Stake dice strategies, explained honestly

Every Stake dice strategy is a rule for how much to bet next and at what win chance. That's it. The differences between them are about one thing: how they spend your balance during losing streaks. Here are the families that actually get used, what each one is really doing, and how to choose.

First, the thing no strategy changes

Stake dice pays out at 99 / win-chance. A 50% bet pays 1.98×, a 10% bet pays 9.9×, a 1% bet pays 99×. That "99" instead of "100" is the 1% house edge, and it is baked into every single roll regardless of what you bet. A strategy can reshape when you win and lose — small frequent wins with rare big losses, or the reverse — but it cannot change the average. Over enough bets, you expect to lose about 1% of everything you wager. If you want the math, read the house-edge breakdown.

So the real question a strategy answers is not "does it win", it's "what does my session look like, and how does it end?"

Flat betting

Same bet every roll. Boring, but it is the baseline everything else is measured against. Your balance drifts down at ~1% of turnover with random noise on top. Bust risk is low unless the bet is large relative to your balance. Flat betting at high win chance (95%+) is how people grind wager for VIP progress and rakeback with minimal variance.

Martingale (double on loss)

Bet at ~49.5% (the 2× payout). On a loss, double the bet; on a win, reset to base. Each win recovers every previous loss in the streak plus one base bet of profit. It feels like a guaranteed +1 base per cycle — and it is, right up until a streak you can't afford.

The problem is how fast the bet grows. From a 0.0001 base:

"1 in 38,000" sounds rare. At 5 bets a second a bot rolls 18,000 times an hour. You will hit a 15-loss streak in an evening, and it costs 65,000× your base bet. This is why a classic martingale always busts eventually — the full explanation is in Martingale on Stake dice.

Delayed martingale

Same as above but you flat-bet through the first N losses (typically 3–6) and only start doubling after that. It cuts the number of doublings you need to survive a given streak, so the peak bet is smaller. The trade-off is that each win now only recovers part of the streak — you're accepting a slower, smaller drawdown instead of a fast one. Most "safe martingale" scripts on Stake are this.

Deep-chance recovery (the 1–2% family)

Bet at a low win chance (1%–5%, paying 20×–99×) and increase the bet slowly on losses — say +3% per loss rather than doubling. You lose almost every roll, but the eventual hit pays back a long run of them. Compared to martingale the bet curve is much flatter: 200 losses at +3% is only a 370× base, whereas 13 martingale losses is 8,192×.

The catch is the length of dry runs. At 1% chance, going 300 rolls without a hit has about a 5% probability — it happens every 20 cycles. At 460 rolls it's 1%. The strategy has to be sized so that a 500-roll dry run doesn't empty the balance. The Deep 1% Recovery strategy on the DiceStrats marketplace is a tuned version of this family with 280 pre-computed steps.

D'Alembert

Add one unit on a loss, subtract one on a win, at ~49.5%. Far gentler than martingale — the bet grows linearly, not exponentially — so busts are rare, but recovery is slow and a long losing run leaves you betting large for a long time to crawl back. Good for long sessions where you want the chart to look like a saw-tooth rather than a cliff.

Paroli / anti-martingale

Double on a win, reset after 3 wins or on a loss. You're betting the house's money during hot streaks and only ever risking one base bet of your own per cycle. Drawdown is tiny and steady; profit arrives in bursts when a streak lands. It never "busts" in the martingale sense, but it also bleeds the 1% edge on every roll like everything else.

How to choose

1. Decide what you're optimising for. Wager volume with low variance (VIP grinding) → flat or Paroli. A shot at a big session → martingale or recovery, with a hard stop.
2. Size the base bet from the worst streak, not the average. Work out the balance needed to survive a 1-in-10,000 streak at your chance. If your balance can't cover it, the strategy is too aggressive for that balance — drop the base or change families.
3. Set take-profit and stop-loss before the first roll. A strategy without an exit is just a slower way to reach zero. DiceStrats has both as fields on the run screen and they stop the loop the moment they hit.
4. Vault the wins. Anything you move to the Stake vault is out of reach of the next streak. Auto-vaulting on profit is the single most effective "strategy" there is.

Running them without a script

Historically this meant pasting JavaScript into a bot page and hoping it wasn't stealing your token. DiceStrats runs strategies from a browser side panel on your own logged-in Stake tab — your session never leaves the browser, there's nothing to paste, and marketplace strategies come with five risk stages so you pick the base bet relative to your balance instead of guessing. See how to run a dice bot on Stake for the setup.