Average win rates and long series of losses

[silentdiver]

Hello all, I'm writing to see if anyone might have advice about using average win rates for favourites as (very rough) predictors of how many losses in a row is likely.

I'm no expert in any of this, happy to be told I'm an idiot, but ideally in constructive way - I think I'm possibly not seeing something straight, and would appreciate any advice.

Say from historical data the win rate for 1st favourites in horse races is 1 in 3, in all conditions/locations/etc/etc - this is a very rough statistic ! - it seems to be more like 35% but the exact figures aren't important, it's the overall reality vs mathematical probability I'm asking about.

Say you were betting on the 1st favourite every time using a Martingale type strategy of increasing the stake after each loss, and so hoping for a win every now and then to make back the losses so far - one would expect, over a zillion bets, that the first bet would win about 33.33% of the time, and lose 66.66% of the time.

So the chance of two losses in a row would be about 66% of 66%, or 0.66 squared, so about 44%

So... the chance of 10 losses in a row would be 0.66 to the power of 10, so just under 2%, with the chances of 20 losses in a row being vanishingly small.

I chose the win rate 1 in 3 as it's close enough to the reality and is the same chance as rolling a 5 or 6 on a perfect 6-sided dice; and I'd expect the maths above to be pretty much reflected in the results if I spent all day rolling that perfect dice.

The thing is, from analysing actual horse racing results, it's not all that uncommon for the first favourite in a mix of races to lose 20 or even 30 times in a row; which from the probabilities above should be extraordinarily unlikely. This is taking results from all around the world merged together, just in race start time order, no filtering on anything.

If I was spending a day rolling a dice, I'd be really surprised if I didn't roll a 5 or 6 once in 20 or even 30 rolls, and statistically, surely I'd be correct to be surprised. Same as if I played Roulette all day and always bet on the 1-12 range, I'd think myself very, very unlucky if I lost 20+ times in a row (and more so if the table didn't have a zero).

So... thanks for reading thus far, my question is where are my calculations and reality diverging or; how come horses don't obey the laws of probability ? - or a slightly (only slightly, I stress) less naive version of the question, how come I see such long series of losses of 1st favourites when the straight maths suggests there should be more wins interspersed ?

If I could remember more of the maths I did at school I might try to work out how many standard deviations from the expected values my observations represent but that was over 40 years ago !

[jamesedwards]

Hello all, I'm writing to see if anyone might have advice about using average win rates for favourites as (very rough) predictors of how many losses in a row is likely.

I'm no expert in any of this, happy to be told I'm an idiot, but ideally in constructive way - I think I'm possibly not seeing something straight, and would appreciate any advice.

Say from historical data the win rate for 1st favourites in horse races is 1 in 3, in all conditions/locations/etc/etc - this is a very rough statistic ! - it seems to be more like 35% but the exact figures aren't important, it's the overall reality vs mathematical probability I'm asking about.

Say you were betting on the 1st favourite every time using a Martingale type strategy of increasing the stake after each loss, and so hoping for a win every now and then to make back the losses so far - one would expect, over a zillion bets, that the first bet would win about 33.33% of the time, and lose 66.66% of the time.

So the chance of two losses in a row would be about 66% of 66%, or 0.66 squared, so about 44%

So... the chance of 10 losses in a row would be 0.66 to the power of 10, so just under 2%, with the chances of 20 losses in a row being vanishingly small.

I chose the win rate 1 in 3 as it's close enough to the reality and is the same chance as rolling a 5 or 6 on a perfect 6-sided dice; and I'd expect the maths above to be pretty much reflected in the results if I spent all day rolling that perfect dice.

The thing is, from analysing actual horse racing results, it's not all that uncommon for the first favourite in a mix of races to lose 20 or even 30 times in a row; which from the probabilities above should be extraordinarily unlikely. This is taking results from all around the world merged together, just in race start time order, no filtering on anything.

If I was spending a day rolling a dice, I'd be really surprised if I didn't roll a 5 or 6 once in 20 or even 30 rolls, and statistically, surely I'd be correct to be surprised. Same as if I played Roulette all day and always bet on the 1-12 range, I'd think myself very, very unlucky if I lost 20+ times in a row (and more so if the table didn't have a zero).

So... thanks for reading thus far, my question is where are my calculations and reality diverging or; how come horses don't obey the laws of probability ? - or a slightly (only slightly, I stress) less naive version of the question, how come I see such long series of losses of 1st favourites when the straight maths suggests there should be more wins interspersed ?

If I could remember more of the maths I did at school I might try to work out how many standard deviations from the expected values my observations represent but that was over 40 years ago !

On Betfair there's probably about 50,000 horse races covered per year. That's 50,000 chances of a race kicking off a long sequence of losing favs. So even assuming your 66% logic is correct, you would expect to come across 20 losing favs in a row around once per month.

The 66% assumption can also be challenged because you're not allowing for a mixed distribution of fav prices and there will be days when the average fav odds will be much higher than others. eg on a Saturday you'll likely find much bigger fields and more handicaps resulting in higher average fav prices.

[Win]

It also goes the other way. If the fav wins 33% of the time, then it loses 67% of the time so if you lay the fav you would expect to profit 2/3rds of the time.
But on occasions the fav will win 10 races in a row.

[Chocorange]

the excellent adrianmassey.com website allows you to create a custom report for this and using many other variables, literally limited only by your imagination.

It has a database of every Uk and Irish horse race back to 1st Jan 2012/

On the front page , if you click on the top option 'fate of the favourites' then on the next page scroll to the bottom option 'create custom report' you can then interrogate all the data.... anything from type of race to weight carried to days since last run, new trainers , literally pretty much anything.

for the report you are interested in, i selected the fav in each race unchecked all other items to do with the fav , then left everything else checked and it generated a report that said if backing favs then there have been the below

Sequences of Consecutive Successful Win Selections
1 Winner 24351
2 Winners 8666
3 Winners 3011
4 Winners 1060
5 Winners 375
6 Winners 126
7 Winners 47
8 Winners 23
9 Winners 5
10 Winners 3
11 Winners 0
12 Winners 2
13 Winners 1
14 Winners 0
15 Winners 0
16 Winners 0
17 Winners 0
18 Winners 0
19 Winners 0
20 Winners 0
21 Winners 0
22 Winners 0
23 Winners 0
24 Winners 0
25 Winners 0
26 Winners 0
27 Winners 0
28 Winners 0
29 Winners 0
30+ Winners 0

so the longest run of consecutive winners in the last 14 or so years is 13.... however if you look at the report for consecutive losers, it shows the below.

Sequences of Consecutive Unsuccessful Win Selections
1 Loser 13252
2 Losers 8643
3 Losers 5474
4 Losers 3604
5 Losers 2382
6 Losers 1556
7 Losers 933
8 Losers 618
9 Losers 422
10 Losers 290
11 Losers 181
12 Losers 100
13 Losers 76
14 Losers 41
15 Losers 38
16 Losers 26
17 Losers 8
18 Losers 7
19 Losers 5
20 Losers 4
21 Losers 5
22 Losers 1
23 Losers 1
24 Losers 0
25 Losers 2
26 Losers 0
27 Losers 0
28 Losers 1
29 Losers 0
30+ Losers 0

So there have been 14 occasions of 20 or more losers , with the longest run topping out at 28.

The past is no predictor of the future, but this may answer the question you asked.

[Revenant]

Start chucking in joint favourites, delayed races the fact that the Betfair favourite won't necessarily be the ISP fav and you're looking at a logistical nightmare with Martingale style betting

[jamesedwards]

Start chucking in joint favourites, delayed races the fact that the Betfair favourite won't necessarily be the ISP fav and you're looking at a logistical nightmare with Martingale style betting

You don't need any of that to make Martingale style betting a nightmare.

It's such a dangerous strategy that the forum has a whole section dedicated to it. viewforum.php?f=77