What are the odds that I have a working system vs I've just been lucky?

Any opinions (preferably backed up with some mathematical evidence) gratefully received.

Quote:MrPogle2I am betting on horse racing. The house edge is about 10%. Applying my "system" I am 3.8% in profit over 1720 level stakes bets.

What are the odds that I have a working system vs I've just been lucky?

Any opinions (preferably backed up with some mathematical evidence) gratefully received.

link to original post

So if you’ve been betting $1 you are up $65 after 1720 bets?

It depends on the odds on the bets you’ve made. If you are only taking favorites with low odds then it is more likely you have a good system

If most of your bets are extreme longshots then it is more likely that variance has been on your side.

Are these all ‘win’ bets, some place or show bets, exactas, perfectas, etc?

Yes. 1720 x $1 bets has resulted in a profit of $65.

As to the odds. The odds vary a lot but I would guess they average about 4/1. The bets are win-only except where a very long shot is picked (25/1 or longer) when the bet is made each-way. Each-way is generally 20% of the win odds paid on the first 3 places.

if you do have a methodology to beat racing in the long run that would be extremely impressive

there are 𝙫𝙚𝙧𝙮 𝙫𝙚𝙧𝙮 𝙛𝙚𝙬 who can legitimately claim that

largely because of the gigantic takeout -

.

I was hoping that someone may be able to calculate the likelihood of my results being just chance. (e.g. tossing heads 3 times in a row doesn't arouse suspicion but doing it 35 times in a row is more likely to be bent than lucky.)

The other problem I have is that my 1720 bets represents nearly a year of betting because only a few horses qualify each day. So the data gathering is very slow with no prospect of modelling or simulation. I guess I'll just having to keep going and see if I'm more than 100 up in 12 months. :)

Quote:MrPogle2Thanks for the reply

Yes. 1720 x $1 bets has resulted in a profit of $65.

As to the odds. The odds vary a lot but I would guess they average about 4/1. The bets are win-only except where a very long shot is picked (25/1 or longer) when the bet is made each-way. Each-way is generally 20% of the win odds paid on the first 3 places.

link to original post

I’d say it is most likely your system will beat random guy throwing darts at the program. I don’t think your results are good enough to be sure your system is enough to beat the high vig.

But if I was you, I’d start betting real money on it and see what happens.

I’m not good enough at the math to figure out how many standard deviations from the norm your results are. There are a dozen or so members here who are. Hopefully they can chime in.

Quote:MrPogle2Applying my "system" I am 3.8% in profit over 1720 level stakes bets.

your idea of betting just on stakes races is a good one

those horses are more consistent and more predictable

if you analyzed your results of just high level stakes - Grades 1, 2, and 3 -

you might find your profitability is greater

I've done a similar thing - I'm not going to claim I beat racing - but I definitely did much better on high level stakes races

.

Quote:lilredroosterQuote:MrPogle2Applying my "system" I am 3.8% in profit over 1720 level stakes bets.

your idea of betting just on stakes races is a good one

Apologies, I didn't make myself very clear. By "level stakes" I meant that I was betting 1 unit each time and not varying the size of the bet.

I agree with you, though, that better class racing tends to go more to form.

I'm not good enough at the maths either, but you should study up on kelly criteria. That will balance your risk of ruin with optimising profitability IF you have an advantage.Quote:SOOPOO

But if I was you, I’d start betting real money on it and see what happens.

I’m not good enough at the math to figure out how many standard deviations from the norm your results are. There are a dozen or so members here who are. Hopefully they can chime in.

link to original post

Empirically you've estimated your advantage at x=65/1720 = 3.7%. We'll work with that.

Now, set aside a bankroll of whatever you are prepared to lose. let's give that a value of B. Whatever your bankroll becomes, re-evaluate B for the next wager.

All as I understand it... And I'm no expert on kelly, but this is a simplification.

Now if you were making even money bets according to Kelly, you would wager x=3.7% of your remaining bankroll each time. If you won or lost, your next wager would be 10% of your newly increased or decreased bankroll. Thus if your system is successful, your bets would increase steadily and vice versa.

But if you are wagering on horses that are not paying even money, you multiply your wager by the probability of winning. That means that any win at any odds will increase your bankroll by the same amount.

So if wagering on a horse with odds that give it 1/4 chance of winning, your wager on that horse becomes 3.7% x1/4 x B

Maybe a better expert will fix up my simplification.

Stake management is something I need to investigate once I am convinced I actually have an edge. Obviously, no staking plan can overcome a house edge so I'm trying to establish one thing at a time but having looked at Kelly I'm not sure it will be possible to apply it.

Kelly seems to rely on knowing both offered odds and estimated probability of winning and the only estimate that I have is 103.7%. Does it make sense to apply this blanket value to every bet? Kelly seems to suggest that you need to evaluate this on a bet by bet basis.

There is also something else that bothers me a bit about adjusting stakes based on current bankroll. My system obviously generates far more losers than winners (hence the slim potential edge) so I worry that when a winner comes up the inevitable previous string of losers will have supressed the stake so much as to negate the win somewhat.

Of course the main issue is that in the unlikely event that I actually have a system that can reliably make $65 a year, I'm going to have to turn $1 bets into multi-100 dollar bets to have a useful "income". I'm not sure I have the nerve for this and (in the UK, at least) bookmakers don't tolerate steady winning of significant amounts and close accounts when it happens.

Quote:MrPogle2Thanks for this.

Stake management is something I need to investigate once I am convinced I actually have an edge. Obviously, no staking plan can overcome a house edge so I'm trying to establish one thing at a time but having looked at Kelly I'm not sure it will be possible to apply it.

Kelly seems to rely on knowing both offered odds and estimated probability of winning and the only estimate that I have is 103.7%. Does it make sense to apply this blanket value to every bet? Kelly seems to suggest that you need to evaluate this on a bet by bet basis.

There is also something else that bothers me a bit about adjusting stakes based on current bankroll. My system obviously generates far more losers than winners (hence the slim potential edge) so I worry that when a winner comes up the inevitable previous string of losers will have supressed the stake so much as to negate the win somewhat.

Of course the main issue is that in the unlikely event that I actually have a system that can reliably make $65 a year, I'm going to have to turn $1 bets into multi-100 dollar bets to have a useful "income". I'm not sure I have the nerve for this and (in the UK, at least) bookmakers don't tolerate steady winning of significant amounts and close accounts when it happens.

link to original post

How many days did it take for your system to find the 1720 bets you made?

Quote:MrPogle2331 days. 45 of those days identified no horses for the day.

link to original post

Ok. Now I got it. At a 3.8% +EV with only around 5 bets a day available, if you can make $1000 bet on each, you will make less than $200 a day. If you can still keep your regular job and this just adds to your income it’s great.

IF you are bankrolled enough for possible losing streaks.

If my $1000 is too high to be tolerated by the books, tell me what you think won’t get you blocked. I think if you are just making 3.8% there will be enough losing bets to provide cover, but I really don’t know how closely they look at horse bettors.

Quote:SOOPOO

If my $1000 is too high to be tolerated by the books,

link to original post

in horse race betting - parimutuel wagering - which means "against ourselves" - the size of the bets is not restricted in anyway

at least not at the racetracks and the online racebooks

the losers pay the winners after the takeout is applied - if a person makes a very large bet it just means all of the other winners will be paid less

it's possible that a casino sports book would accept horse racing bets and not place them into the pools - just pay whatever is indicated - and they might restrict the size of the bets

but I don't really know about that - anyway - anybody could simply open an online racebook account and bet however much they wanted to

.

However, last year I was excluded by a huge on-line casino for a trivial amount of money by simply being very, very lucky 3 weeks in a row. It rather implies that they are very vigilant in monitoring this stuff.

Quote:MrPogle2Of course the main issue is that in the unlikely event that I actually have a system that can reliably make $65 a year, I'm going to have to turn $1 bets into multi-100 dollar bets to have a useful "income". I'm not sure I have the nerve for this and (in the UK, at least) bookmakers don't tolerate steady winning of significant amounts and close accounts when it happens.

link to original post

You're in the UK? Snap!

Let's do stuff in £££££s

$:o)

You just need to spread your action across bookies, online and in shop.

Meanwhile, I think you just had good variance.

pop your wagers and payouts into excel so a column becomes 'wager profit' then use the standard deviation formula on that column.

multiply the stdev value by number of wagers to show your profit compared to the std dev of your profit.

You'll end up with something like

profit =£67 +/- stdev of £150

I.e your 67 profit sits in a range of expected values of maybe -83 to 217

Anyhow..... back to money management betting a fixed, modest percentage of rolling bankroll is the only way to let any edge work. If you flat bet, the limit to BR growth is suppressed. you need to embrace the variance or edge, whichever it is.

Quote:lilredrooster

in horse race betting - parimutuel wagering - which means "against ourselves" - the size of the bets is not restricted in anyway

at least not at the racetracks and the online racebooks

In the UK this is called TOTE betting and I suspect that using this system I wouldn't be able to turn a profit. When I place bets I choose the best odds I can get from about a dozen on-line bookmakers. They all offer a deal where if the SP is better than the odds I have taken, I get the SP instead. The TOTE gives me a price that is calculated based on the whole betting pool so is undeniably "fair". However there is no chance of getting a "better than fair" price, which I often do. I expect that this would wipe out my slim edge.

Quote:OnceDear

pop your wagers and payouts into excel so a column becomes 'wager profit' then use the standard deviation formula on that column.

multiply the stdev value by number of wagers to show your profit compared to the std dev of your profit.

I'll give this a go tomorrow.

A quick question.

Given I have made $65 from $1702 but the odds I was being given only offered me a likely return of 90% of $1702. Is my "profit" $65 or $235? After all, if my system was just random horses I would have returned $1532 and I have done $235 better than that.

Quote:MrPogle2Quote:OnceDear

pop your wagers and payouts into excel so a column becomes 'wager profit' then use the standard deviation formula on that column.

multiply the stdev value by number of wagers to show your profit compared to the std dev of your profit.

I'll give this a go tomorrow.

A quick question.

Given I have made $65 from $1702 but the odds I was being given only offered me a likely return of 90% of $1702. Is my "profit" $65 or $235? After all, if my system was just random horses I would have returned $1532 and I have done $235 better than that.

link to original post

Great question. Here comes a great answer.

Your profit is $65. That is unequivocal.

Your skill at picking winners is $235. Which is QUITE good, by the way!

It’s analogous to this question…..I can pick the correct team in a 50-50 game 51% of the time. Have I beaten the bookies? The answer is no, as the normal ‘vig’ needs around 53% to show a profit.

Quote:MrPogle2Quote:OnceDear

pop your wagers and payouts into excel so a column becomes 'wager profit' then use the standard deviation formula on that column.

multiply the stdev value by number of wagers to show your profit compared to the std dev of your profit.

I'll give this a go tomorrow.

A quick question.

Given I have made $65 from $1702 but the odds I was being given only offered me a likely return of 90% of $1702. Is my "profit" $65 or $235? After all, if my system was just random horses I would have returned $1532 and I have done $235 better than that.

link to original post

I've not got my stats head on, but your profit is $65 ( or £65) and in the absence of any skill your expected profit was -170.

Because of variance, that -170 would have been in a range. Without doing any maths, the expectation might have been -170 +/- 300 and maybe you've just naturally had a good result well within expected range.

It sounds like you have a fun way to wager which MIGHT have an edge. If I were you, I'd ramp up the wagering a little and carry on. Whatever you do, don't take any winning results as a confirmation of your prowess. Just enjoy it while it lasts.

If you lose, shrug, but don't chase losses.

I have wished for a VERY long time that the horsebook would FINALLY offer fixed odds on the races. (I think they do in Australia?) It's just about impossible to find this in the USA (unless I'm missing something).

There is nothing more annoying then placing a bet, and then find the odds actually changing ONCE THE RACE HAS STARTED!!!! And then to see if you would of followed the "new" odds, you would of won.

The track seems to be in a bit of a slump here, and I believe this is why.

Like the cockroach said to the dinosaur, "adapt or die".

You can download a simple p-value calculator here: https://www.football-data.co.uk/

Quote:audionutVery impressive, MrPogle! Are you using a bet exchange, or a regular book???

I have wished for a VERY long time that the horsebook would FINALLY offer fixed odds on the races.

I'm using a regular book but the "best odds guarantee", which is readily available in the UK, is pretty valuable. You can often see odds double/halve and you get the better price regardless.

Quote:blackjackladBased on 1720 bets with a 3.8% yield and average odds of 5.0 there is a 78% chance you have an edge, and a 22% chance what you're seeing is just randomness.

link to original post

Thanks for this. That is just what I was looking for. It's nice to know that I probably have an edge even if it's too small to replace the day job.

If I set my yield to zero I get a value of 0.5 which (I think) means expected value. However, if I can break even at the bookies I am overcoming their edge so I'm doing better than expected value. So, in terms of generating a p-value should my yield be my actual profit plus the house edge?

Quote:MrPogle2I've had a brief go at playing with the p-value calculator and I have a question which is the same as a previous one.

If I set my yield to zero I get a value of 0.5 which (I think) means expected value. However, if I can break even at the bookies I am overcoming their edge so I'm doing better than expected value. So, in terms of generating a p-value should my yield be my actual profit plus the house edge?

link to original post

It depends what your goal is. If it is to prove you have a system that will make you money, you need to only include your profit.

If it is to prove you are better than someone making random picks, you would include profit plus house edge.

The p value isn't an indication of your expected value, it's an indication of whether your results are caused by the randomness in a too-small sample size or not. Over 1000 bets might seem a lot, but depending on the average odds and the yield of the results, even 10000 bets might not be enough to say with confidence whether or not you have an ongoing edge.

I would consider a p-value higher than 0.05 to still be in the territory of an opportunity I wouldn't risk any serious money on.

If you're making money from online bookies in the UK and your results are inclusive of BOG then even if you have found an edge you will have to get creative sooner or later, due to the fact that the bookies will limit or close your account. Bricks and mortar bookies tend to have less generous odds, but depending on what your angle is you might still be able to turn a profit from them and spread yourself across lots of shops in order to maintain anonymity and avoid them accurately profiling you.

Exchanges are great because they have the best odds. As to whether the long-term value of exchange odds is better than the long-term value of bookies odds with BOG included, I'm not sure.

I would think that there is historical data available that can be used to simulate or model this.Quote:MrPogle2... So the data gathering is very slow with no prospect of modelling or simulation...link to original post

I also image that the data might show the opening, closing, high and low odds, while your history probably rarely includes making bets at any one of those values, which would mean any model/simulation would not produce reliable results.

Of course I'd bet that the cost of getting that data makes this whole project not worth the effort.

Quote:blackjacklad

Exchanges are great because they have the best odds. As to whether the long-term value of exchange odds is better than the long-term value of bookies odds with BOG included, I'm not sure.

link to original post

Me neither, however I have to pick one or the other. In terms of gathering data, results from an exchange are not repeatable and require a lot of effort from me. Is now a good time to bet? Are the odds likely to lengthen or shorten in the next hour? Does the current price represent value?

As I don't know the answer to any of those questions it just adds more variables to my fundamental question of whether or not I actually have an edge. Over a great many bets the variance created by the exchange probably averages out but I don't have that many bets. Anecdotally, I would say that exchange prices are better than SP at race-start but worse than those available at 10am. At some point between 10am and race-start, a better price than BOG is probably available on the exchange but I can't spend all day waiting for it and would only recognise it in hindsight anyway.

As things stand, at approx. the same time each morning I generate my picks and take the best odds available from about 10 bookies who offer BOG. This gives me the most repeatable strategy for the least effort. If I had a much bigger profit margin this might not matter but, as my margin is wafer thin, I suspect that consistently getting fair-to-good odds does.

Quote:MrPogle2Quote:blackjacklad

Exchanges are great because they have the best odds. As to whether the long-term value of exchange odds is better than the long-term value of bookies odds with BOG included, I'm not sure.

link to original post

Me neither, however I have to pick one or the other. In terms of gathering data, results from an exchange are not repeatable and require a lot of effort from me. Is now a good time to bet? Are the odds likely to lengthen or shorten in the next hour? Does the current price represent value?

As I don't know the answer to any of those questions it just adds more variables to my fundamental question of whether or not I actually have an edge. Over a great many bets the variance created by the exchange probably averages out but I don't have that many bets. Anecdotally, I would say that exchange prices are better than SP at race-start but worse than those available at 10am. At some point between 10am and race-start, a better price than BOG is probably available on the exchange but I can't spend all day waiting for it and would only recognise it in hindsight anyway.

As things stand, at approx. the same time each morning I generate my picks and take the best odds available from about 10 bookies who offer BOG. This gives me the most repeatable strategy for the least effort. If I had a much bigger profit margin this might not matter but, as my margin is wafer thin, I suspect that consistently getting fair-to-good odds does.

link to original post

Great post. I hope you realize that the overriding question you are trying to answer is the last sentence of your first paragraph.

DOES THE CURRENT PRICE REPRESENT VALUE?

If you believe it does, AND can quantify that value (e.g., I have a 10% edge on this bet), then you can use Kelly to tell you how much to bet. You needn’t worry about trying to guess which way the line MIGHT move if you have value NOW.

Of course it’s smart to shop for the best line; why wouldn’t you if it is easily available to do?

Also worth noting that you can download historical data from betfair free of charge. As it's the most liquid exchange it's the best representation of the fair price at any given time (though it's not a completely efficient market), so it can be useful for backtesting ideas.

https://historicdata.betfair.com/#/home

I've yet to turn the downloadable data into something usable - annoyingly it doesn't come in a spread sheet format - but I will once I find the time.

In the late 1980s, fixed-odds football (soccer) betting took the form of a paper coupon that you picked up from the bookies and filled in with a biro. It contained about 60-70 matches for the Saturday and, because it was printed on Thursday, the odds couldn't change. The prices for each match were based on the relative quality of the two teams involved but highly unlikely outcomes had odds that never varied. They used to offer 66/1 on any match having a scoreline of 3-3 and for about a season and a half my cousin and I would bet £1 on every match resulting in this scoreline. Almost every week one match would oblige and some weeks 2 or 3 would. We were coining it in until one week the odds offered dropped from 66/1 to 50/1 and our system died overnight. It didn't help that we were turning in ridiculous looking coupons with every line filled in whereas most punters had about a dozen entries on theirs. Even in those days, there was enough feedback to plug leaks in the odds-setting mechanism. It wouldn't take them a season and a half nowadays.