Fix my betting system.

December 16, 2004 11:23 AM

What's the flaw in this horse racing betting system? [MI]

1. Get some reliable tips, like 25%-30% accuracy over the last 2 years. You might want to compare tipsters for each race.

2. Initialise your "bling" score, indicating the number of currency units you want to make per bet, e.g. 10.

3. Never bet odds-on (i.e. where the first digit of the odds fraction is smaller than the second).

4. Divide bling by the odds fraction. Round up to something, e.g. half your initial bling. Stake this amount.

a) If you win, begin the cycle again at stage 2.

b) If you lose, add together your current bling, your initial bling, and your stake. This becomes your new bling. Bet on the next race from stage 4.

5. If you get a long losing streak and are getting scared by the bet amounts, start the process again and come back to the current cycle when you have some more money.

I don't know what I'm talking about, of course, but I can't see the risk with this system, as long as you have enough money to buffer with so you can get over a streak of losing tips. However, the amount is unlikely to be very much, since you are extremely unlikely to get more than 10 losers in a row, say.

Of course I don't believe in get-rich-quick schemes, so where's the flaw in the system I'm missing here?

posted by Dreamghost to sports, hobbies, & recreation (41 comments total)

Not quite sure I'm getting this, but isn't this the same as going to the roulette table and just betting black, doubling up each time you lose, because with the first win you'll come out ahead? If so, I think the flaws in both strategies are that a) losing streaks can extend well past your ability to pony up the geometrically-increasing amount of dough and b) the house has a slight edge.

(disclosure: I know nothing about gambling.)
posted by stupidsexyFlanders at 11:28 AM on December 16, 2004

I also know nothing, but I doubt you can compare this to roulette, since this involves using tipsters.
posted by Pretty_Generic at 11:32 AM on December 16, 2004

Sounds like a variant of the Martingale system to me. The problem, as noted, is losing streaks. Just because it's unlikely that you'll lose x number of times in a row doesn't mean it won't happen. And when it does, you'll be broke.
posted by uncleozzy at 11:38 AM on December 16, 2004

Of course there's a risk of that happening, but if you have enough money to "buffer" with to make the risk of going broke small enough, this seems like a very sensible method of betting. Am I wrong?
posted by Pretty_Generic at 11:41 AM on December 16, 2004

And in the casino of course, the odds are always geared to benefit the house. When you're betting using tipsters, that isn't necessarily the case, is it?
posted by Pretty_Generic at 11:43 AM on December 16, 2004

Is there a house limit? That could break you absolutely in the case of a losing streak, even if you're not "getting scared by the bet amounts".
posted by mr_roboto at 11:44 AM on December 16, 2004

I would guess the chance of getting from a $5 bet to whatever the limit for online horse racing bets is (surely thousands and thousands), before the tipster provides a winner, is negligible.
posted by Pretty_Generic at 11:49 AM on December 16, 2004

as long as you have enough money to buffer with

Works fine, as long as you have an infinite amount of money to buffer with. (But if you already have an infinite amount of money, why waste your time gambling?) If you work out the math, it turns out that any finite amount of money, no matter how large, is not a large enough buffer.

since you are extremely unlikely to get more than 10 losers in a row, say.

The chance of getting 10 losers in a row may be small, but it's balanced by the fact that on the rare occasion that you do get 10 losers in a row, you've lost a lot of money. It's like making a bet that you have a 99% chance of winning--but when you win, you only win $1, and when you lose, you lose $1000. (And if you have an even larger buffer, then it would maybe take 15 losers in a row unstead of 10 to break you, but after the even more unlikely event of 15 losers in a row, you've lost even more money.)
posted by DevilsAdvocate at 11:50 AM on December 16, 2004

DevilsAdvocate, are you absolutely sure a high rate of winning couldn't make the risk of that happening a worthwhile one? Surely it all depends on the accuracy of tips.
posted by Pretty_Generic at 11:56 AM on December 16, 2004

I mean, if I had 99% accurate tips, it would obviously be worth doing. If so, what about 30% accurate?
posted by Pretty_Generic at 11:58 AM on December 16, 2004

I would guess the chance of getting from a $5 bet to whatever the limit for online horse racing bets is (surely thousands and thousands), before the tipster provides a winner, is negligible.

I don't know; using a simple Martingale system, a $5 bet will inflate to over $1000 in less than ten losses; you get to over $15,000 in less than 15 losses. Are you sure that sites that allow bets as small as $5 have limits much larger than this? I don't know about online horse racing, but I do know that casinos tables with $5 minimums typically have $500 limits.

It's not like the casinos haven't thought of this.
posted by mr_roboto at 11:58 AM on December 16, 2004

Plus, you have to be sure that the tipsters you're using are actually using inside information from the crooked race staff - you'd better be doing this in England, if at all. You also must factor in the cost of valuable tipster services. Anything less than a few hundred dollars a week is worthless. And don't take their word for their track record. Track them yourself for about a year (enough to cover all the major races and festivals) before you commit to one, or even better, a basket (although far more expensive).

I reiterate, you are talking about inside info right? Not just "talented analysis"? And everyone else is right about this being the Martingale system, except that a good tipster is like having a biggish edge over the house, as opposed to the tiny percentage you'd have in say, counting cards.
posted by loquax at 12:00 PM on December 16, 2004

And for reference, with 30% accurate tips, you have about a 3% chance of losing ten times in a row.

If you're looking at hitting the house limit after only 5 losses, there's a 17% chance of losing all of your money with 30% accurate tips. Of course, you'd be losing less money than in the first case.
posted by mr_roboto at 12:05 PM on December 16, 2004

Using this system with an initial bling of 5, a $5 bet will inflate to only $345 after 10 2/1 losses. A less pessimistic average odds figure of 4/1 means it will inflate to only $55.

I think.
posted by Pretty_Generic at 12:06 PM on December 16, 2004

Anything less than a few hundred dollars a week is worthless.

The Sun's free tips in Britain have had 29% accuracy over 2 years.
posted by Pretty_Generic at 12:07 PM on December 16, 2004

And for reference, with 30% accurate tips, you have about a 3% chance of losing ten times in a row.

Um? What sum are you using for that? Isn't the chance 0.7^2^2^2... ten times? Which would be 4e-80?

I'm not feeling particularly bright right now, so excuse me for basic maths errors (if any)
posted by Pretty_Generic at 12:10 PM on December 16, 2004

OH GOD THAT WAS STUPID. Ignore me.
posted by Pretty_Generic at 12:14 PM on December 16, 2004

it's not about % accuracy, it's about whether you win more than you spend, so you have to factor in the odds. if you were 50% accurate on average even odds then you've won nothing.

in the end, it comes down to: is your info better than the bookies, plus whatever edge they give themselves though the odds.

more generally, if you're asking questions like "oooh this sounds reasonable" without being able to calculate whether it's reasonable or not, you don't know enough to take the chance.
posted by andrew cooke at 12:15 PM on December 16, 2004

This system not only ensures that you won't lose too much, it ensures that you won't win too much. If you want to do it as a hobby, you look safe. But it will take a discouraging amount of time to do much better than even. Including some awful runs of bad luck that will keep you up nights and some pitiful wins that won't seem very rewarding after you've put months into your system.
posted by Mayor Curley at 12:17 PM on December 16, 2004

(0.7^2^2^2^2^2^2^2^2^2^2) != (0.7^10)
(0.7^10) == 0.0282475249 ~~ 3%
posted by sonofsamiam at 12:17 PM on December 16, 2004

This system not only ensures that you won't lose too much, it ensures that you won't win too much.

Every time you win a bet, it means you will have won X initial bling amount (e.g. $10) for EVERY bet you have placed, win or lose. That seems like enough to me.
posted by Pretty_Generic at 12:26 PM on December 16, 2004

Using this system with an initial bling of 5, a $5 bet will inflate to only $345 after 10 2/1 losses. A less pessimistic average odds figure of 4/1 means it will inflate to only $55.

Yeah; stupid of me. I forgot that we're not dealing with even odds. Horseracing. andrew cooke is right: it's all about the odds.
posted by mr_roboto at 12:26 PM on December 16, 2004

OK, we've both made stupid mistakes, we're even.
posted by Pretty_Generic at 12:29 PM on December 16, 2004

The tips are just a red herring here. If you have good tips, you'll win money. If you don't have good tips, you'll lose money. Martingale makes no difference.

The system described would give you a lot of really tiny wins and an occasional devastating loss, but in the end, you'd wind up with the same winnings or losses as if you'd just bet a fixed amount on each race.

On preview, I disagree strongly with Mayor Curley. The math is easy to check. Try it with a coin toss, and you'll see that in the improbable but all-too-common case of a long losing streak, you'll take a big hit.

Another thing that's at work here is the gambler's fallacy, which is the notion that losing/winnings streaks are more likely to end than not. It doesn't work that way. From a gambler's perspective, horse races are independent events. The fact that you lost nine in a row has no bearing on the outcome of the next race.

Again, using a coin toss as an example, we know that there's a 50/50 chance that a single toss will land heads side up. Suppose we've tossed it nine times now and it's been heads every time. What are the chances that the tenth toss will be heads? 50/50.
posted by chrchr at 12:31 PM on December 16, 2004

It won't work because parimutual betting is an efficient market. Everyone has access to the same tipsters so the market adjusts the odds to account for that information.

You probably could slowly make money by simply betting on favourites if the track and the government weren't skimming their share off the top, but they are.
posted by timeistight at 12:34 PM on December 16, 2004

The fact that you lost nine in a row has no bearing on the outcome of the next race.

But by making your bets dependant on the outcome of previous events, you're no longer betting on each individual race: you're betting against the streak. And long streaks are relatively improbable. The longer the streak, the lower the probability.
posted by mr_roboto at 12:37 PM on December 16, 2004

chrchr, what if you limited your maximum bet to a certain percentage of your assets, and went back to make those big bets when you had eventually built up more money from smaller ones? Then the devastating losses wouldn't be so devastating any more.
posted by Pretty_Generic at 12:38 PM on December 16, 2004

timeistight: betting in Britain is tax free. A £1 bet on all favorites in 2002 would have made a profit (although this is unusual).
posted by Pretty_Generic at 12:39 PM on December 16, 2004

I know the odds are not exactly the same as flipping a coin is based upon a 50 - 50 chance, but I once flipped a coin heads ten times in a row. The odds of that are less than 1 in a thousand. If I can flip ten heads, you can loose ten bets.
posted by pwb503 at 12:43 PM on December 16, 2004

mr_roboto, you're not betting against the streak, though. You're only betting on the next toss. You could, for instance, bet against somebody flipping ten heads in a row, but, what this system does is different. It says that now that we've seen nine heads in a row, we're going to bet BIG that the tenth will be tails. As far as this bet is concerned, the nine previous tosses have no bearing. Why would you bet more on this particular 50/50 chance than on the previous nine 50/50 chances?

pretty_generic, the people who take this stuff seriously size their bets as a percentage of their assets factored against their calculated likelihood of winning. I.e., if you calculate a 100% chance of winning, you'll bet the farm. If you think you have a 3% edge, you'll be far more modest.

But again, the likelihood of winning has everything to do with the tips/handicapping skill/whatever, and absolutely nothing to do with whether or not you won the last race.
They're independent trials. Think about it. The horses and jockeys have no idea that you've lost ten bets in a row. What magical force impels them to break your losing streak?
posted by chrchr at 12:53 PM on December 16, 2004

...and went back to make those big bets when you had eventually built up more money from smaller ones?

Is this even possible, though? The most you can profit from a successful series of bets (one in which you don't hit the house limit) is your initial stake. Let's take the case of $5, again, assuming a $1000 limit, and even odds. When you hit the limit and lose, you lose $640 + $320 + $160 + $80 + $40 + $20 + $10 + $5 = $1,255. With even chances, you encounter a loss of this magnitude about 4 times out of every 1000, for a total of $5,020. In those same 1000 trials, you win $5 996 times, for a total of $4980. You wind up a net loser.

Of course, with uneven odds and tips, the horseracing example is more complicated. This is just the fundamental flaw in Martingale systems.

Why would you bet more on this particular 50/50 chance than on the previous nine 50/50 chances?

Because if you lose on this trial, you'll double your bet on the next trial, covering all the money that you've lost so far. Odds are that eventually the streak will break, and the bet that you make on the trial at which it breaks will be sufficient to cover all previous losses (since you've been doubling at each trial) and provide you winnings equal to your initial stake. Read uncleozzy's link about the Martingale system above.
posted by mr_roboto at 1:02 PM on December 16, 2004

Because if you lose on this trial, you'll double your bet on the next trial, covering all the money that you've lost so far.

I read the link, and I've read plenty about the Martingale system. This is a classic example in probability.

If your assets are infinite, you're right. The Martingale system "works". In the real world, as the link points out, there are table limits, and you have a finite amount of money, and you'll be unable to cover your previous losses.

My point is that the Martingale system doesn't actually change your odds of winning. In the coin toss example, that's 50/50, and always 50/50. In the horse racing example, it depends on the accuracy of the tips.

If you believe you have an advantage in the game, you should size your bets according to the probability of winning and the amount of money you have to play with in order to minimize risk. Leave Martingale out of it.

And anyway, the idea that you should size your bets based on the amount you've lost is probably not a good one to begin with. The money you've lost is gone. It's a sunk cost. When you place the big bets in a Martingale progression, you're not winning your money back. You're only risking more of your money.
posted by chrchr at 1:19 PM on December 16, 2004

I don't think anyone doubts that this system would be a long-term flop in a casino. But with fairly good tips and horse-racing odds, it seems like a fantastic method of betting to me.
posted by Pretty_Generic at 1:45 PM on December 16, 2004

In the real world, as the link points out, there are table limits, and you have a finite amount of money, and you'll be unable to cover your previous losses.

And if you had been reading the thread, you would realize that this was my initial point.

My point is that the Martingale system doesn't actually change your odds of winning.

Of course it doesn't change the odds of each individual trial. No one here (apart from yourself) suggested such an outrageous thing. It's a risk management system; it fails because of house limits.

You said that "another thing that's at work here is the gambler's fallacy". The Martingale system is not a product of this fallacy; it fully recognizes that each trial is an independent event. Neither is Dreamghost's system a product of the gambler's fallacy. He increases his "bling" to minimize risk, not out of the misapprehension that the odds change because of his losing streak. What he's missing is that the small losses covered by progressively increasing bets will come back in the form of the occassional massive loss. Maybe. It's still not clear to me whether these losses will actually overcome the winnings.

If you believe you have an advantage in the game, you should size your bets according to the probability of winning and the amount of money you have to play with in order to minimize risk.

It's true that in casinos, these systems turn out to be equivalent to flat bets.
posted by mr_roboto at 2:00 PM on December 16, 2004

Inside of a casino, or at the track, or on a plane, or in a train, a betting method makes exactly zero difference in the amount of money you have in your pocket when the game is over.

There's nothing about the odds and the tips that make Martingale more applicable to horse racing than to a given casino game. It's the same deal. Progression betting doesn't change your odds of winning; not for a single trial, and not for a series of trials.

I do not say that Dreamghost can't make money at this. If the tips are good, he'll win. But he'll win just as much betting the same amount every race as he would if he bet a progression, and a long losing streak won't hurt nearly as much.

On preview, mr_roboto, I'm sorry if I misunderstood your post.

It's still not clear to me whether these losses will actually overcome the winnings.

If he has an advantage from the tips, we can expect that the winnings will exceed the losses after many, many trials. The same is true of flat betting.

I say that the Martingale is a product of the gambler's fallacy because it rates the value of an opportunity based on the outcome of previous opportunities.

For example, if we determine that a bet on a particular horse is worth $6 at 10:1 odds, it's always worth $6. The fact that we lost our last three bets doesn't suddenly make it worth $20. The value of the opportunity is the same.

There's no risk mitigation in Martingale either. It just shifts the risk around. The same amount of money is at risk, assuming the average Martingale bet is the same as the flat bet.
posted by chrchr at 2:38 PM on December 16, 2004

The problem is that the system tries to merge two mutually exclusive sets of information--tips on race outcomes, and game theory.

If you had good tips on race outcomes, you could use any betting system and win money.

Adding in game theory (which appears to be poorly understood in this case, anyway) will get you no further ahead. My guess is that this "system" is superstitious pseudo-game theory rather than any actual extrapolation of statistical outcomes (I can't be bothered to do the math to check, but it looks sketchy to me).
posted by Sidhedevil at 3:05 PM on December 16, 2004

If you want to read up on the math, the Wikipedia articles on Martingale and the
Gambler's Fallacy are good starting points.
posted by chrchr at 3:21 PM on December 16, 2004

Inside of a casino, or at the track, or on a plane, or in a train, a betting method makes exactly zero difference in the amount of money you have in your pocket when the game is over.

This statement is clearly false. To use an extreme example, if a person started with £100 at the beginning of the 2002 season, and put it all on the favorite in the first race, they would probably be broke immediately. If they put £1 on every favorite in the season, they would end up with a profit.
posted by Pretty_Generic at 4:50 PM on December 16, 2004

probably?
posted by jacobsee at 5:37 PM on December 16, 2004

If you put 1 currency unit on every favorite in the season, you could very well end up without a profit. You don't even need to do the math on this one--the favorite doesn't always win, and the favorite is rarely at very advantageous odds.
posted by Sidhedevil at 5:46 PM on December 16, 2004

I think upthread someone mentioned that the 2002 season was special in that betting the favourites would have actually made money. But what's the point? Are you saying you have a time machine and can go back to 2002? If that's the case, why don't you just look up who actually did win the race and just bet on those horses?

(I know that's not what you're really saying, but using the benefit of hindsight in your argument doesn't make it very strong)

I read a story once about a kid going up against the greatest poker player of all time. He bet it all on the first hand because he figured that the more games he played, but more likely the great player's skill would overcome his luck. Makes sense to me.
posted by jacobsee at 2:18 PM on December 17, 2004