A friend and I have been discussing an approach, and I'm looking for the most clear and succinct way to refute it. We agreed that I'd put it to the wise folk here, so here goes.
I'm happy to be shot down in flames; that's precisely what I'm after, but ideally expressed in an irrefutable and convincing form that I can pass on.
The essence of it is, to take a roulette analogy; waiting for 10 blacks, then betting on red and increasing the stake on each subsequent black until there's a red.
The theory being that since there have already been 10 blacks... that... and I feel silly saying it... but that the next red is 'due' 10 spins earlier than if we'd not waited for the 10 blacks first.
I'm having trouble expressing why this is no different to not waiting until there are 10 blacks in a row, and just betting in exactly the same way, I know it's probability-istically (!) identical, and so pointless pursuing (and a sure loser ultimately); but I can't convince my friend why.
The actual context is horse racing, and a Martingale staking plan; waiting until X losses in a row, then starting to increase the stake until there's a win - the same as the roulette analogy.
The counter-argument put to me is that since the losses 'started' 10 bets earlier (bets that weren't placed) that the losses from those 10 not-placed bets won't be lost and the win will come sooner - but of course, since the bets weren't placed, then there aren't 10 losses already, we're not at 'bet #11' in a losing streak, we're at bet #1.
Much more to the point, we are always at bet #1 and always will be, regardless of how many wins or losses have happened or whether we bet on them or not.
The whole '10 losses in a row' thing is purely an intellectual/statistical construct that has been superimposed on 10 unconnected events in the past; it is of no use in predicting anything.
Can anyone more eloquent than I come up with a way to make that crystal clear; specifically squashing the counter-argument ?
I've looked at several YouTube videos on the fallacy, but not found one that deals specifically with the 'Wait for X losses first' idea.
The best i can come up with is along the lines of "The horse has no idea how many bets we recently placed on other horses at other tracks around the world - so how can the bets that were or weren't placed have any effect on whether this one wins or not" but that doesn't do the job, nor do paraphrases of any of the above.
Thanks
Gambler's Fallacy again, looking for a succinct explanation of a variant of Martingale
The flaw is that waiting does not change the probability of the next outcome.
After 10 blacks, the chance of red on the next spin is exactly the same as it was before those 10 blacks happened. The wheel has no memory. Those 10 blacks do not make red arrive any sooner, and they do not improve the value of the next bet.
So the proposed method is mathematically identical to saying:
“I will ignore a random stretch of results, then begin a Martingale.”
But ignoring previous results does not create an edge. It just creates the illusion that you are joining the sequence at bet number 11, when in reality your first actual bet is still bet number 1.
That is the key point. Unbet losses are not losses. You do not inherit them. You cannot treat 10 results you merely observed as though they were 10 losing bets in your staking plan.
If the strategy really worked, then simply choosing a starting point after any arbitrary pattern would create profit, and that would mean past random events were predicting future random events. They are not.
The cleanest way to say it is this:
Waiting for 10 losses before starting a Martingale does not avoid the Martingale problem. It is still a Martingale, just delayed.
You still have exactly the same problem as always:
no edge,
same strike rate,
same eventual long losing run,
same exponential stake growth,
same risk of busting the bank.
In horse racing terms, waiting for X losers in a row before increasing stake does not make the next winner more likely. It only makes it feel as though a winner is due. But due is not a mathematical concept, it is just the gambler’s fallacy dressed up as staking logic.
After 10 blacks, the chance of red on the next spin is exactly the same as it was before those 10 blacks happened. The wheel has no memory. Those 10 blacks do not make red arrive any sooner, and they do not improve the value of the next bet.
So the proposed method is mathematically identical to saying:
“I will ignore a random stretch of results, then begin a Martingale.”
But ignoring previous results does not create an edge. It just creates the illusion that you are joining the sequence at bet number 11, when in reality your first actual bet is still bet number 1.
That is the key point. Unbet losses are not losses. You do not inherit them. You cannot treat 10 results you merely observed as though they were 10 losing bets in your staking plan.
If the strategy really worked, then simply choosing a starting point after any arbitrary pattern would create profit, and that would mean past random events were predicting future random events. They are not.
The cleanest way to say it is this:
Waiting for 10 losses before starting a Martingale does not avoid the Martingale problem. It is still a Martingale, just delayed.
You still have exactly the same problem as always:
no edge,
same strike rate,
same eventual long losing run,
same exponential stake growth,
same risk of busting the bank.
In horse racing terms, waiting for X losers in a row before increasing stake does not make the next winner more likely. It only makes it feel as though a winner is due. But due is not a mathematical concept, it is just the gambler’s fallacy dressed up as staking logic.
The roulette wheel has no memory of what has happened before - do horses have any memory of what has happened before?
In roulette, even after 100 blacks in a row or even after 40 blacks and 60 reds, there is still exactly the same probability of the ball landing in either a red or a black pocket, assuming there is no roulette wheel wear bias.
In roulette, even after 100 blacks in a row or even after 40 blacks and 60 reds, there is still exactly the same probability of the ball landing in either a red or a black pocket, assuming there is no roulette wheel wear bias.
-
jamesedwards
- Posts: 5499
- Joined: Wed Nov 21, 2018 6:16 pm
This'll get him...
Ask him, (ignoring the pesky green 0);
> What are the chances of spinning 10 reds in a row?
Answer: 1 in 1,024
> Then, what are the chances of spinning 20 reds in a row?
Answer: 1 in 1,048,576
> Then, ask him what are the chances of spinning 20 reds in a row from spin 11 after 10 straight reds?
Answer: 1,048,576 / 1,024 which equals 1 in 1,024.
Therefore the chances of spinning 10 reds in a row from spin 11 is exactly the same as from spin 1.
Ask him, (ignoring the pesky green 0);
> What are the chances of spinning 10 reds in a row?
Answer: 1 in 1,024
> Then, what are the chances of spinning 20 reds in a row?
Answer: 1 in 1,048,576
> Then, ask him what are the chances of spinning 20 reds in a row from spin 11 after 10 straight reds?
Answer: 1,048,576 / 1,024 which equals 1 in 1,024.
Therefore the chances of spinning 10 reds in a row from spin 11 is exactly the same as from spin 1.
Last edited by jamesedwards on Fri Apr 03, 2026 11:29 am, edited 1 time in total.
This comes up repeatedly and can be difficult for people to get their head around, as you've found.
Perhaps coming at it from a different angle might help: If it were easy to make $$ this way, then there must be many people successfully doing it. Where are they?
Perhaps coming at it from a different angle might help: If it were easy to make $$ this way, then there must be many people successfully doing it. Where are they?
Equally: go ahead and do it, and share the results.silentdiver wrote: ↑Fri Apr 03, 2026 6:39 amA friend and I have been discussing an approach, and I'm looking for the most clear and succinct way to refute it. We agreed that I'd put it to the wise folk here, so here goes.
I'm happy to be shot down in flames; that's precisely what I'm after, but ideally expressed in an irrefutable and convincing form that I can pass on.
The essence of it is, to take a roulette analogy; waiting for 10 blacks, then betting on red and increasing the stake on each subsequent black until there's a red.
The theory being that since there have already been 10 blacks... that... and I feel silly saying it... but that the next red is 'due' 10 spins earlier than if we'd not waited for the 10 blacks first.
I'm having trouble expressing why this is no different to not waiting until there are 10 blacks in a row, and just betting in exactly the same way, I know it's probability-istically (!) identical, and so pointless pursuing (and a sure loser ultimately); but I can't convince my friend why.
The actual context is horse racing, and a Martingale staking plan; waiting until X losses in a row, then starting to increase the stake until there's a win - the same as the roulette analogy.
The counter-argument put to me is that since the losses 'started' 10 bets earlier (bets that weren't placed) that the losses from those 10 not-placed bets won't be lost and the win will come sooner - but of course, since the bets weren't placed, then there aren't 10 losses already, we're not at 'bet #11' in a losing streak, we're at bet #1.
Much more to the point, we are always at bet #1 and always will be, regardless of how many wins or losses have happened or whether we bet on them or not.
The whole '10 losses in a row' thing is purely an intellectual/statistical construct that has been superimposed on 10 unconnected events in the past; it is of no use in predicting anything.
Can anyone more eloquent than I come up with a way to make that crystal clear; specifically squashing the counter-argument ?
I've looked at several YouTube videos on the fallacy, but not found one that deals specifically with the 'Wait for X losses first' idea.
The best i can come up with is along the lines of "The horse has no idea how many bets we recently placed on other horses at other tracks around the world - so how can the bets that were or weren't placed have any effect on whether this one wins or not" but that doesn't do the job, nor do paraphrases of any of the above.
Thanks
The risk/reward profile is all wrong. You risk the world (repeatedly) for the sake of a one unit win. Also, in reality, you'd always come up against house limits or liquidity constraints.
Summary: the idea is perfection; the implementation is impossible (and pretty pointless).
Summary: the idea is perfection; the implementation is impossible (and pretty pointless).
-
wearthefoxhat
- Posts: 3698
- Joined: Sun Feb 18, 2018 9:55 am
What everyone else said, and...
There is no fool-proof roulette system, if there is no wheel or dealer bias.
There have been examples, if the casino are poor on their wheel checking procedures, then a wheel can develop a bias for a particular section of the wheel. The only way of checking this, is with a team, you manually record many 1000's of spins, note the wheel, and analyse the results.
Sometimes new dealers are focussed on running the game, they fall into a pattern when they spin. Wait for them to come back and see if it continues.
In both cases, there is still randomness in play, but over a period of time, there would be an edge.
--
If you treat roulette as fun, then Norman Leigh wrote a book about busting the bank of monte-carlo using a REVERSE-LABOUCHERE (La-Boo-Share) staking system. In essence, you increase your stake from winning and reduce your stake from losing.
It doesn't change the maths built in to roulette, but can be fun. You have as much chance of hitting 10 winning coups @ Evs, as well as 10 losing coups @ Evs. If that occurred, there are maximums at each table. If you reach those, then stop and cash out. (also operate a stop-loss)
---
Horse racing lends itself to a little more nuance. If an Evs chance should be odds on, then you get stuck in. The over round is minimal on Betfair, and so is the 2%. If you hit your win streak, the only maximum is what the other person is willing to risk on Betfair.
---
There is no fool-proof roulette system, if there is no wheel or dealer bias.
There have been examples, if the casino are poor on their wheel checking procedures, then a wheel can develop a bias for a particular section of the wheel. The only way of checking this, is with a team, you manually record many 1000's of spins, note the wheel, and analyse the results.
Sometimes new dealers are focussed on running the game, they fall into a pattern when they spin. Wait for them to come back and see if it continues.
In both cases, there is still randomness in play, but over a period of time, there would be an edge.
--
If you treat roulette as fun, then Norman Leigh wrote a book about busting the bank of monte-carlo using a REVERSE-LABOUCHERE (La-Boo-Share) staking system. In essence, you increase your stake from winning and reduce your stake from losing.
It doesn't change the maths built in to roulette, but can be fun. You have as much chance of hitting 10 winning coups @ Evs, as well as 10 losing coups @ Evs. If that occurred, there are maximums at each table. If you reach those, then stop and cash out. (also operate a stop-loss)
---
Horse racing lends itself to a little more nuance. If an Evs chance should be odds on, then you get stuck in. The over round is minimal on Betfair, and so is the 2%. If you hit your win streak, the only maximum is what the other person is willing to risk on Betfair.
---
-
silentdiver
- Posts: 22
- Joined: Sat Jan 04, 2025 1:53 pm
It might be instructive to do that; the problem is that I'll need to code it and add the functions to the web app that controls it etc. (waste of time #1), then he'll spend time trying out simulating different delays (waste of time #2).I'm trying to avoid both wastes of time, and not add pointless functionality to the software we're developing together.
-
silentdiver
- Posts: 22
- Joined: Sat Jan 04, 2025 1:53 pm
Yep, and the logic is even more compelling for horse racing, when the previous events aren't on the same table in the same room, but for a different horse, probably many miles away, or on the other side of the planet etc.
-
silentdiver
- Posts: 22
- Joined: Sat Jan 04, 2025 1:53 pm
I did try an approach along the lines of "OK, how about, instead of starting by waiting for 10 losses, we just start now by looking back in time for a random set of 10 potential bets that we might have placed but would have lost, call those 'loss 1-10' and start betting immediately, it's just the same as waiting for them".
It is tricky, my friend is a consummate cricket better in terms of live human betting in-play; and has a flair for, and a proven record at that. Trading bets on the ups and downs of a single individual match is quite different to the more mathematical, automated, hands-off approach of betting on the final outcome of a series of events like horse races.