Auto Roulette Strategy: Same Edge on Every Bet, So Play the Clock (2026 Data)

Auto Roulette has no dealer, no multipliers and no bonus round, which leaves a strategy question most roulette pages skip: if every bet returns 97.30%, what is there to decide? The answer is variance, coverage and pace. This page ranks the bets by how they lose rather than whether they lose, works the Martingale sums against streaks we have actually recorded, and shows why an automated wheel dealing about 70 spins an hour makes time the main cost. Our figures come from the 275,852 spins logged on the live tracker since December 2025; the rules are covered in the Auto Roulette guide.

Is there a best bet on Auto Roulette?

No. Every bet on the layout returns 97.30% and loses 2.70% of stake in the long run, because each one is priced against 36 numbers while the wheel has 37. The zero takes the same 1/37 bite whether you back a single number, a corner or black.

What the bets differ in is how often they pay and how large the swings are. The table lists each bet with its hit chance, payout and the standard deviation of the result per unit staked.

Bet	Numbers	Chance per spin	Payout	Return	Swing per unit (std dev)
Straight-up	1	2.70%	35:1	97.30%	5.84
Split	2	5.41%	17:1	97.30%	4.08
Street	3	8.11%	11:1	97.30%	3.29
Corner	4	10.81%	8:1	97.30%	2.80
Line	6	16.22%	5:1	97.30%	2.21
Dozen / column	12	32.43%	2:1	97.30%	1.40
Red/black, odd/even, low/high	18	48.65%	1:1	97.30%	1.00

A straight-up bettor experiences swings almost six times wider than an even-money bettor for the same expected loss. That is the entire "best bet" debate in one number: the return column is flat, the swing column is not.

Our data backs the return column. Across the last 90 days (103,221 spins) red landed 48.63%, black 48.64%, the first dozen 32.49%, the second 32.29%, the third 32.49%, and the single most frequent number (10) hit 2.78% against a theoretical 2.70%. Nothing is out of line. The RTP page tracks these shares against theory over time.

How do coverage plans change the picture?

Coverage raises how often something pays and does nothing to the return. Whether you spread one chip or nine, the expected loss is 2.70% of the total on the table.

The clearest demonstration is the "cover almost everything" plan. Put one unit straight-up on 35 numbers. You win 94.59% of spins, each win returning 36 units for 35 staked, and the other 5.41% of spins lose all 35. Expected result per spin: (35/37 × 1) minus (2/37 × 35) = minus 0.946 units, exactly 2.70% of 35.

Combining outside bets works the same way. The table gives the combined hit rate for the most common plans.

Plan	Numbers covered	Chips (equal units)	Hit rate	Net on a hit	Net on a miss
Two dozens	24	2	64.86%	+1	-2
Tiers + Orphelins	20	11	54.05%	varies by sector	-11
Voisins + Tiers	29	15	78.38%	varies by sector	-15
Voisins + Tiers + Orphelins	36	20	97.30%	varies by sector	-20 (zero only)
35 straight-ups	35	35	94.59%	+1	-35

Cover 36 numbers with all three sectors and the only losing spin is zero, which arrived 2.73% of the time in our 90-day window. Our longest recorded zero drought was 396 spins, ending 2026-08-23, so a full-sector player could go five hours without a losing spin and still be down for the session, because most "wins" return less than the 20 chips laid.

Two dozens is the compact version: 24 numbers, 64.86% hit rate, win one unit or lose two. Over 100 spins at one unit per dozen you expect about 65 wins worth 65 and 35 losses worth 70, a deficit of 5.4 units, which is 2.70% of the 200 staked.

What do the racetrack sector bets actually do?

The racetrack bets are pre-built coverage of physical sections of the wheel, and they return the same 97.30% as everything else. What they add is a mix of bet sizes inside one click, so the payout depends on which pocket inside the sector the ball finds.

Evolution's client offers the four French call bets plus neighbour bets:

• Voisins du Zero: 17 numbers either side of zero, 9 chips (a two-chip trio on 0/2/3, a two-chip corner on 25/26/28/29, and five splits). Hit chance 45.95%.
• Tiers du Cylindre: 12 numbers opposite the zero, 6 chips on six splits. Hit chance 32.43%.
• Orphelins: the 8 numbers left over, 5 chips (one straight-up on 1 and four splits). Hit chance 21.62%.
• Jeu Zero: 7 numbers hugging the zero, 4 chips. Hit chance 18.92%.

Each component pays its normal odds. A Tiers hit returns 18 units for 6 staked; a Voisins hit returns 18 to 27 against 9; Orphelins pays 18 on a split or 36 on the straight-up, against 5.

Our sector shares over 90 days: Voisins 45.62%, Tiers 32.61%, Orphelins 21.77%, against theory of 45.95%, 32.43% and 21.62%. The wheel is not favouring any section of itself; the streaks and sectors page shows how quickly "the ball keeps landing near zero" impressions dissolve.

Does Martingale beat Auto Roulette?

No. Martingale on an even-money bet doubles after each loss so that the first win recovers everything plus one unit. It needs two things the game refuses to supply: short losing runs and an unlimited table.

The losing probability on red is 19/37 per spin (18 blacks plus the zero), which is 51.35%. Runs of losses then multiply:

Straight losses on red	Probability from any starting spin	Next bet from a 0.10 start	Total already lost
6	1.83%	6.40	6.30
8	0.48%	25.60	25.50
10	0.13%	102.40	102.30
13	0.017%	819.20	819.10
16	0.0023%	6,553.60	6,553.50

The two bottom rows are not hypothetical. Inside one 90-day window our tracker recorded a 13-spin black run and a 16-spin red run, and those figures count only same-colour runs; a zero landing in the middle extends a losing streak for an even-money bettor without breaking the colour sequence. A player who started at 0.10 against that 16-red run needed 6,553.60 on the seventeenth spin to win back 6,553.50 plus 0.10 profit. Every Auto Roulette table has a per-spot maximum well below the point where the doubling keeps going, so the progression ends with the loss locked in, not recovered.

At roughly 1,690 spins a day, a 0.0023% event is expected about twice per 100,000 spins, roughly every two months on this one wheel.

The deeper flaw is arithmetic. Doubling means the largest bets are made at the end of the worst runs, so the 2.70% edge is applied to the biggest stakes. Martingale does not fight the edge; it feeds it.

What about Fibonacci, D'Alembert and other progressions?

They move the same loss around without changing its size. Fibonacci raises stakes more slowly than Martingale (1, 1, 2, 3, 5, 8, 13, 21, 34, 55, 89, 144), which means a 12-loss run costs 376 units with 233 needed next, and a single win no longer recovers everything. D'Alembert adds one unit after a loss and removes one after a win, so the bust is slower still, but a run of 16 losses still leaves you betting 17 units to win back a 136-unit hole one unit at a time.

Reverse progressions, Labouchere, Oscar's Grind: every one of them is a rule about stake size, and stake size is the only thing the house edge multiplies. Sum any sequence of bets, take 2.70% of it, and that is the expected loss whatever order you placed them in.

Flat betting is the one plan whose worst case you control: bet the same unit every spin and the most you can lose in an hour is about 70 units, a bound set by you and the clock rather than the table maximum.

Should you chase a cold number or a zero drought?

No, and Auto Roulette gives unusually clean evidence why, because the tracker records every spin with no host and no session breaks.

The chance a specific number misses N spins in a row is (36/37) to the power N. That gives 36.3% for 37 spins, 6.5% for 100 spins and 0.42% for 200 spins. In our last 90 days number 29 went 469 spins without landing, a 0.0003% event for one chosen number. It looks like a sign of something. It is not: with 37 numbers and 103,221 spins, a gap of 400-plus is expected to show up somewhere in the window, and the numbers 21 and 6 that hit least often (2.62% and 2.63%) still sat within a whisker of 2.70%.

Zero behaves the same. The average gap between zeros was 36.6 spins (theory 37), the median 24, the longest 396 spins, ending 2026-08-23; two other droughts of 277 and 270 spins both ended on 2026-07-25. A "zero is due" bettor placing one unit on 0 through the 396-spin gap lost 396 units waiting for a 35-unit payout.

Hot numbers are the mirror image. Number 10 topped our 90-day list with 2,870 hits and number 21 sat at the bottom with 2,703, a gap of 167 hits in 103,221 spins. That is the range random variation produces on a fair wheel, and it will have shuffled by the time you read this. The live stats page shows the current hot and cold list with that caveat printed next to it.

How should you size an Auto Roulette bankroll?

Size it by the hour, because the wheel sets the pace. Our feed shows a spin roughly every 51 seconds, about 70 an hour and 1,690 a day, with no break between rounds.

Multiply the stake per spin by 70 and by 0.027 and you have the expected cost of an hour. One unit a spin on red: 70 units exposed, 1.89 units expected loss, typical swing about 8.4 units either way (the standard deviation of 70 even-money bets). Ten units a spin: 700 exposed, 18.9 expected loss, swing about 84.

Straight-ups need a different scale. One unit a spin on a single number loses the same 1.89 units an hour in expectation, but the swing is about 49 units over 70 spins. The average wait for a specific number in our data is 37.0 spins, so a one-number player should expect about two hits an hour and should not be surprised by none (roughly 15% chance across 70 spins).

From that, a workable sizing rule: hold at least 100 units per unit staked for outside bets, and 300 to 500 units per unit staked for straight-ups and small sectors, where a unit is your flat bet. Set a stop-loss in units and a session limit in minutes before the first spin, and treat reaching either as the end. Keep the stake flat. If you want more excitement, move from red to a dozen or a corner, which changes the swing column in the first table; do not change the amount.

Why is speed the real cost on an automated wheel?

Because the house edge is a rate charged per spin, and Auto Roulette delivers spins faster than a dealer table with nothing to interrupt them. A player who stakes 2 units a spin on Auto Roulette for an hour puts 140 units through the edge. The same player at a hosted table running 35 to 40 spins an hour puts through 70 to 80. Identical bet, identical edge, roughly double the expected loss.

The wheel does not care who is watching. Player traffic peaks at 15:00 UTC in our 30-day data, yet the spin count per hour barely moves, because no dealer pauses for a busy table. There is only the number of spins you choose to sit through.

The practical advice is to skip spins. Nothing forces a bet every round, and betting one spin in three at the same stake cuts the hourly cost by two thirds while the session lasts just as long. It is the only "strategy" on this page that lowers the house's take, and it does so by the least glamorous method possible: fewer bets.

Auto Roulette strategy cheat sheet

Every row returns 97.30%. The columns that differ are hit rate, swing, and what the plan does to the pace of loss.

Approach	Hit rate	Swing per unit	Hourly cost at 1 unit/spin	Verdict
Flat on red/black or other even-money	48.65%	1.00	about 1.9 units	Longest bankroll life; gentle swings
Flat on a dozen or column	32.43%	1.40	about 1.9 units	Slightly livelier, same cost
Two dozens	64.86%	about 1.4 per 2 units	about 3.8 units	Frequent one-unit wins, two-unit losses
Flat straight-up	2.70%	5.84	about 1.9 units	Same cost, six times the swing; expect droughts
Voisins / Tiers / Orphelins	45.95% / 32.43% / 21.62%	varies inside sector	9, 6 or 5 chips per spin	Coverage with mixed payouts, no edge change
Cover 35 or 36 numbers	94.59% / 97.30%	low, with rare deep holes	35 or 20 chips per spin	Many tiny wins, occasional big loss, same 2.70%
Martingale on colour	48.65% per spin	unbounded until the cap	grows with each loss	Bust is scheduled by streaks we have already logged
Chasing cold numbers or zero	2.70% per spin	5.84	about 1.9 units	The wheel has no memory; 469-spin droughts happen
Bet every third spin	as chosen	as chosen	one third of the above	Only line that reduces the house's take

Choose a row for the variance you want, set the unit from the bankroll rule, and let the clock decide when to leave.

What does a realistic Auto Roulette session look like?

A realistic session is a slow drain with swings that dwarf the drain in the short run. Two hours of flat 1-unit bets on red is about 140 spins, an expected loss of 3.8 units and a standard deviation of roughly 11.8. Roughly one session in three ends ahead; roughly one in three ends down by more than 10. The average sits just below zero and does not move.

Straight-up sessions look nothing like that and cost exactly the same. Two hours on one number means about 140 spins and roughly four hits, each paying 35, so the bankroll falls by 30 to 40 units, jumps by 35, falls again. About 2% of such sessions see no hit at all. Our 469-spin gap on number 29 was over six and a half hours of continuous wheel time.

If you want to check these figures, the live tracker refreshes from the same feed and shows the current streaks, droughts and sector shares over 30 and 90 days. Evolution's rules and the 97.30% return figure are in the game client's help panel and on evolution.com. For roulette variants that carry multipliers, see the Lightning Roulette guide.

Auto Roulette runs 24 hours a day and will spin whether or not you are betting. Decide how long you will play and how much you will risk before you open the table, and close it when either limit arrives. Support for gambling problems is listed on our observatory page.

Frequently asked questions

What is the best bet on Auto Roulette?

There is no bet with a better return. Evolution rates every wager on Auto Roulette at 97.30%, from a single number to red or black. The bets differ only in variance: even-money bets win 48.65% of spins and swing gently, a straight-up wins 2.70% of spins and swings hard. Pick by how long you want a bankroll to last, not by any imagined edge.

Does Martingale work on Auto Roulette?

No. Doubling after each loss on red needs losing runs to stay short, but a run of 13 losses costs 819 units from a 0.10 start and our tracker has logged both a 13-spin black run and a 16-spin red run in one 90-day window. Every spin still pays 97.30% on average, so stacking losing bets only stacks the loss.

Are the racetrack bets on Auto Roulette worth playing?

They are worth playing only as a coverage choice. Voisins du Zero puts 9 chips on 17 numbers and hits 45.95% of spins, Tiers puts 6 chips on 12 numbers for 32.43%, Orphelins 5 chips on 8 numbers for 21.62%. Each returns the standard 97.30%. Our 90-day sector shares (45.62%, 32.61%, 21.77%) match those figures closely.

Should I bet on a number that has not hit for a long time?

Not as a strategy. The wheel has no memory, so a number that has missed 200 spins still has a 1-in-37 chance on the next one. Our tracker saw number 29 go 469 spins without landing in the last 90 days. Someone betting one unit on it every spin because it was due would have lost 469 units before the first payout of 35.

How much bankroll do I need for Auto Roulette?

Budget by time. The wheel spins about 70 times an hour, so one hour of flat 1-unit bets on red exposes 70 units and loses about 1.9 units in expectation, with typical swings of 8 to 10 units either way. A 100-unit session bankroll at 1 unit a spin is comfortable for two or three hours on outside bets; straight-ups need three to five times as much per unit staked.