to get SSPilot free.
Sign up to StakeDouble after every loss, reset on a win. Each win recovers the whole losing run plus one base bet. It wins often and small, then loses everything at once — the losing session is the one where the run outlasts your balance.
Double after every win, reset on a loss. Losing runs cost one base bet each, so the balance bleeds slowly instead of collapsing. You need a streak to make anything back, and you must decide in advance where to cash the streak out.
Add one unit after a loss, remove one after a win. The progression is linear instead of exponential, so a bad run costs multiples of your unit rather than powers of two. It survives far longer than Martingale and recovers far more slowly.
Same stake every bet, whatever happened before. It is the only system whose loss rate matches the house edge exactly, which makes it the honest baseline: anything that beats flat betting over a large sample is doing so through variance, not edge.
Your own percentages on win and on loss, so you can sit anywhere between flat betting and Martingale. Useful for testing a hypothesis — but only if you record enough bets to tell the result apart from noise.
A profit target and a loss limit, enforced by the bot rather than by you at 3am. This is the one setting that changes outcomes reliably: it does not improve your edge, it caps the size of the session that goes wrong.
Martingale is the system people underestimate, so it is worth putting numbers on it. At a 49.5% win chance — the standard dice setting for a 2× payout — each bet loses 50.5% of the time. Below is how often a run of consecutive losses shows up, and the reserve you need to still be doubling when it ends. Read the last column as the bet you must be willing to place, in multiples of your opening stake.
| Losses in a row | How often | Bankroll needed | Next bet |
|---|---|---|---|
| 5 | 1 in 30 | 31× | 32× |
| 8 | 1 in 236 | 255× | 256× |
| 10 | 1 in 926 | 1,023× | 1,024× |
| 13 | 1 in 7,174 | 8,191× | 8,192× |
| 15 | 1 in 28,209 | 32,767× | 32,768× |
A 10-loss run turns up about once in 926 bets. An automated session places several thousand — so it is not a tail risk you might avoid, it is an event you should expect several times an evening. That is the whole argument for a loss limit: it decides in advance which run you refuse to keep doubling into.
Take the longest losing run your progression can absorb, and check what fraction of your balance that run consumes. If the answer is more than half, your base bet is too large — no matter how comfortable it feels on bet one.
A limit you hold in your head is renegotiated the moment it is hit. A limit in the stop rules is not. This is the single change that most reliably alters how a session ends.
A few hundred bets tell you almost nothing: the noise is wider than any effect you are looking for. Thousands of bets, recorded and compared against flat betting on the same settings, tell you something real.
Reloads, drop codes and rakeback are the only parts of Stake that can carry positive expectation. They come from claiming, not from the progression you run afterwards — keep the two in separate columns when you count what a month made.
This page is the map. Each guide below takes one part of it and works through the numbers properly.
No, and any page telling you otherwise is selling something. Stake's original games run at a 1% house edge, which means every bet returns 99 cents on the dollar on average. A betting system changes the shape of your results — how often you win, how large the swings are, how long a session lasts — but it multiplies the same negative expectation. What a strategy can do is control variance and enforce discipline, and that is worth having.
Martingale, by a wide margin — and it also has the worst losing session. It converts many tiny wins into one catastrophic loss, so most sessions end green and the rare red one wipes out everything the green ones made. Flat betting does the opposite: sessions end close to the expected loss, with no surprises in either direction.
More than most people budget for. Doubling means a run of n losses costs 2ⁿ−1 base bets before you place the recovery bet. At a 49.5% win chance, a 10-loss run happens roughly once every 926 bets and requires 1,023 base bets in reserve. Run a bot for an evening and you will see several thousand bets, so plan for the run you will hit, not the one you expect.
Kelly tells you what fraction of a bankroll to stake given an edge. With a 1% house edge, that edge is negative, and the Kelly fraction of a negative-edge bet is zero — the maths says bet nothing. Kelly becomes useful on Stake only when a bonus, a reload or a rakeback push a specific wager into positive expectation.
It removes the two failure modes that have nothing to do with maths: deviating from the plan after a bad run, and miscalculating a bet size. It also multiplies your bet count, which brings your realised loss closer to the theoretical expectation — variance protects you less at 5,000 bets than at 50. Bet count is why stop rules matter more with a bot than without one.
Dice and Limbo, because you set the win probability yourself and it stays constant across every bet — which makes the outcome sequence comparable and a backtest meaningful. Mines and Plinko change their payout structure with each configuration, so a strategy tested at one setting says nothing about another.
Download SSPilot and execute strategies with perfect precision. Track results, optimize your approach. Free to download.
Download SSPilot Free